# Welcome AutoReach finds your ideal buyers on X/Twitter, LinkedIn, and Instagram, scores them with AI, and runs personalized outreach on autopilot so you can focus on closing deals. *** ## Jump right in * [**Quickstart**](/getting-started/quickstart) - Connect your accounts, create your first offer, and launch outreach in minutes. * [**Core Concepts**](/core-concepts/overview) - Understand how offers, leads, buyer scoring, and sequences work together. * [**Autopilot**](/autopilot/overview) - One click to activate fully automated discovery, engagement, and responses. *** ## How it works 1. **Discover** - Find leads from tweets, LinkedIn posts, job listings, companies, and lookalike audiences that match your ideal customer profile. 2. **Enrich** - Pull in profile data, company info, emails, websites, and recent activity for every lead. 3. **Score** - AI scores each lead on fit, intent, and timing to surface who is ready to buy right now. 4. **Engage** - Send personalized DMs, emails, connection requests, likes, comments, follows, and story views automatically across X, LinkedIn, Instagram, and your inbox. 5. **Respond** - When leads reply on any channel, AI generates stage-aware responses that keep the conversation moving toward a meeting. 6. **Resurface** - Leads that are not ready today get monitored. When new buying signals appear, they come back to the top. 7. **Book** - Qualified conversations flow into meeting bookings with auto-generated call briefs. *** ## Explore the docs | Section | What you will find | | ------------------------ | -------------------------------------------------------------------------------------------------------- | | **Getting Started** | Connect your social and email accounts, create your first offer, and launch your first campaign | | **Core Concepts** | How leads, buyer scoring, offers, and the knowledge base work | | **Finding Leads** | Tweet search, LinkedIn people / company / job search, Instagram discovery, lookalikes, and the lead pool | | **Enrichment** | Profile data, emails, websites, and company info for every lead | | **Outreach & Sequences** | Building sequences, DM and email generation, scheduling, and A/B testing | | **AI & Conversations** | The unified inbox, AI responses, tone examples, and conversation optimization | | **Autopilot** | One-click automation: what it does, how to enable it, and how to tune it | | **Engagement Engine** | Content strategy, daily engagement, and account health | | **AI Video & Reels** | Generate short-form vertical videos from a topic and your offer | | **Meetings & CRM** | Booking meetings, call briefs, and the Chrome extension pipeline | | **Settings** | AI model config, multi-account management, costs, and account safety | # FAQ Answers to the most common questions about AutoReach. For detailed guides, follow the links to the relevant documentation pages. *** ## Getting Started ### How do I get started with AutoReach? Sign up at [autoreach.tech](https://autoreach.tech), pick a proxy location, install the Chrome Extension and connect at least one social account, create an Offer, and start Autopilot. AI runs on the free credits included with your plan, so there are no API keys to set up. The onboarding wizard walks you through each step. See the [Quickstart](/getting-started/quickstart) for a full walkthrough. ### Do I need both X and LinkedIn accounts connected? No. You need at least one social account connected to use AutoReach. You can connect X only, LinkedIn only, or both. Connecting both platforms gives you cross-platform profile matching and more outreach options. ### How does AI usage get billed? By default, AI runs on credits included with your plan. You start with free credits and can top up anytime from the **Credits** page, with nothing to configure. If you would rather use your own provider, add an OpenAI, Anthropic, or DeepSeek key in **Settings > AI & Models** for unlimited usage billed directly by your provider with no markup. Either way, your actual usage depends on how many leads you process and which features you run. ### Can I use AutoReach without Autopilot? Yes. Autopilot is a convenience feature that automates initial setup. You can manually create searches, build sequences, score leads, and manage everything yourself. Autopilot simply handles the initial configuration so you can get started faster. *** ## Finding Leads ### Where do leads come from? AutoReach discovers leads from multiple sources: X tweet search, LinkedIn content search, LinkedIn people search, LinkedIn company search (By Company), LinkedIn job search, lookalike audience discovery, follower extraction, comment extraction, the Lead Pool, CSV import, manual add, and the Chrome Extension. See [Finding Leads](/finding-leads/overview) for a breakdown of each method. ### Can I search by company on LinkedIn? Yes. Use **By Company** when adding a LinkedIn lead source. Enter descriptors like "B2B SaaS Sales Tooling" or "Lead Generation Agency" (or generate them from your offer with AI), and AutoReach finds matching companies and adds the top decision-maker from each. See [LinkedIn Company Search](/finding-leads/linkedin-company-search). ### How many leads should I target at once? Quality matters more than quantity. Start with 100-200 leads from intent-driven sources (tweet search or LinkedIn content search) rather than thousands from bulk imports. Leads found through keyword searches carry inherent intent signals and score higher than cold imports that need full enrichment from scratch. ### Can I import leads from a CSV file? Yes. Go to the Leads page and use the CSV Import feature. You can import leads with X handles, LinkedIn URLs, or email addresses. Imported leads are queued for full enrichment and scoring. Duplicate detection prevents the same lead from being added twice. ### What is the Lead Pool? The Lead Pool is a shared database of pre-enriched lead profiles stored as vector embeddings. When you create an offer, AutoReach uses semantic similarity to instantly match existing profiles against your ICP. This is the fastest way to get scored leads because they skip enrichment entirely. See [Lead Pool](/finding-leads/lead-pool) for details. ### Can I export my leads to CSV? Yes. You can export leads from the All Leads page or the Buyers page. Use the export controls to download your lead data as a CSV file. *** ## Scoring and Buyers ### How does buyer scoring work? AutoReach scores each lead across three dimensions: Fit (does this person match your ICP?), Intent (are they actively looking for a solution?), and Timing (is this the right moment to reach them?). These combine into a single Buyer Score from 0-100 that determines the lead's buyer state. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence) for the full breakdown. ### Why is a lead scored as Poor Fit? A lead lands in Poor Fit when their Buyer Score is low. This usually means they have a significant mismatch with your ICP, such as wrong industry, wrong seniority, or wrong company size. Check the scoring breakdown to see which dimension scored lowest. Poor Fit leads are still monitored and can be promoted if new signals appear, like a job change or new social activity. See [Buyer Intelligence](/core-concepts/buyer-intelligence) for details. ### Can a lead's score change over time? Yes. Scores are not static. They update when new signals are detected (a relevant post, a job change, company funding), when you update your offer definition, or when you trigger a manual rescore. The resurfacing scheduler automatically rechecks Monitor and Poor Fit leads on a regular cycle. See [Continuous Operations](/autopilot/continuous-operations). ### What is the difference between Active and Monitor buyer states? Active leads are ready for outreach and eligible for automatic enrollment into sequences. Monitor leads have potential but are not quite ready - perhaps strong fit but weak intent. Monitor leads are automatically rechecked on a regular schedule and promoted to Active if their score improves. See [Buyer Intelligence](/core-concepts/buyer-intelligence). *** ## Sequences and Outreach ### How do sequences work? A sequence is a multi-step outreach campaign that you build visually in the flow editor. You define actions (like a post, follow, send a DM, send a connection request), set delays between steps, and add conditions that branch based on lead behavior. Once started, AutoReach executes the steps automatically for each enrolled lead. See [Building Sequences](/outreach-and-sequences/building-sequences). ### What happens if I run out of daily send limits? When a daily limit is reached, remaining actions are automatically rescheduled to the next day at a random time within your activity window. Leads stay in the sequence and no actions are lost. Limits reset at midnight in your timezone. See [Scheduling & Send Limits](/outreach-and-sequences/scheduling). ### How do I pause a sequence? Pausing temporarily halts all pending actions without losing progress. You can resume a paused sequence and it picks up where it left off. To run the same campaign against a different audience, duplicate the sequence and start the copy. See [Building Sequences](/outreach-and-sequences/building-sequences). ### How do I duplicate a sequence? Open the sequence you want to copy, then use the duplicate option from the sequence menu. This creates a new draft sequence with the same steps, templates, prompts, and settings. You can modify the copy before starting it. Tone examples and enrolled leads are not copied to the duplicate. ### Can I run multiple sequences at the same time? Yes. You can have multiple active sequences running simultaneously, even on the same platform. Each sequence has its own daily limits, enrolled leads, and settings. Be mindful of total daily activity across all sequences to keep your accounts safe. *** ## AI and Messaging ### How does AI personalize my DMs? When a DM step executes, AutoReach replaces template variables (like first name, company, role) with real lead data, then passes any remaining placeholders to the AI along with the lead's profile, recent posts, your offer context, and knowledge base content. The AI fills in context-specific details and ensures the message sounds natural. See [Cold DM Generation](/outreach-and-sequences/dm-personalization). ### How do I customize the tone of AI messages? Add tone examples to your sequence. These are conversation samples that teach the AI how you sound. The AI retrieves the most relevant examples at response time via semantic search. You can also adjust the AI prompt and DM generation prompt in your sequence settings. AutoReach auto-captures winning conversation patterns from meetings you book. See [Tone Examples](/ai-and-conversations/tone-and-knowledge). ### What does "max AI responses = 0" mean? A value of 0 means unlimited. The AI will continue responding to a conversation for as long as it remains active, with no cap on the number of replies. Set it to a positive number (like 3 or 5) to limit how many times the AI auto-replies in a single conversation before stopping. See [Scheduling & Send Limits](/outreach-and-sequences/scheduling). ### What's a healthy reply rate? A reply rate above 10% on cold DMs indicates good targeting and messaging. Below 5% usually means your ICP is too broad or your message needs work. The 5-10% range is typical for cold outreach in most B2B niches. If you're stuck below 5%, the [Low Reply Rates](/troubleshooting#low-reply-rates) section walks through the diagnostic loop. *** ## Email ### Can I send emails through AutoReach? Yes. Open a parent X or LinkedIn account from the **Accounts** page, connect Gmail or Outlook from the **Email** card, then add an email step to any sequence. Subject and body support the same template variables and AI personalization that DMs use. Replies appear in the unified Inbox alongside X and LinkedIn conversations. See [Email Channel](/outreach-and-sequences/email-channel). ### How do I connect Gmail? Gmail uses an App Password rather than OAuth. Enable 2-Step Verification on your Google account, generate an App Password named "AutoReach", and paste it into the connect dialog along with your Gmail address. Workspace admins may need to enable IMAP and App Passwords. See [Connecting Email](/getting-started/connecting-email). ### How do I connect Outlook? Outlook uses Microsoft OAuth. Click **Connect Outlook**, sign in through the popup, and grant permissions. AutoReach refreshes tokens automatically. ### Can I connect both Gmail and Outlook? Yes. When both are connected, AutoReach routes outbound mail by recipient domain (Gmail recipients use Gmail, Outlook recipients use Outlook, everyone else uses your **Primary** mailbox). This improves deliverability and authenticity. ### Are bounces handled automatically? Yes. Bounces are detected on inbound mail and the lead is flagged. Future sends to a bounced address are skipped automatically. ### Do email actions count toward my daily limit? Yes. Email sends count toward the same per-sequence daily action limit that covers DMs, likes, comments, follows, and connection requests. On top of that, each mailbox is rate-paced at one send at a time and up to 10 per minute to keep delivery clean. *** ## Conversations ### How does the AI know what stage a conversation is in? AutoReach classifies conversations into 7 stages: Opener Reply, Discovery, Value Prop, Objection Handling, Soft Close, Follow Up, and Exit. Stage detection uses a combination of heuristics (message count, objection count) and AI classification based on conversation history. The AI adapts its tone and goals for each stage. See [Conversation Stages](/ai-and-conversations/ai-response-engine). ### Can I take over a conversation from the AI? Yes. Toggle AI off for any individual conversation using the AI ON/OFF switch in the Inbox. When AI is off, only your manual messages are sent. Sending a manual message also auto-cancels any pending AI response. You can toggle AI back on at any time. See [Manual Controls](/ai-and-conversations/inbox). ### How do follow-ups work when a lead goes silent? Enable conversation follow-ups in your sequence settings. You configure how many days to wait before following up (default 3) and how many follow-ups to send (default 2). AutoReach periodically checks for stale conversations and generates a fresh-angle message to re-engage the lead. See [Manual Controls](/ai-and-conversations/inbox). *** ## Engagement Engine ### What does the Engagement Engine do? The Engagement Engine builds a credible social presence on your accounts by automatically posting content, engaging with relevant posts (likes, replies, comments), and growing your visibility. This makes your accounts look active and authentic before and during outreach. See [Engagement Engine Overview](/engagement-engine/overview). ### Do I need to approve everything the Engagement Engine posts? By default, yes. All generated content enters an approval queue where you can approve, reject, or edit it. You can enable auto-approval in your Engagement Engine settings to skip manual review and let content post automatically. Pending approvals expire if not reviewed within a set period. See [Approving Content](/engagement-engine/content-generation). *** ## Autopilot ### What does Autopilot actually automate? Autopilot runs five continuous operations: lookalike rotation (finding new seed accounts), auto-enrollment (adding scored leads to sequences), buyer expansion (daily discovery of new prospects), monitor resurfacing (rechecking lower-scored leads for new signals), and signal search (finding prospects posting about your offer topics). See [What Autopilot Does](/autopilot/overview). ### Is Autopilot safe for my accounts? Yes. Autopilot respects all the same safety measures as manual operation: daily send limits, activity windows, human behavior simulation, negative content screening, and gradual resumption. It does not bypass any account safety features. See [Account Safety](/settings-and-configuration/account-safety). *** ## Account Safety ### How does AutoReach avoid getting my account banned? AutoReach uses multiple layers of protection: consistent browser identity management per account, human behavior simulation (randomized delays, session clustering, typing simulation), time-of-day and day-of-week activity patterns, negative content screening, daily send limits, and automatic emergency pausing when issues are detected. See [Account Safety](/settings-and-configuration/account-safety). ### What should I do if my account gets paused? Check the Accounts page to see the error type and reason. Auto-recoverable errors (rate limits, timeouts) resolve after a cooldown period. Non-recoverable errors (bot detection, captcha) require manual intervention. After resolving the issue, manually resume the account. All associated sequences and warmup activities are paused together and need to be resumed. See [Account Safety](/settings-and-configuration/account-safety). *** ## Billing and Usage ### How much does AutoReach cost to run? AutoReach is **$99/mo** on a subscription that starts after a **7-day software trial**. A non-refundable **$20 activation fee is charged at signup** and includes starter AI credits and managed-proxy activation if selected. The subscription includes all platforms (LinkedIn, X, Instagram, and email) and every feature. You can top up AI credits as needed, with nothing to configure. If you prefer, you can bring your own OpenAI, Anthropic, or DeepSeek key for unlimited usage billed directly by your provider. The cost estimation tool in Settings shows credit estimates based on your configuration. A proxy is optional: managed-proxy activation is included in the signup fee and costs **$15/mo after the trial** if retained (see "Do I need a proxy?" below), or bring your own at no additional AutoReach charge. A Done-For-You option is also available: $299 setup plus $200 per booked meeting, where the AutoReach team sets everything up and runs it for you. See [Pipeline Cost Estimation](/settings-and-configuration/cost-estimation). ### How does the 7-day trial work? Signup starts a **7-day software trial**. A non-refundable **$20 activation fee is charged today**. If you cancel before the trial ends, the $99/month recurring subscription does not begin; the activation fee is not refunded because starter AI credits and onboarding infrastructure are made available immediately. Promotion codes are not accepted on signup Checkout. ### Do I need a proxy? A proxy is **optional**. You choose during onboarding between two options: * **Managed proxy** (IPRoyal ISP residential static, provided by AutoReach): activation is included in the $20 signup fee, then **$15/mo** per managed proxy after the trial if retained. One managed proxy can be shared across your LinkedIn, X, and Instagram accounts. * **Bring your own proxy (BYOP)**: free. Enter your host, port, username, password, and type (HTTP or SOCKS5). AutoReach verifies connectivity before saving. Proxies are managed on the [Proxies](/settings-and-configuration/multi-account#proxy-configuration) page, where you can add, edit, remove, and assign proxies to accounts. ### How many accounts can I connect? One subscription includes **one account slot per platform**: 1 LinkedIn, 1 X, and 1 Instagram. Offers and sequences are unlimited. Each social account can also attach up to 2 email mailboxes (one Gmail, one Outlook). To connect more than one account on the same platform, contact support at . There is no self-serve add-on for extra accounts. ### How do I reduce my AI costs? Three strategies: disable web enrichment for leads that don't need it (it's the most expensive per-lead operation), use lower-cost models for high-volume categories like classification and keyword generation in your AI Model settings, and choose efficient fallback models since fallback models also contribute to your overall cost. You can also run smaller, more targeted searches rather than broad ones to reduce the number of leads entering the pipeline. See [Pipeline Cost Estimation](/settings-and-configuration/cost-estimation) and [AI Model Configuration](/settings-and-configuration/ai-models). ### How do I set up webhooks for meeting tracking? Calendar configuration is **per-account** - open an account from the Accounts page and use the Calendar section. Add your booking URL, then in your booking platform add the custom invitee question shown in the form (this is what AutoReach uses to match bookings back to a lead). For Calendly, paste a Personal Access Token, click Auto-fetch to retrieve your Organization URI, then click Register webhook - AutoReach creates the webhook for you (Calendly paid plan required). For Cal.com, copy the webhook URL AutoReach generates and paste it into Cal.com. See [Meeting Booking](/meetings-and-crm/meetings). # Glossary Key terms used throughout AutoReach, listed alphabetically. *** **Active**- A buyer state assigned to leads with a high Buyer Score. Active leads are considered ready for outreach and are eligible for automatic enrollment into sequences. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **Activity Window**- The time range during which AutoReach executes outreach actions (e.g., 9:00 AM to 9:00 PM). Actions scheduled outside this window are deferred until it reopens. Configured **per-account** on each connected LinkedIn/X/Instagram account from the Accounts page → Configuration tab → Activity Window. Falls back to the user-level window if not customized. See [Scheduling & Send Limits](/outreach-and-sequences/scheduling). **API Calls Today**- A counter shown on each account card in the Accounts page, displayed as `used / total`. It tracks how much of the account's daily activity allowance has been consumed by the actions AutoReach performs through that account (DMs, connection requests, profile enrichment, engagement, and searches). When the allowance is used up, AutoReach pauses activity on the account until the counter resets the next day, keeping daily activity at a safe level. See [Multi-Account Management](/settings-and-configuration/multi-account#account-health). **Auto-Enrollment**- Auto-enrollment periodically adds newly scored Active leads into a running sequence. See [Continuous Operations](/autopilot/continuous-operations). **Autopilot**- AutoReach's one-click automation system that configures searches, creates sequences, enrolls leads, and runs five continuous background operations (lookalike rotation, auto-enrollment, buyer expansion, monitor resurfacing, and signal search). See [Autopilot Overview](/autopilot/overview). **Blacklist**- A list of accounts that AutoReach will never discover, enrich, or engage with. Adding an account to the blacklist immediately cancels all pending actions and removes the account from all sequences. See [Blacklisting](/settings-and-configuration/blacklisting). **Booking Link**- Your calendar scheduling URL (Calendly, Cal.com, or custom) that AutoReach includes in outreach messages via the `{{booking_link}}` template variable. When webhook integration is configured, AutoReach automatically detects when meetings are booked. See [Meeting Booking](/meetings-and-crm/meetings). **Buyer Expansion**- An Autopilot operation that runs daily to discover new prospects from your existing pipeline. On X, it re-extracts followers from seed accounts. On LinkedIn, it rotates through role-based people searches. See [Continuous Operations](/autopilot/continuous-operations). **Buyer Intelligence**- AutoReach's AI-powered scoring system that evaluates each lead across three dimensions (fit, intent, and timing) to predict purchase probability. Also referred to as the scoring engine. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence). **Buyer Score**- A composite score from 0-100 that combines fit, intent, and timing into a single number. The Buyer Score determines a lead's buyer state and pipeline position. The balance between dimensions adapts automatically based on your market. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence). **Buyer State**- One of six statuses assigned to every lead: Active, Monitor, Poor Fit, Disqualified, Manual Outreach, or Not Scored. The state determines where a lead appears in the UI and whether it is eligible for outreach. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **Call Brief**- An AI-generated preparation document for an upcoming meeting with a lead. It pulls data from the lead's profile, conversation history, and buyer intelligence to give you talking points. See [Meetings](/meetings-and-crm/meetings). **Chrome Extension**- A browser extension required to connect your X and LinkedIn accounts to AutoReach. It extracts session cookies from your logged-in browser sessions. On both LinkedIn and X, it also provides an "Add to Leads" button on profile pages, a full CRM Kanban pipeline, and an AI messaging assistant (first message, reply suggestion, follow-up, proofread, lead analysis, SDR co-pilot) in DM and reply composers. See [Installing the Chrome Extension](/getting-started/chrome-extension). **Cold DM**- The first direct message sent to a lead who has not previously interacted with you. AutoReach generates personalized cold DMs using AI, your offer context, lead profile data, and knowledge base content. See [DM Personalization](/outreach-and-sequences/dm-personalization). **Condition Step**- A sequence step that branches the flow based on lead behavior (replied, followed back, connection accepted, or has profile). Condition steps have two outgoing branches (true and false) and can retry on a configurable interval before falling to the false branch. See [Building Sequences](/outreach-and-sequences/building-sequences). **Connection Request**- A LinkedIn-specific sequence action that sends a connection request to a lead, optionally with a personalized note (up to 200 characters). Includes configurable auto-withdraw if not accepted within a set number of days. See [Building Sequences](/outreach-and-sequences/building-sequences). **Conversation Analyzer**- An AI system that reviews your outreach conversations, compares positive outcomes (meetings booked) against negative ones (lost), and suggests improvements to tone examples and prompts. See [Offers & Knowledge Base](/core-concepts/offers-and-knowledge-base). **Conversation Stage**- One of seven stages that classify where a conversation currently sits: Opener Reply, Discovery, Value Prop, Objection Handling, Soft Close, Follow Up, or Graceful Exit. The AI adapts its response style based on the detected stage. See [AI Response Engine](/ai-and-conversations/ai-response-engine). **Daily Limit**- The maximum number of outreach actions AutoReach will execute per sequence per day. A single unified limit covers all action types across all channels (likes, comments, follows, DMs, connection requests, and emails). When the limit is reached, remaining actions are rescheduled to the next day. See [Scheduling & Send Limits](/outreach-and-sequences/scheduling). **Disqualified**- A buyer state for leads with very low fit and buyer scores. Disqualified leads are excluded from automatic rescoring and resurfacing. They can be manually overridden to Manual Outreach if needed. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **DM Template**- A message template with `{{variable}}` placeholders that AutoReach fills with real lead data at send-time. Templates use a two-pass system: direct substitution for known variables, then AI inference for custom placeholders. See [DM Personalization](/outreach-and-sequences/dm-personalization). **Email Channel**- A native outreach channel that lets you send and receive emails through your connected Gmail and Outlook mailboxes. Email steps in sequences support the same template variables and AI personalization as DMs. Inbound replies, bounces, and auto-responses are handled automatically. See [Email Channel](/outreach-and-sequences/email-channel). **Email Step**- A sequence action that sends an email from a connected mailbox. Configurable with a subject template, body template, AI personalization toggle, and step delay. See [Email Channel](/outreach-and-sequences/email-channel). **Engagement Engine**- A system that builds social credibility on your accounts by automatically posting content, engaging with relevant posts (likes, replies, comments, follows), and maintaining human-like activity patterns. See [Engagement Engine Overview](/engagement-engine/overview). **Engagement Pod**- An opt-in LinkedIn-only feature (on by default) where AutoReach users automatically engage with each other's posts within a 60-minute window to boost algorithmic reach. Toggled per account from the Engagement Engine card. See [Engagement Engine Overview](/engagement-engine/overview#engagement-pod-linkedin-only). **Enrichment**- The automatic process of gathering additional data about a lead after discovery. Includes profile enrichment (bio, experience, skills), activity enrichment (recent posts), cross-platform matching, web enrichment, email finding, and website finding. See [The Enrichment Pipeline](/enrichment/pipeline). **Fit Score**- A score from 0-100 measuring how well a lead matches your ideal customer profile. Evaluates industry, company size, role/seniority, geographic match, and skill/background alignment. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence). **Follow-Up**- An automatic message sent when a lead goes silent after a conversation. Configurable per sequence with wait time (1-30 days) and max count (1-10). Follow-ups are generated with a fresh angle to re-engage the lead. See [Inbox](/ai-and-conversations/inbox). **Graceful Exit**- A conversation stage triggered when a lead declines or the conversation reaches a dead end. The AI exits respectfully and leaves the door open for the future. See [AI Response Engine](/ai-and-conversations/ai-response-engine). **Heat Score**- An account-level metric that aggregates recent signals across all leads at a company. Companies with multiple recent signals have a higher heat score, indicating the account is worth prioritizing. See [Account Signals](/settings-and-configuration/account-signals). **ICP (Ideal Customer Profile)**- A description of the type of person and company most likely to buy your product. Defined in your Offer's target audience field. Your ICP powers lead discovery, scoring, and personalization throughout AutoReach. See [Offers & Knowledge Base](/core-concepts/offers-and-knowledge-base). **Intent Score**- A score from 0-100 measuring whether a lead is actively looking for a solution. Evaluates signals like asking for recommendations, switching tools, complaining about current solutions, engaging with competitors, and mentioning relevant pain points. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence). **Intent Signal**- A data point indicating that a lead is interested in, evaluating, or planning to buy a solution like yours. Examples include asking for recommendations, mentioning competitor tools, complaining about current processes, or posting about relevant challenges. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **Instagram Lead Discovery**- Finding prospects on Instagram by mining hashtags, follower and following lists, and post engagers (likers and commenters), plus manual username adds. Discovered profiles flow into the same enrichment and scoring pipeline as X and LinkedIn leads. See [Instagram Lead Discovery](/finding-leads/instagram-search). **Knowledge Base**- A collection of documents (PDF, DOCX, TXT) uploaded to an Offer that AutoReach uses for AI responses. Documents are chunked, embedded as vectors, and retrieved via semantic search to provide context during conversations. Limited to 10 documents per offer, 10MB per file. See [Offers & Knowledge Base](/core-concepts/offers-and-knowledge-base). **Lead**- A unified profile of a potential buyer that combines data from X, LinkedIn, and Instagram. Leads are automatically discovered, enriched, scored, and tracked throughout the outreach lifecycle. See [How Leads Work](/core-concepts/leads). **Lead Pool**- A global database of pre-enriched lead profiles stored as vector embeddings. When you create an offer, AutoReach uses semantic similarity search to instantly match existing profiles against your ICP, delivering scored leads in seconds. See [Lead Pool](/finding-leads/lead-pool). **LinkedIn Company Search (By Company)**- A LinkedIn lead discovery method that finds companies matching your descriptors (e.g., "B2B SaaS Sales Tooling") and adds the top decision-maker from each as a lead. See [LinkedIn Company Search](/finding-leads/linkedin-company-search). **Lookalike Audience**- A set of leads discovered by finding influencers, thought leaders, or communities whose audiences match your ICP, then extracting their followers. AutoReach uses AI to find the right seed accounts automatically. See [Lookalike Audiences](/finding-leads/lookalike-audiences). **Manual Outreach**- A buyer state you set manually on any lead, regardless of their score. Treated as Active for enrollment eligibility. Use this for VIP accounts, strategic partnerships, or leads you want to reach despite low scores. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **Monitor**- A buyer state for leads with moderate Buyer Scores. Monitor leads have potential but are not yet ready for outreach. They are automatically rechecked on a regular schedule and promoted to Active if their score improves. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **Offer**- The foundation of AutoReach. An Offer describes what you sell, who you are targeting, and your outreach goals. It powers lead discovery, scoring, DM personalization, and AI responses. You can create multiple Offers, each with its own pipeline and settings. See [Offers & Knowledge Base](/core-concepts/offers-and-knowledge-base). **Offer Refinement**- A feedback flow triggered by clicking "Should be a buyer" or "Should NOT be a buyer" on a lead's detail view. AutoReach analyzes why the score was off and suggests edits to your offer's target audience, pain points, or ICP criteria. Once applied, the offer is updated and the lead is rescored. See [Score Feedback](/core-concepts/buyer-intelligence#score-feedback). **Pipeline**- The end-to-end flow a lead moves through: discovery, enrichment, scoring, engagement, response, and resurfacing. Also refers to the enrichment pipeline that processes leads through profile data collection, activity fetching, and scoring. **Poor Fit**- A buyer state for leads with low Buyer Scores who still have some baseline relevance. Poor Fit leads are monitored periodically and can be promoted if new signals appear. See [Buyer Intelligence](/core-concepts/buyer-intelligence). **RAG (Retrieval-Augmented Generation)**- A technique where AI responses are enhanced with relevant context retrieved from your knowledge base and tone examples via semantic search. AutoReach retrieves relevant context from your knowledge base and tone examples for each AI response. **Reel**- A short-form vertical video generated by AutoReach from a topic and your offer. Reels produce a finished MP4 plus a caption that you download and post manually; they are a content tool, not an outreach action. See [AI Video & Reels](/ai-video-and-reels/overview). **Resurfacing**- The process of automatically rechecking lower-scored leads for new signals on a regular schedule. When new signals boost a lead's score above the Active threshold, they are promoted and become eligible for auto-enrollment. See [Continuous Operations](/autopilot/continuous-operations). **Scoring**- The process of evaluating a lead across fit, intent, and timing dimensions to produce a Buyer Score. AutoReach uses AI-powered Deep Analysis to score leads. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence). **Sequence**- A multi-step outreach campaign built visually in the flow editor. Sequences define the actions (like, follow, DM, connection request), timing between steps, and branching logic for each enrolled lead. See [Building Sequences](/outreach-and-sequences/building-sequences). **Sequence Lead**- A lead that has been enrolled in a specific sequence. Each sequence lead tracks its own progress through the sequence steps, current status (pending, active, replied, meeting booked, completed, etc.), and action history. **Signal**- Any data point that contributes to a lead's buyer score. Signals include intent signals (asking for recommendations, switching tools), company signals (funding, hiring, expansion), engagement signals (competitor engagement, own-post engagement), and timing signals (job changes, new hires). See [Buyer Intelligence](/core-concepts/buyer-intelligence). **Simulation**- A preview environment where you can test DM personalization with real lead data, see variable substitution, and simulate AI responses to different lead replies, all without sending anything live. See [Simulation & A/B Testing](/outreach-and-sequences/simulation-and-testing). **Timing Score**- A score from 0-100 measuring how urgent or immediate a lead's needs appear. Evaluates signals like recent job changes, company funding, hiring activity, product launches, and recency of relevant social engagement. See [Buyer Intelligence & Scoring](/core-concepts/buyer-intelligence). **Tone Example**- A conversation sample stored per-sequence that teaches the AI how you want to sound. Tone examples are classified by conversation stage and retrieved via semantic search at response time. AutoReach auto-captures winning examples from booked meetings. See [Tone & Knowledge](/ai-and-conversations/tone-and-knowledge). **Tone Summary**- An AI-generated style guide extracted from your tone examples. It analyzes vocabulary, brevity, value-led approach, and anti-patterns. Auto-regenerated as new winning examples are captured. See [Tone & Knowledge](/ai-and-conversations/tone-and-knowledge). **Warmup**- The process of building a credible social presence on your accounts before launching outreach. The Engagement Engine handles warmup by posting content, engaging with relevant posts, and growing visibility with human-like patterns. See [Engagement Engine Overview](/engagement-engine/overview). **Watch Story**- An Instagram-only sequence and engagement action that views a lead's active stories, putting you on their story-viewer list. Often used as a light-touch warm-up before a follow or DM. Skipped if the lead has no active stories. See [Supported Actions](/outreach-and-sequences/supported-actions). **Primary Mailbox**- When you have multiple email accounts connected under a parent X or LinkedIn account, the Primary mailbox is used to send to recipients on neutral domains (anything that is not an obvious Gmail or Outlook domain). Set it from the Email card on the parent account's detail page. See [Connecting Email](/getting-started/connecting-email). **Webhook**- An integration that allows external services (Calendly, Cal.com) to notify AutoReach when events occur, such as meeting bookings. Webhooks enable automatic lead status updates when meetings are booked. See [Meeting Booking](/meetings-and-crm/meetings). **Withdraw Connection**- A LinkedIn-only action that withdraws a pending connection request. Automatically scheduled by a Connection Request step's auto-withdraw setting; not available in the flow builder palette and cannot be added to sequences manually. Does not count toward sequence progress. See [Supported Actions](/outreach-and-sequences/supported-actions). # Best Practices Practical tips for getting the most out of AutoReach. Each section covers a key area of the platform with actionable advice you can apply immediately. *** ## Where to Focus First If you only do five things, do these. Everything else is optimization. 1. **Spend 30 minutes on your Offer.** It drives lead scoring, message personalization, and AI responses. Vague offer = generic everything. → [Setting Up Your Offer](#1-setting-up-your-offer) 2. **Use intent-driven lead sources, not cold CSV imports.** Tweet/content search + lookalikes outscore mass imports. → [Finding the Right Leads](#2-finding-the-right-leads) 3. **Warm up new accounts for 1-2 weeks before outreach.** Cold DMs from a silent account get ignored or flagged. → [Account Safety](#6-account-safety) 4. **Start at 15-20 actions/day and ramp slowly.** Sudden spikes trigger bot detection. → [Account Safety](#6-account-safety) 5. **Run Simulation before launching any sequence.** Catch broken templates and weak personalization before 200 leads see them. → [Writing Better DMs](#4-writing-better-dms) Skip ahead to [Common Mistakes to Avoid](#8-common-mistakes-to-avoid) if you want the same advice in negative form. *** ## Sections | # | Section | When to read | | - | --------------------------------------------------------------- | -------------------------------------------- | | 1 | [Setting Up Your Offer](#1-setting-up-your-offer) | Before launching anything | | 2 | [Finding the Right Leads](#2-finding-the-right-leads) | Choosing discovery methods | | 3 | [Building Effective Sequences](#3-building-effective-sequences) | Designing your outreach flow | | 4 | [Writing Better DMs](#4-writing-better-dms) | Reply rate is low or you're tuning templates | | 5 | [Managing Conversations](#5-managing-conversations) | Replies are coming in | | 6 | [Account Safety](#6-account-safety) | Before launching, and weekly thereafter | | 7 | [Measuring Success](#7-measuring-success) | Reading dashboards, deciding what to fix | | 8 | [Common Mistakes to Avoid](#8-common-mistakes-to-avoid) | Quick troubleshooting checklist | *** ## 1. Setting Up Your Offer Your Offer is the foundation that powers everything in AutoReach: lead discovery, scoring, DM personalization, and AI responses. Time spent here pays off across the entire platform. * **Be specific about your target audience.** "VP of Demand Gen at B2B SaaS companies with $10M+ ARR" scores better than "marketing professionals." Include who to target, who to avoid, and any context that helps scoring. * **Write detailed pain points.** Name the exact problems your buyers face: "Outbound outreach timing is guesswork" is more powerful than "help with sales." Pain points drive intent matching and message personalization. * **Be specific in your description and pain points.** The more detailed your offer, the better AutoReach generates search keywords and matches leads via the Lead Pool. * **Add your known competitors.** This helps AutoReach detect leads engaging with competitor content, which is one of the strongest intent signals. * **Upload 3-5 high-quality knowledge base documents.** Case studies, pricing sheets, and objection-handling guides give the AI real context for conversations. A few focused documents outperform a large volume of generic content. * **Review and update quarterly.** As you learn which types of companies convert, refine your target audience and pain points to match. See [Creating Your First Offer](/getting-started/create-offer) and [Offers & Knowledge Base](/core-concepts/offers-and-knowledge-base). *** ## 2. Finding the Right Leads Not all lead sources are equal. Prioritize quality and intent over volume. * **Start with intent-driven sources.** X tweet search and LinkedIn content search find people actively discussing relevant topics. These leads carry inherent intent signals and score higher than cold imports. * **Use LinkedIn people search for systematic targeting.** When you need specific roles at specific companies or in specific locations, people search gives you precise control. * **Try lookalike audiences for scale.** Once you know which seed accounts attract your ideal buyers, lookalike discovery can tap into entire communities of relevant prospects. * **Enable buyer expansion for continuous growth.** Buyer expansion runs daily, re-extracting followers from seed accounts and rotating through role-based searches so your pipeline never dries up. * **Use the Lead Pool for instant results.** When you create a new offer, the Lead Pool delivers scored leads in seconds from pre-enriched profiles. Use it alongside other discovery methods, not as a replacement. * **Aim for 100-200 quality leads to start.** It is better to run a tight campaign on highly-scored leads than to blast 1,000 random prospects. You can always scale up after seeing what works. * **Include commenters in content searches.** People who comment on relevant posts often show stronger engagement and intent than passive readers. See [Finding Leads Overview](/finding-leads/overview). *** ## 3. Building Effective Sequences Sequences are your outreach campaigns. Structure them to build familiarity before making a direct ask. * **Lead with soft engagement.** Like a post, view a profile, or follow before sending a DM. This builds familiarity so the lead recognizes your name when your message arrives. * **Keep sequences to 3-5 steps.** Longer sequences are not necessarily better. A focused flow of like, follow, then DM often outperforms complex 10-step campaigns. * **Use conditions to branch intelligently.** On LinkedIn, always add a condition step after a connection request to check if it was accepted before sending a DM. This avoids wasting messages on unconnected leads. * **Set appropriate delays between steps.** Same-day engagement followed by a next-day DM feels natural. A DM within minutes of a like feels automated. * **Mind the total across sequences.** Daily limits are set per sequence, but your account's safe ceiling is a single number. If you run two sequences on the same account at 20/day each, that's 40 daily actions on the account. Size each sequence's limit so the combined total stays within your account-level threshold. See [Account Safety](#6-account-safety) below for starting limits. * **Use the simulation tool before launching.** Test your DM template with 3-5 different leads to check variable substitution, tone, and edge cases where data might be missing. See [Simulation & A/B Testing](/outreach-and-sequences/simulation-and-testing). * **Associate an offer with each sequence.** This connects your sequence to the right scoring and personalization context. See [Building Sequences](/outreach-and-sequences/building-sequences). *** ## 4. Writing Better DMs Your first message determines whether a lead responds. Make it count. * **Reference something specific.** Use `{{post}}` to reference a lead's recent content, or `{{current_role}}` and `{{company_name}}` to show you did your research. Generic messages get ignored. * **Keep it to 50-100 words.** Short, direct messages with one clear question at the end outperform long pitches. * **End with exactly one question.** A single, easy-to-answer question gives the lead a natural way to respond. Multiple questions feel overwhelming. * **Avoid banned openers.** "Saw you..." and "Noticed you..." are overused. Open with a statement or insight instead. * **Use the variable picker.** Browse available template variables in the flow editor rather than guessing syntax. The more lead data you reference, the more personalized the message feels. * **Test with leads who have incomplete data.** Some leads may not have a company name or recent posts. Make sure your template still reads well when variables are missing. * **Set a custom DM generation prompt for DM-specific instructions.** This is separate from the general AI prompt and lets you control cold DM tone independently from conversation replies. See [DM Personalization](/outreach-and-sequences/dm-personalization). *** ## 5. Managing Conversations Conversations are where deals happen. Balance AI automation with personal touch. * **Let AI handle the first 2-3 replies.** AI responses are stage-aware and handle common scenarios well: acknowledging interest, asking discovery questions, and handling basic objections. * **Take over manually for hot prospects.** When a lead shows strong buying intent or asks detailed product questions, toggle AI off and respond personally. Your highest-value conversations deserve your direct attention. * **Add tone examples from your best conversations.** When a conversation leads to a meeting, review the exchanges and add the best ones as tone examples. AutoReach auto-captures winning patterns, but manual curation improves quality. * **Enable conversation follow-ups selectively.** Automatic follow-ups work well for mid-funnel leads who went silent. Set a 3-5 day wait with a max of 1-2 follow-ups to avoid being pushy. * **Use the Conversation Analyzer monthly.** After you have enough data (20+ conversations), run the analyzer to identify patterns in what worked versus what did not, and apply the suggested improvements. * **Review the AI's graceful exits.** When the AI detects a dead-end conversation, it exits respectfully. Review these to make sure the AI's exit language matches your style. See [AI Response Engine](/ai-and-conversations/ai-response-engine), [Tone & Knowledge](/ai-and-conversations/tone-and-knowledge), and [Inbox](/ai-and-conversations/inbox). *** ## 6. Account Safety Protecting your social accounts is non-negotiable. AutoReach has built-in protections, but your behavior matters too. * **Warm up new accounts before outreach.** Run the Engagement Engine for at least a week on new or inactive accounts before launching sequences. Build a visible activity history first. * **Use a quality proxy.** Proxies are optional and managed on the [Proxies](/settings-and-configuration/multi-account#proxy-configuration) page. You can share one residential proxy across your accounts or assign a dedicated proxy per account. A dedicated proxy per account isolates IP-based risk best; a shared residential proxy is fully supported. * **Start with conservative daily limits.** Begin at 15-20 actions per day and increase slowly over weeks, not days. * **Monitor the Accounts page regularly.** Check error rates, cooldown status, and account health indicators. Catching issues early prevents escalation. * **Respect emergency pauses.** When AutoReach pauses an account for bot detection or IP blocks, do not immediately resume. Wait for the full cooldown period and investigate the cause. * **Keep your Chrome Extension active.** Session cookies expire. Keep the extension running to ensure your accounts stay connected. * **Sign in to LinkedIn from one browser only.** LinkedIn rotates its session token (`li_at`) every time the same account signs in from a new browser, phone, or device. That rotation invalidates the cookies AutoReach holds and forces a reconnection. Pick one browser as your daily LinkedIn workspace and avoid logging in elsewhere. To audit existing sessions, open LinkedIn and go to **Settings & Privacy > Sign In & Security > Where you're signed in**, then end any sessions you do not actively use. * **AutoReach's proxy will not appear in LinkedIn's "Where you're signed in" list.** This is normal. AutoReach reuses your browser's existing session cookie rather than logging in again, so no new session entry is created. As long as the Accounts page shows the account as active, the connection is healthy. * **Do not run AutoReach and manual mass-actions simultaneously.** If you are manually liking 50 posts while AutoReach is also running engagement, the combined activity can trigger rate limits. See [Account Safety](/settings-and-configuration/account-safety) and [Engagement Engine Overview](/engagement-engine/overview). *** ## 7. Measuring Success Track the right metrics to know whether your outreach is working and when to adjust. * **Reply rate is your primary signal.** A reply rate above 10% on cold DMs indicates good targeting and messaging. Below 5% means either your ICP is too broad or your message needs work. * **Track meetings booked per week.** This is the ultimate outcome metric. If replies are high but meetings are low, your conversation AI or follow-up strategy needs adjustment. * **Compare lead sources.** Check which discovery method (tweet search, LinkedIn search, lookalike, Lead Pool) produces the highest-scoring leads and best reply rates. Double down on what works. * **Monitor your Active-to-Enrolled ratio.** If many Active leads are not being enrolled, check your auto-enrollment settings and sequence status. * **Watch the Monitor-to-Active promotion rate.** A healthy pipeline has leads regularly graduating from Monitor to Active as new signals appear. If promotions are rare, your offer's pain points and target audience may need refinement. * **Review AI response quality monthly.** Read through 10-20 AI-generated conversations to check for tone drift, missed objections, or generic responses. * **Check cost per lead.** Use the cost estimation tool to understand your per-lead AI spend and optimize model selections for high-volume categories. See [Pipeline Cost Estimation](/settings-and-configuration/cost-estimation) and [AI Model Configuration](/settings-and-configuration/ai-models). *** ## 8. Common Mistakes to Avoid The top mistakes new users make and how to avoid them. 1. **Skipping the Offer setup.** A vague offer definition produces generic scoring, weak personalization, and irrelevant leads. Spend 30 minutes crafting a detailed offer before doing anything else. 2. **Importing thousands of cold CSV leads first.** CSV imports have no intent signals and require full enrichment. Start with intent-driven sources (tweet search, LinkedIn content search) that produce higher-scoring leads faster. 3. **Setting daily limits too high on new accounts.** Jumping to 100 actions per day on a fresh account is a fast path to getting flagged. Start at 15-20 and increase gradually over weeks. 4. **Ignoring the simulation tool.** Launching a sequence to 200 leads without testing your DM template is risky. Always simulate with 3-5 diverse leads first to catch issues with variable substitution, tone, and missing data. 5. **Leaving AI on for every conversation.** AI handles routine responses well, but your highest-value prospects deserve personal attention. Toggle AI off for conversations that show strong buying intent and handle them yourself. 6. **Never updating tone examples.** Tone examples define how your AI sounds. If you never add examples from real conversations, the AI relies on generic patterns. Add examples from your best conversations regularly. 7. **Running outreach without warming up accounts.** Sending cold DMs from an account with no recent activity looks suspicious. Run the Engagement Engine for at least a week first to build credibility. 8. **Using broad, untargeted searches.** Searching for generic terms like "marketing" produces thousands of irrelevant leads. Use specific intent phrases like "looking for a marketing automation tool" to find people with real buying signals. 9. **Forgetting to set up webhook integration.** Without a Calendly or Cal.com webhook, AutoReach cannot automatically detect booked meetings. Set up the webhook during onboarding to get automatic meeting tracking. 10. **Not reviewing the Buyers page.** When auto-enroll is off, scored leads appear on the Buyers page waiting for your review. Check it regularly to enroll high-quality leads and adjust your targeting based on what you see. # Troubleshooting A centralized guide for diagnosing and fixing common issues in AutoReach. Each section follows a **Symptom / Likely Cause / Fix** pattern. *** ## Account Connection Issues ### X Account: "Cookies Invalid" **Symptom:** Your X account shows as disconnected or you see a "cookies invalid" error. **Likely Cause:** Your X session has expired. Session cookies typically last 2-4 weeks before they need to be refreshed. **Fix:** Open the Chrome Extension, click the **three dots** next to the affected account, and click **Reconnect**. Make sure you are logged into X in your browser before reconnecting. *** ### X Account: "Proxy Connection Failed" **Symptom:** Actions fail with a proxy-related error. **Likely Cause:** The proxy assigned to the account has gone offline or its credentials have changed. **Fix:** Open the **Proxies** page to check the account's assigned proxy. If you use a bring-your-own proxy, verify the credentials and re-save them (AutoReach re-tests connectivity on save), or assign a different proxy. If you use a managed proxy and see persistent errors, contact support at . *** ### LinkedIn Account: Session Expired **Symptom:** LinkedIn actions fail or the account shows as inactive. You may receive a "LinkedIn cookie expiry" email. **Likely Cause:** LinkedIn issued a new session token (`li_at`) for your account, which invalidates the cookies stored in AutoReach. The most common triggers are: 1. **Signing in to the same LinkedIn account from another browser, phone, or device.** LinkedIn rotates session tokens whenever it sees a new device, especially if the IP differs from the proxy AutoReach uses. 2. **Changing your LinkedIn password** or accepting a security prompt. 3. **Choosing "Sign out of all sessions"** in LinkedIn settings. 4. LinkedIn detecting unusual activity and forcing a session refresh. **Fix:** 1. Open the Chrome Extension on linkedin.com, click the **three dots** next to the account, and click **Reconnect**. You must be logged into LinkedIn in the same browser. 2. To prevent this from repeating, see [LinkedIn Account: Repeatedly Disconnects](#linkedin-account-repeatedly-disconnects) below. *** ### LinkedIn Account: Repeatedly Disconnects **Symptom:** Your LinkedIn account expires and needs reconnection multiple times per week, even though you reconnect promptly each time. **Likely Cause:** You are logging in to the same LinkedIn account from multiple browsers or devices (for example, your work laptop, a personal browser, the LinkedIn mobile app, or a second computer). Each new sign-in causes LinkedIn to rotate the `li_at` cookie, which invalidates the one AutoReach is using. **Fix:** Pick one access path and stick to it. 1. Open LinkedIn in your browser and go to **Settings & Privacy > Sign In & Security > Where you're signed in**. 2. Review the list of active sessions. Each entry shows a device, location, and last activity time. 3. Sign out of any sessions you do not actively use (old browsers, phones you no longer carry, shared computers). Click the three dots next to the session and select **End**. 4. Going forward, sign in to this LinkedIn account from a single browser only. If you need to use the LinkedIn mobile app, sign in there once and avoid switching devices frequently. 5. Reconnect the account in AutoReach via the Chrome Extension. > **Note:** AutoReach's proxy IP will not appear in "Where you're signed in" because AutoReach reuses your existing session cookie via API calls rather than performing a full login. Not seeing the proxy listed is normal and not a sign of a problem. See the next entry for details. *** ### LinkedIn Settings: AutoReach's Proxy Is Not Listed Under "Where You're Signed In" **Symptom:** You check **Settings & Privacy > Sign In & Security > Where you're signed in** in LinkedIn and you do not see AutoReach's proxy IP or location in the list of active sessions. **Likely Cause:** This is expected behavior, not a misconfiguration. LinkedIn's "Where you're signed in" page only shows sessions that completed a full login flow (username, password, browser fingerprint). AutoReach connects to LinkedIn by reusing the session cookie that your own browser created when you first connected the account through the Chrome Extension. Because no new login flow happens, no new session entry appears. **Fix:** No action needed. Your account is connected correctly as long as the Accounts page in AutoReach shows the account as **active**. *** ### LinkedIn Account: Emergency Pause **Symptom:** All LinkedIn activity has stopped and the account shows as paused. **Likely Cause:** AutoReach detected an issue with the account (rate limit, bot detection, IP block, captcha, or session expired) and paused all associated automation - sequences, warmup, and pending actions are all halted. **Fix:** What to do depends on the error type shown on the Accounts page: * **Auto-recoverable** (rate limit, timeout, proxy error, IP block, expired auth): cooldowns range from 2h to 7 days. Activity resumes automatically once the cooldown expires - do not attempt to bypass the pause. * **Manual fix required** (bot detection, captcha, AI provider out of credits): you will receive an email and must address the underlying issue, then manually resume the account from the Accounts page. Bot detection in particular is serious - treat repeat occurrences as a sign to reduce activity significantly. *** ### Chrome Extension: "License Key Invalid" **Symptom:** The extension rejects your license key. **Likely Cause:** The key has expired or was entered incorrectly. **Fix:** 1. Go to your AutoReach settings and copy your current license key 2. Paste it into the extension 3. Click **Activate** *** ### Chrome Extension: "Extension Can't Access This Page" **Symptom:** The extension shows this message when you click it. **Likely Cause:** You are on a page other than linkedin.com or x.com. **Fix:** Navigate to linkedin.com or x.com first, then click the extension icon. *** ## Sequence Not Sending Actions ### Sequence is active but nothing is happening **Symptom:** You started a sequence but no actions are being executed. **Likely Cause (1):** No leads are enrolled. A sequence needs at least one lead to run. **Fix:** Open the sequence, go to the Leads tab, and add leads from your pipeline. **Likely Cause (2):** The acting account's activity window is closed. Actions only execute during the account's configured activity hours. **Fix:** Activity windows are per-account. Open **Accounts → \[account] → Configuration → Activity Window** and check the start/end/timezone. If it is set to business hours and it is currently outside those hours, actions will resume when the window opens. **Likely Cause (3):** Your daily action limit has been reached for today. **Fix:** Wait until tomorrow, or increase your daily action limits in the sequence settings (keeping in mind platform safety guidelines). **Likely Cause (4):** The connected account is paused or has errors. **Fix:** Check the **Accounts** page. If the account shows errors or a paused state, resolve the underlying issue (see Account Connection Issues above). *** ### Actions are stuck in "pending" status **Symptom:** Actions appear in the timeline but never execute. **Likely Cause:** The actions are scheduled for a future time based on the delays configured in your flow. **Fix:** Check the scheduled time on each pending action in the lead timeline. If the timing looks correct, just wait- actions execute at their scheduled time within your activity window. If the delays are wrong, pause the sequence, adjust step delays in the flow editor, then resume. *** ### Specific action type keeps failing **Symptom:** DMs, likes, or follows repeatedly fail with errors. **Likely Cause:** The platform is rate-limiting that specific action type, or the lead's account has restrictions (e.g., DMs are closed, account is private). **Fix:** 1. Check the error message on the failed action 2. If rate-limited, reduce your daily limits for that action type 3. If the lead's account has restrictions, the action may not be possible - consider removing that lead or skipping the step 4. Retry the failed action using the retry button *** ## Leads Not Getting Scored ### Leads stuck without a buyer score **Symptom:** Leads appear in your pipeline but have no score (showing as "unscored" or no buyer state). **Likely Cause (1):** Enrichment has not completed yet. Scoring requires profile data, which is gathered during enrichment. **Fix:** Check if enrichment is still in progress. New leads go through enrichment before scoring, which can take a few minutes per lead. **Likely Cause (2):** You are out of AI credits (or, if you use your own API key, that key has run out of credits). **Fix:** Check the health warnings on the Dashboard and the **Credits** page. If you run on included credits, top up there. If you use your own OpenAI / Anthropic / DeepSeek key, add credits to that provider account. **Likely Cause (3):** Your offer is missing key ICP information. **Fix:** Open your offer and make sure you have filled in target audience, pain points, and qualifying criteria. Scoring quality depends on a well-defined ICP. *** ## AI Messages Not Generating ### DMs or replies show as failed with AI errors **Symptom:** Message generation fails, and the action is marked as failed. **Likely Cause (1):** You are out of AI credits, or (if you use your own key) the key is invalid or out of credits. **Fix:** Check the **Credits** page and the Dashboard health warnings. If you run on included credits, top up. If you use your own key, verify it is correct in **Settings > AI & Models** and has available credits. **Likely Cause (2):** Both your primary and fallback AI models are down. **Fix:** This is rare but can happen during provider outages. AutoReach will show a health warning. Wait for the provider to recover, or switch your model configuration in **Settings > AI & Models**. *** ### AI messages sound generic **Symptom:** Messages are technically correct but lack personality or feel templated. **Likely Cause:** The AI does not have enough context about your voice and style. **Fix:** 1. Add **tone examples** to your sequence (real messages you have sent that represent your style) 2. Customize the **AI prompt** and **DM generation prompt** in sequence settings with specific tone instructions 3. Use the **Simulation** tool to preview and iterate on message quality *** ## Low Reply Rates ### Reply rate below 3% after 100+ DMs **Symptom:** You have sent a meaningful volume of messages but very few people are responding. **Likely Cause (1):** Weak targeting. Your leads may not be a good fit for your offer. **Fix:** 1. Open your sequence and check the score distribution of enrolled leads. If most are below 50, your offer's ICP is too broad or doesn't match the leads your searches are surfacing. 2. The fastest way to fix this: open a lead that scored wrong and click **Should be a buyer** (thumbs up) or **Should NOT be a buyer** (thumbs down). This launches the **Offer Refinement** flow, where AutoReach analyzes why the score was off and suggests specific edits to your target audience or pain points. Review the suggestions and apply them. See [Score Feedback](/core-concepts/buyer-intelligence#score-feedback). 3. After updating the offer, new leads will be scored against the updated ICP. To re-evaluate leads already in your pipeline, select them on the Leads page and run a rescore. See [Rescoring](/core-concepts/buyer-intelligence#rescoring) for how this works. 4. Switch your discovery method. If you've been using broad people-search by role, try **From Signals** (intent-based) or **From a Lookalike** (audience of an account whose followers match your ICP). These produce higher-intent leads than role-only search. 5. Watch the next 50-100 leads. If scores trend higher and reply rate improves, the offer was the issue. If not, move to Cause (2). **Likely Cause (2):** Messages are not compelling or personalized enough. **Fix:** Use Simulation to preview your DMs for several different leads. If the messages read as generic, the fix is usually in the offer (richer pain points give the AI more to work with) or the template (add a `{{post}}` reference on X, or `{{current_role}}` + `{{company_name}}` on LinkedIn). Refine your message template and AI prompt until messages feel natural and relevant. See [DM Personalization](/outreach-and-sequences/dm-personalization). > **How to tell which cause is yours:** Run Simulation on 10 diverse leads. If the generated messages look generic and interchangeable, fix the message (Cause 2). If the messages look personalized but the leads themselves seem off-target, fix the offer (Cause 1). **Likely Cause (3):** No warmup before the DM. **Fix:** Add engagement steps before the DM in your sequence flow (like a post, follow the account, view their profile). People respond better to names they recognize. **Likely Cause (4):** Wrong platform or timing. **Fix:** Try switching to the other platform (X vs. LinkedIn). Also review the sending account's activity window (per-account, set on the account's Configuration tab) to ensure messages arrive during business hours in the lead's time zone. *** ## Account Health Warnings ### "Rate Limited" status **Symptom:** Account shows as rate limited, and some actions are paused. **Likely Cause:** You hit the platform's daily or hourly limit for a specific action type. **Fix:** AutoReach automatically applies cooldown periods when rate limits are hit. Wait for the cooldown to expire, then reduce your daily action limits to stay under the platform's thresholds. Start conservative (15-20 DMs/day for X, stay well under LinkedIn's daily connection limit). *** ### "Bot Detected" status **Symptom:** Account shows bot detection warning, and all activity is paused. **Likely Cause:** The platform flagged a pattern of activity that looks automated (too many actions in rapid succession, unusual hours, or suspicious IP). **Fix:** Bot detection is treated as a serious event that requires manual intervention - AutoReach will email you and the account stays paused until you manually resume it from the Accounts page. Before resuming: 1. Investigate what the platform flagged (the error message on the account is your best clue) 2. Reduce your daily action limits significantly 3. Make sure the account's activity window (per-account, on the Configuration tab) matches realistic usage patterns 4. Verify your proxy is working (residential IPs are safest) 5. Manually resume the account from the Accounts page once you have addressed the underlying cause > **Warning:** Repeated bot detection warnings can lead to permanent account suspension. Take them seriously and reduce your activity levels. *** ### "Auth Error" on LinkedIn **Symptom:** LinkedIn account shows authentication error. **Likely Cause:** Your session cookies have been revoked, often because LinkedIn detected a mismatch between your login location and the proxy location. **Fix:** Reconnect your LinkedIn account via the Chrome Extension. If this happens repeatedly, contact support - your proxy configuration may need adjustment. *** ## Chrome Extension Issues ### Extension not detecting the current page **Symptom:** The extension popup is blank or shows "not connected" even though you are on linkedin.com or x.com. **Likely Cause:** The extension does not have permission to access the current tab, or the page has not fully loaded. **Fix:** 1. Make sure the page has fully loaded before clicking the extension 2. Try refreshing the page 3. Check Chrome's extension settings to ensure AutoReach has permission to access linkedin.com and x.com 4. If the issue persists, remove and reinstall the extension *** ### "Connection Not Detected" for LinkedIn **Symptom:** A LinkedIn connection was accepted but AutoReach does not detect it. **Likely Cause:** The extension was not running or the LinkedIn tab was closed when the acceptance happened. **Fix:** Use the **Check Acceptances** button in the Chrome Extension's CRM panel to manually trigger a detection check. *** ## Enrichment Failures ### Leads stuck in "enriching" state **Symptom:** Leads have been in the enrichment stage for a long time without progressing to scored. **Likely Cause (1):** The enrichment queue is backed up. If you added many leads at once, they are processed sequentially. **Fix:** Wait for the queue to clear. Enrichment processes leads in order, and large batches may take time. **Likely Cause (2):** The platform is rate-limiting profile data requests. **Fix:** This usually resolves on its own as AutoReach respects rate limits and retries. If leads are stuck for more than an hour, try the recovery option in the lead's menu. **Likely Cause (3):** The lead's profile is private or has been deleted. **Fix:** Private or deleted profiles cannot be enriched. These leads will be marked accordingly and can be removed from your pipeline. *** ### Cross-platform match not found **Symptom:** A LinkedIn lead does not get matched to an X profile (or vice versa). **Likely Cause:** The lead does not have a public presence on the other platform, or their usernames are too different for automated matching. **Fix:** This is expected behavior. Not everyone has accounts on both platforms. You can manually add a profile URL for the other platform if you know it. *** ## Getting More Help If your issue is not covered here: 1. Check the specific feature's documentation page - most have their own troubleshooting section 2. Email support at **** with a description of the issue and any error messages you see # Quickstart Get AutoReach up and running and launch your first outreach sequence. The onboarding wizard will guide you through each step. ## Step 1: Create Your Account Visit [autoreach.tech](https://autoreach.tech) and sign up with your email. Signup starts a **7-day software trial** with a **non-refundable $20 activation fee charged today**. The activation includes starter AI credits and managed-proxy activation if selected. Your subscription begins automatically at **$99/mo** after the trial unless you cancel before the seven days end. There is no confirmation email. You can start using AutoReach immediately after signing up. > **Note:** Prefer to have us run everything for you? The Done For You tier is **$299 setup + $200 per booked meeting**. Contact support to set it up. ## Step 2: Your AI Is Included AutoReach uses AI to find and score leads, personalize messaging, and reply on your behalf. There is nothing to set up: every account starts with **AI credits included in the activation fee**, and you can top up anytime from the **Credits** page. Prefer to run on your own provider? You can add your own **OpenAI, Anthropic, or DeepSeek** key in **Settings > AI & Models** at any time for unlimited usage billed directly by your provider. This is optional, the onboarding wizard does not ask for any keys. ## Step 3: Set Your Activity Window Your activity window controls when sequences are active and outreach actions are performed. The onboarding wizard will prompt you to configure this. 1. Set your **Start Time** and **End Time** (e.g., 9:00 AM to 9:00 PM) 2. Select your **Timezone** Outside your activity window, AutoReach pauses all outreach actions. This makes your activity look natural. > **Note:** Activity windows are **per-account**. The setting you pick during onboarding is the user-level default. After connecting accounts, you can give each LinkedIn, X, and Instagram account its own start/end/timezone from **Accounts → \[account] → Configuration → Activity Window**. Each account can have up to 8 hours/day. Accounts without their own customization inherit the user-level default. ## Step 4: Set Up Your Proxy and Chrome Extension The onboarding wizard will guide you through these steps: 1. **Choose your proxy (optional)**: A proxy is optional and chosen here during onboarding, not at signup. You have two options: * **Managed**: AutoReach provisions a secure ISP residential static proxy for you. Activation is included in the signup fee; if you keep the managed proxy after the trial, it costs **$15/mo**. One managed proxy is shared across your LinkedIn, X, and Instagram accounts. * **Bring your own (BYOP)**: Free. Enter your proxy host, port, username, and password (HTTP or SOCKS5). AutoReach verifies it connects before saving. You can also manage proxies later from the **Proxies** page (`/proxies`). 2. **Download the Chrome Extension**: You can only connect your X, LinkedIn, and Instagram accounts through the Chrome Extension. See [Installing the Chrome Extension](/getting-started/chrome-extension) for installation instructions. > **Important:** Social accounts (X, LinkedIn, Instagram) can ONLY be linked through the Chrome Extension. There is no other way to connect them. ## Step 5: Generate Your License Key The onboarding wizard will prompt you to generate your license key. Click **Generate License Key** and copy it. ## Step 6: Add the License Key to the Extension 1. Click the **AutoReach icon** in your Chrome toolbar 2. Paste your license key 3. Click **Activate** ## Step 7: Connect Your LinkedIn Account 1. Visit [linkedin.com](https://linkedin.com) while logged in 2. Click the **AutoReach extension icon** 3. Enter your **first name** and **last name** 4. Click **Connect Account** The extension will automatically extract your session cookies and link your LinkedIn account to AutoReach. ## Step 8: Connect Your X/Twitter Account 1. Visit [x.com](https://x.com) while logged in 2. Click the **AutoReach extension icon** 3. Enter your **name** and a **4-digit PIN** for X DM Chat 4. Click **Connect Account** The extension will automatically extract your session cookies and link your X account to AutoReach. > **Tip:** You can also connect Instagram from the extension. Your subscription includes one account slot per platform (1 LinkedIn + 1 X + 1 Instagram). At minimum, you need one social account connected to use AutoReach. Need more accounts? Contact support. > **Tip:** You can also add email as a channel. See [Connecting Email](/getting-started/connecting-email) to link Gmail and/or Outlook. ## Step 9: Set Up Your Calendar Link Once an account is connected, the onboarding wizard will ask you to configure a meeting booking link **for that account** (calendar config is per-account). AutoReach uses it to inject your booking link into outreach messages and to track when meetings are booked. Choose your calendar provider: * **Calendly**: Requires a paid Calendly subscription. Paste your booking link, paste a Calendly Personal Access Token, click **Auto-fetch** to retrieve your Organization URI, then click **Register webhook** and AutoReach creates the webhook for you. **No form field setup needed**: AutoReach now uses invisible URL-based attribution (`utm_content`), so bookers see only Calendly's default name + email fields. * **Cal.com**: Free, no subscription required. Two steps: (1) paste the webhook link AutoReach generates into **Cal.com → Settings → Developer → Webhooks** and subscribe to `BOOKING_CREATED`, then (2) add a hidden `username` short-text question to each event type under **Advanced → Booking Questions**, with **Disable input if the URL identifier is prefilled** checked. AutoReach prefills it via the URL so bookers never see it. * **Custom calendar link**: Use any booking URL. No webhook setup needed. See [Meetings](/meetings-and-crm/meetings) for full details. > **Note:** Webhooks are optional but strongly recommended. Without one, AutoReach can still inject your booking link into messages, but it can't automatically detect when a meeting is booked, so you'll be tracking bookings in a spreadsheet. With a webhook, AutoReach auto-matches bookings to leads via the URL tracking parameter (or falls back to email matching). ## Step 10: Create Your First Offer > **This is the most important step.** Steps 1-9 are mechanical setup that takes minutes. Your Offer is the one thing that compounds: it shapes which leads AutoReach finds, how they get scored, what your DMs say, and how the AI replies on your behalf. A vague offer produces vague results everywhere downstream. A sharp offer makes every other feature work better. Plan to spend \~30 minutes here, not 5. Once your accounts are connected, the onboarding wizard will guide you to create your first Offer. An **Offer** describes what you sell, who you're targeting, and what you want to achieve. 1. The easiest way: **add your website URL** and AutoReach will auto-populate your offer details. Review and modify what's needed. 2. Or fill in the details manually: name, description, target audience, goal, pain points, etc. What to focus on: * **Target audience**: be specific about who to target *and who to avoid*. "VP of Demand Gen at B2B SaaS, 50+ employees, avoid agencies" beats "marketing leaders." * **Pain points**: name the exact problems your buyers feel. These drive intent matching and message personalization. * **Known competitors**: leads engaging with competitor content is one of the strongest buying signals AutoReach detects. See [Creating Your First Offer](/getting-started/create-offer) for a detailed guide with examples. ## Step 11: Start Autopilot After creating your offer, start **Autopilot**. This is the recommended way to get your first outreach running. When you enable Autopilot, it automatically: 1. **Fetches 100 leads** from the database and scores them against your offer 2. **Starts 1 role-based search** on LinkedIn to find matching prospects 3. **Finds 1 lookalike** account and starts a search on their followers 4. **Starts an intent signal search** to find prospects showing buying signals 5. **Creates your first sequence** with the right workflow for your platform 6. **Starts the sequence**, and outreach begins automatically > **Note:** When Autopilot is active, the auto-enroll setting is turned ON in your settings. This means all ready buyers found by the system are automatically added to the sequence and will **not** appear on the Buyers page. The Buyers page only shows buyers before they are added to a sequence. > **Note:** Autopilot also sets all searches with Buyer Expansion enabled, meaning searches will run every day to keep your pipeline full with fresh leads. ## What's Next? Your outreach is now running on autopilot. Here's what to do: * **Read the** [**Day 2 Guide**](/getting-started/getting-results): what to expect in the first 24-48 hours, when first replies arrive, and how to read your early metrics. Read this *before* you start refreshing your dashboard. * **Check your Inbox**: Replies and conversations appear here in real-time. * **Review the Buyers page**: See scored leads before they're enrolled (when auto-enroll is off). * **Monitor Autopilot**: Check the [Autopilot Dashboard](/autopilot/overview) for stats and activity. * **Review your Sequence**: Go to your sequence's **Advanced Settings** and review the prompt, the configuration, and the tone examples. Tone examples define how your outreach sounds and can be customized per sequence. * **Customize your Offer**: Upload [knowledge base documents](/getting-started/create-offer#knowledge-base) and refine your target audience. Need help? Email . # Connecting Your Accounts Connect your X (Twitter), LinkedIn, and Instagram accounts to AutoReach through the Chrome Extension. This is the only supported method for linking social accounts. > **Important:** Make sure you have [installed the Chrome Extension](/getting-started/chrome-extension) and activated your license key first. > **Email accounts:** Gmail and Outlook are connected separately, not through the Chrome Extension. See [Connecting Email](/getting-started/connecting-email) for setup. *** ## Connecting X (Twitter) 1. Visit [x.com](https://x.com) and make sure you are logged in 2. Click the **AutoReach extension icon** in your Chrome toolbar 3. Enter your **name** 4. Enter your **4-digit PIN** for X DM Chat 5. Click **Connect Account** The extension links your active browser session to AutoReach automatically. > **Warning:** Do not use automated engagement on X aggressively. Aggressive automated activity is the number one reason accounts get suspended. Start conservatively and increase gradually. ## Connecting LinkedIn 1. Visit [linkedin.com](https://linkedin.com) and make sure you are logged in 2. Click the **AutoReach extension icon** in your Chrome toolbar 3. Enter your **first name** and **last name** 4. Click **Connect Account** The extension links your active browser session to AutoReach automatically. ## Connecting Instagram 1. Visit [instagram.com](https://instagram.com) and make sure you are logged in 2. Click the **AutoReach extension icon** in your Chrome toolbar 3. Select **Instagram** and confirm the account 4. Click **Connect Account** The extension links your active browser session to AutoReach automatically. When you are signed in to a single Instagram account, the extension can also refresh the session for you after you log back in. If you use multiple Instagram accounts, reconnect the specific one from the extension when prompted. > **Warning:** Instagram is sensitive to automated activity. Start conservatively, lean on the Engagement Engine to build activity before outreach, and keep daily action volumes modest on new accounts. *** ## Proxy Configuration A proxy is **optional** in AutoReach, and you choose how to set one up during onboarding (not at signup). You have two options: * **Managed**: AutoReach provisions a secure ISP residential static proxy for you. Activation is included in the $20 signup fee; if you keep the managed proxy after the trial, it costs **$15/mo**. A single managed proxy is shared across your LinkedIn, X, and Instagram accounts. * **Bring your own (BYOP)**: Free. Enter your proxy host, port, username, and password (HTTP or SOCKS5). AutoReach verifies the proxy connects before saving it. ISP residential static proxies use real internet service provider IPs, making your activity indistinguishable from a normal user. You can add, edit, or switch proxies any time from the **Proxies** page (`/proxies`). *** ## LinkedIn Daily Connection Limits LinkedIn enforces daily limits on connection requests. AutoReach tracks these automatically: AutoReach lets you set a daily connection request limit per account. Any whole number from 1 to 100 is allowed (default 15, which is the recommended starting point). Choose a limit that matches your account type and risk tolerance - lower limits are safer for new or free accounts. These limits reset daily. AutoReach tracks your usage and pauses connection requests when you approach the limit. If a connection request gets rate-limited, it enters a **deferred state** and automatically resumes the next day. *** ## Inbound Engagement Detection AutoReach monitors your accounts for inbound engagement and processes it as buying signals. **X (Twitter):** Retweets, likes on your posts, and new followers. **LinkedIn:** Reactions on your posts, comments on your content, mentions of you, and new connections. Someone engaging with your content may be a potential buyer. AutoReach can automatically create leads from high-intent engagers and update signals on existing leads. *** ## Account Health AutoReach continuously monitors your account health and warns you when something is wrong. See [**Account Safety**](/settings-and-configuration/account-safety) for details on error handling, cooldowns, and emergency pausing. ### Prevention Tips 1. Use a residential proxy (managed or bring-your-own) for your accounts 2. Set realistic daily action limits (start conservatively on new accounts) 3. Space out your sequences with wait periods 4. Do not engage with spam, adult, or hateful content 5. Start with the Engagement Engine to build activity history before launching sequences 6. Monitor your account health regularly in the Accounts page *** ## Troubleshooting **"Cookies Invalid"** - Your session has expired. Open the Chrome Extension, click the three dots next to the account, and click **Reconnect**. **"Proxy Connection Failed"** - Check your proxy on the **Proxies** page. If you bring your own proxy (BYOP), verify the host, port, username, and password are still valid and re-save to re-run the connectivity check. If you use a managed proxy and see persistent errors, contact support at . **"Rate Limited"** - You have hit the platform's action limit. Wait for the cooldown to expire. Consider lowering your daily limits. **LinkedIn account keeps disconnecting** - The most common cause is signing in to the same LinkedIn account from multiple browsers, phones, or devices. Each new sign-in causes LinkedIn to rotate your session token and invalidate the cookies AutoReach is using. Open LinkedIn and go to **Settings & Privacy > Sign In & Security > Where you're signed in**, then end any sessions you do not actively use. Going forward, keep LinkedIn signed in on a single browser only. See [Troubleshooting: LinkedIn Account: Repeatedly Disconnects](/troubleshooting#linkedin-account-repeatedly-disconnects) for the full procedure. **AutoReach's proxy is not visible in LinkedIn's "Where you're signed in" list** - This is expected. AutoReach reuses your browser's session cookie rather than performing a full login, so no separate session entry is created in LinkedIn. The Accounts page in AutoReach is the source of truth for connection health. *** ## Next Steps Once your accounts are connected: 1. [**Connect Email**](/getting-started/connecting-email) to add Gmail and/or Outlook as a third channel (optional but recommended) 2. [**Create an Offer**](/getting-started/create-offer) to define your target audience and message tone 3. [**Build a Sequence**](/outreach-and-sequences/building-sequences) to automate outreach 4. [**Launch Autopilot**](/autopilot/overview) for fully automated pipeline management # Connecting Email (Gmail & Outlook) Connect Gmail and Outlook mailboxes to send cold emails, monitor replies, and keep email conversations in your unified Inbox alongside X and LinkedIn DMs. Email connections are scoped to a parent X or LinkedIn account. You can connect one or both providers per parent account, and AutoReach will route emails to the right mailbox based on the recipient's domain. *** ## Connecting Gmail Gmail uses an **App Password** (not OAuth). This is faster to set up and works for both personal and Google Workspace accounts (subject to admin policy). ### Prerequisites * Gmail or Google Workspace account * 2-Step Verification enabled on the Google account * IMAP enabled on the account ### Setup 1. **Enable 2-Step Verification** at [myaccount.google.com/security](https://myaccount.google.com/security) 2. **Generate an App Password**: * Visit [myaccount.google.com/apppasswords](https://myaccount.google.com/apppasswords) * Name it "AutoReach" * Copy the 16-character password 3. In AutoReach, open the parent X or LinkedIn account from the **Accounts** page, find the **Email** card on the **Configuration** tab, and click **Connect Gmail** 4. Enter your full Gmail address and the App Password 5. Click **Connect** AutoReach validates SMTP and IMAP access before saving. If validation fails, the error will explain why (usually IMAP disabled, App Password rejected, or admin block on Workspace). ### Workspace (G Suite) Notes Some Workspace admins block App Passwords or IMAP at the org level. If your admin has these locked down: * Ask the admin to allow App Passwords for your account, **or** * Ask the admin to enable IMAP for your account Contact support if you need an alternative connection method. *** ## Connecting Outlook Outlook uses Microsoft OAuth. AutoReach never sees your password. ### Setup 1. In AutoReach, open the parent X or LinkedIn account from the **Accounts** page, find the **Email** card on the **Configuration** tab, and click **Connect Outlook** 2. A popup opens to the Microsoft sign-in page 3. Sign in and grant the permissions AutoReach requests (read mail, send mail, manage profile) 4. The popup closes automatically once authorization succeeds If your browser blocks the popup, allow popups for AutoReach and click **Connect Outlook** again. If the popup closes before the handshake completes, the connection list refreshes shortly after to detect any successful connection. ### Token Refresh Outlook tokens are refreshed automatically in the background. If a token can no longer be refreshed (revoked permission, password changed, MFA reset), the account shows an **Expired** badge and you'll need to **Reconnect**. *** ## Managing Email Accounts Open the parent X or LinkedIn account from the **Accounts** page and use the **Email** card on the **Configuration** tab. The card lists every mailbox connected under that parent account with provider icons (Gmail / Outlook), the email address, a primary badge when applicable, and per-mailbox controls. ### Available Actions | Action | Effect | | ---------------- | ---------------------------------------------------------------------------- | | **Pause** | Temporarily stops outbound sends from this mailbox without disconnecting | | **Resume** | Re-enables sends | | **Make Primary** | Sets this account as the default for recipients on neutral domains | | **Reconnect** | Refresh credentials when a mailbox shows as expired | | **Disconnect** | Permanently remove the mailbox; in-flight email actions on this mailbox stop | ### Multiple Mailboxes and Domain Routing When you have **both Gmail and Outlook connected**, AutoReach picks which mailbox to send from based on the recipient's email domain: * `@gmail.com` and `@googlemail.com` recipients use your Gmail account * `@outlook.com`, `@hotmail.com`, `@live.com`, `@msn.com` (and regional variants) use your Outlook account * For other domains, AutoReach also checks the domain's MX records to detect Google Workspace and Microsoft 365 hosting and routes accordingly * Domains it cannot classify use your **Primary** mailbox This makes outbound mail look more authentic and improves deliverability. Click **Make Primary** on any mailbox to change the default for unclassified domains. *** ## What Happens After Connecting Once a mailbox is connected: * **Email steps in your sequences become available** when you build or edit a sequence * **AutoReach starts polling for replies** in the background * **Bounce detection** runs automatically on every reply * **Inbound emails appear in the unified Inbox** alongside X and LinkedIn conversations You can connect mailboxes at any time. Existing sequences pick up the new mailbox automatically the next time an email step runs. *** ## Troubleshooting **Gmail connection fails with "IMAP access disabled"** * Visit Gmail settings > Forwarding and POP/IMAP and enable IMAP. For Workspace, your admin may need to enable it. **Gmail connection fails with "Authentication failed"** * Make sure 2-Step Verification is on and you used an App Password (not your regular password). Regenerate the App Password if needed. **Outlook popup closes immediately** * Check that popups are allowed for AutoReach. Try again. The page will reload to detect any successful connection. **Mailbox shows "Expired" status** * Click **Reconnect** on the mailbox card. For Gmail, regenerate the App Password if it was revoked. For Outlook, sign in again to renew the token. **Replies aren't appearing in the Inbox** * Confirm the mailbox is **Active** (not Paused or Expired) * Replies are polled periodically, so allow a short delay * Check that the original outbound email was actually sent (visible in the conversation timeline) *** ## Next Steps * [**Email Channel and Email Steps**](/outreach-and-sequences/email-channel): Add email steps to sequences and configure templates * [**Inbox & Real-Time Messaging**](/ai-and-conversations/inbox): How email conversations show up alongside X and LinkedIn * [**Email Finding**](/enrichment/optional-enrichment): Discover work emails for leads who don't have one yet # Creating Your First Offer An **Offer** is the foundation of AutoReach. It describes your product or service, your target audience, and your outreach goals. AutoReach uses your Offer to power AI personalization, lead scoring, and DM generation. ## What Is an Offer? Think of an Offer as your "outreach profile." It contains: * **What you sell**: Your product or service description * **Who you're targeting**: Your ideal customer profile (ICP) * **Context about your business**: Pain points you solve, competitors, locations, industry focus AutoReach's AI uses this information to: 1. **Find the right leads**: Score prospects based on how well they match your ICP 2. **Personalize messages**: Generate DMs that reference the prospect's situation and your Offer 3. **Guide conversations**: Provide context to AI auto-replies so they stay on-brand 4. **Optimize outreach**: Learn what works and suggest refinements ## Creating Your Offer Go to **Offers** and click **New Offer**. You'll see two modes at the top: * **Website** (default)- Paste your website URL and click **Extract**. AutoReach's AI visits your site and extracts your product description, target audience, pain points, competitors, and industries. Review and adjust what's needed, then save. * **Manual**- Fill in all the details yourself from scratch. ### AI-Assisted Creation Most fields include a **Generate** (or **Regenerate**) button that uses AI to create or refine content based on your description. The description field has an **Enhance** button that rewrites your text to be clearer and more specific. You can regenerate individual fields as many times as you like without affecting the others. For new offers, the **Target Audience** field starts as a **Generate from description** button. Click it to have AI create your target audience from your description- you can then edit the result. This is required before you can save. > **Tip:** Start by pasting your URL and letting the AI do the first pass. Then spend a few minutes reviewing and tweaking - this produces better results than writing everything from scratch. ## Offer Fields ### Name The name of your offer or product. Keep it short and clear. **Examples**: * "SaaS Security Audits" * "Fractional CFO Services" * "AI Content Writing Tool" ### Description A detailed description of what you do, what you offer, and your pricing. Be specific about the transformation or value you provide. You can use the **Enhance** button to improve your description with AI. This field supports longer descriptions. Include your value proposition, key features, and pricing details so AutoReach's AI has full context for personalization. **Good descriptions include**: * What your product or service does * Key benefits and differentiators * Pricing structure and what's included * Any relevant details for outreach conversations **Avoid**: * Generic descriptions ("We're a software company") * Marketing fluff ("Industry-leading solution") * Jargon without context ("SaaS optimization platform") ### Target Audience Who you're trying to reach (required). For new offers, this starts as a **Generate from description** button- click it to have AI create your ICP, then edit the result. Use **Regenerate** to get a fresh suggestion. You can be as descriptive as you want. This is not just a list of titles. You can include who to target, who to avoid, and any other context that helps AutoReach understand your ideal buyer. For example, you can write "Avoid freelancers and solopreneurs" or "Only target companies with 50+ employees" or "Do not reach out to agencies." **Example**: > Owners, Directors, MDs, CROs, VP Sales, VP Business Development, VP-level leaders responsible for lead generation or pipeline growth, Sales Directors, Head of Revenue, Head of Demand Gen, SDR Directors, Sales Managers, and other senior leaders directly accountable for outbound pipeline at B2B SaaS companies and professional services firms. Avoid freelancers, solopreneurs, and companies with fewer than 10 employees. ### Pain Points You Solve A list of problems your offer fixes. These help AI match leads who mention these issues on social media and personalize messages accordingly. Type a pain point and click the **+** button (or press Enter) to add it. Use the **Generate** button to have AI suggest pain points. **Examples**: * "Lead generation quality is inconsistent" * "Building sales funnels takes too long" * "Hard to identify high-intent buyers at scale" * "Outbound outreach timing is guesswork" * "Too much time spent on unqualified leads" ### Known Competitors Companies or tools that compete with you. This helps AI detect leads who engage with competitors and understand your market position. Type a competitor name and click the **+** button (or press Enter) to add it. Use the **Generate** button to have AI suggest competitors. **Examples**: * Common Room * 6sense * ZoomInfo Copilot * Clay ### Target Industries LinkedIn industries to filter people search. These are used in "By Role" and Lookalike searches to find prospects in specific sectors. Use the **Generate** button to have AI suggest industries. **Examples**: * Computer Software * Marketing and Advertising * Information Technology and Services * Management Consulting * Staffing and Recruiting ### Deal Value The monetary value of your offer (e.g., $5,000). This helps AutoReach estimate pipeline value. ### Locations Geographic regions where you want to focus outreach. Toggle between **Lead** location and **Company HQ** location to control how leads are matched. Search and add countries, regions, or cities. Leave blank for global targeting. ### Target Language The language for your outreach messages. Choose from English, Italian, Spanish, French, German, or Portuguese. AutoReach will generate DMs and responses in this language. ### Website Your website URL (e.g., ). Used by the Enhance and Generate features to extract business details. ### Active Toggle Toggle your offer on or off. Only active offers are used by Autopilot and sequences for lead scoring and outreach. ## Knowledge Base Upload your strategy docs, case studies, and other context. AutoReach's AI uses these documents to generate personalized, contextual responses. ### What to Upload Upload up to 10 documents (PDF, DOCX, or TXT) with: * **Case studies**: How you solved problems for past clients * **Playbooks**: Your outreach strategy, sales process, or methodology * **Price sheets**: Pricing information (optional, for discovery calls) * **Product guides**: Overview of your product or service * **Strategy docs**: How your offer solves specific problems ### How to Upload 1. On the Offers page, click **Knowledge Base** on your offer card 2. Click or drag a file into the upload area (PDF, DOCX, or TXT, max 10MB) 3. The document is processed automatically- you'll see it move from "Processing" to "Ready" AutoReach will: * Chunk your document into sections * Generate vector embeddings for each chunk * Index them for semantic search When someone replies to your outreach, AutoReach's AI will search your knowledge base for relevant context and use it to generate smart, personalized responses. > **Tip:** The more detailed your knowledge base, the better AutoReach's AI performs. Include real examples of your work, your problem-solving approach, and customer success stories. ### How Knowledge Base Powers AI Responses When a prospect replies, AutoReach: 1. **Searches your knowledge base** for relevant sections (e.g., if they ask about pricing, it finds your price sheet) 2. **Includes the context** in the AI prompt 3. **Generates a response** that is personalized and grounded in your actual strategy **Example**: * Prospect asks: "How do you handle data security?" * AutoReach searches your knowledge base and finds your security playbook * AI generates a response citing your specific approach * Response is sent automatically or queued for your review ## How Your Offer Powers AutoReach ### 1. Lead Scoring When a lead enters the system, AutoReach scores them based on: * Profile data (role, company, location) * Social signals (recent posts, engagement) * Match to your Offer's target audience and pain points Higher scores mean a stronger fit with your ICP. ### 2. Keyword Generation AutoReach automatically generates search keywords from your offer's description and pain points, then uses them to: * Search for relevant prospects on X and LinkedIn * Filter leads in Autopilot ### 3. DM Personalization When sending DMs, AutoReach: * Looks up the prospect's profile and recent activity * References relevant pain points from your Offer * Crafts a personalized opener matching your outreach tone ### 4. AI Response Context When prospects reply, AutoReach: * Searches your knowledge base for context * Generates smart replies that stay on-brand * Maintains your outreach voice across conversations ## Quality Bar: Weak vs. Strong Examples Quick gut-check before saving. If your fields read like the weak column, the AI will produce generic outreach. | Field | Weak | Strong | | --------------- | ------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- | | Description | "We help businesses with automation." | "We automate repetitive B2B workflows using AI. Save your team 20+ hours/week on email, data entry, and report generation." | | Target Audience | "Marketing professionals" | "VP of Demand Gen at B2B SaaS companies with $10M+ ARR, focused on ABM strategy" | | Pain Points | "Help with revenue" | "Reduce time to close deals", "Improve sales team visibility", "Decrease win-loss variance" | For broader strategy on offer setup, knowledge base content, and quarterly review cadence, see [Best Practices: Setting Up Your Offer](/best-practices#1-setting-up-your-offer). ## Next Steps 1. **Configure your accounts**: Go to the Accounts page to enable inbound engagement detection, set your limits, and start the Engagement Engine 2. **Start Autopilot**: Let AutoReach find and engage leads automatically 3. **Review your Sequence**: Check the prompt, configuration, and tone examples in Advanced Settings See [Quickstart](/getting-started/quickstart) for the full setup flow. # Installing the Chrome Extension The AutoReach Chrome Extension is the central tool for connecting your social accounts, managing your sales pipeline, and using AI-powered messaging directly from your browser. ## Installation 1. Visit the [AutoReach Chrome Extension](https://chromewebstore.google.com/detail/autoreach/gokfpebaahgioniohencjfoknidmadmh) page on the Chrome Web Store 2. Click **Add to Chrome** 3. In the popup, click **Add extension** 4. The AutoReach icon will appear in your Chrome toolbar ## License Key Activation After installation, activate your license key: 1. Click the **AutoReach icon** in your Chrome toolbar 2. Paste your **License Key** (generated during onboarding or found at [autoreach.tech/settings](https://autoreach.tech/settings)) 3. Click **Activate** > **Important:** Keep your license key private. Never share it with others or post it online. ## Platform Modes The extension automatically detects which platform you are on and adjusts its interface. **X (Twitter) now has full feature parity with LinkedIn**: the same CRM pipeline, AI messaging assistant, profile lead capture, and account management are available on both platforms. * **LinkedIn**: Full CRM pipeline, AI messaging assistant, profile "Add to Leads" button, comment/post reply assist, connection tracking, account management, and proxy detection * **X/Twitter**: Full CRM pipeline, AI messaging assistant, profile "Add to Leads" button, tweet reply assist, DM reply assist, account management, and proxy detection * **Instagram**: Account connection and management (connect, enable/disable, reconnect) from the extension popup * **Other sites**: Extension popup is not active *** ## Features on Both X and LinkedIn The features below work the same way on both platforms. Subsequent sections only flag platform-specific differences. ### Account Connection * **X**: Enter your name and a 4-digit PIN (used for DM Chat encryption). Cookies extracted automatically. * **LinkedIn**: Enter your first and last name. Cookies extracted automatically. * See [Connecting Accounts](/getting-started/connecting-accounts) for details. ### Account Management * **Reconnect**: When cookies expire, click the three dots next to the account and click Reconnect * **Enable/Disable**: Toggle accounts on or off * **Edit PIN** (X only): Update your 4-digit DM encryption PIN * **Status display**: Healthy, Inactive, Expired, Paused, Rate Limited ### Proxy Detection * The extension detects and displays your current proxy configuration * Supports AutoReach provisioned proxies and external proxy managers (GoLogin, Multilogin) ### First-Run Tour The first time you open the extension on a new platform (X or LinkedIn), an interactive 4-step tour walks you through the profile button, message assist buttons, and sequence picker. *** ### CRM Pipeline (Kanban Board) A full sales pipeline built into the extension popup. View and manage all your leads in a Kanban-style board. #### Pipeline Stages | Stage | Description | | ------------- | ----------------------------------------------------------- | | **New** | Lead added to your CRM | | **On Hold** | Lead temporarily paused (cooldown period after withdrawal) | | **Requested** | Connection request sent (auto-detected) | | **Accepted** | Connection request accepted (auto-detected or manual check) | | **Contacted** | You sent a DM (auto-detected) | | **Replied** | Lead replied to your message (auto-detected) | | **Meeting** | Meeting booked (AI-suggested, confirm to move) | | **Won** | Deal closed successfully (AI-suggested, confirm to move) | | **Lost** | Opportunity over (AI-suggested, confirm to move) | You can also create **custom stages** that appear between Replied and Meeting. Custom stages can be renamed, reordered, and deleted. #### Pipeline Features * **Search**: Filter leads by name, headline, or company * **Stage changes**: Move leads between stages via dropdown * **Delete leads**: Remove leads with confirmation * **Infinite scroll**: Loads more leads as you scroll within each column * **Pipeline stats**: Total leads, reply rate, and follow-up count displayed at the top * **Per-account view**: Switch between LinkedIn accounts to see each account's pipeline * **Active sequence badges**: See which sequence a lead is currently in *** ### Automatic Stage Detection The CRM automatically detects and moves leads through stages based on your activity on LinkedIn. You do not need to update stages manually. #### Connection Request Sent When you send a connection request manually on LinkedIn, the extension intercepts it and: * If the person is already a lead: moves them to **Requested** * If they are not a lead: creates a new lead and sets them to **Requested** * Captures the invitation URN for later withdrawal tracking #### DM Sent When you send a DM on LinkedIn, the extension detects the recipient and: * Leads in New/Requested/Accepted are moved to **Contacted** * Leads already in Contacted are checked for replies * Leads in later stages get an activity timestamp update #### Reply Detection The extension continuously monitors open conversations. When a lead replies: * The lead is automatically moved from **Contacted** to **Replied** * A toast notification confirms the detection * Uses fuzzy name matching to handle name variations, abbreviations, accents, and Cyrillic characters #### Deal Signal Detection (Meeting, Won, Lost) When a conversation is open, the extension uses AI to analyze the lead's messages for deal progression signals: * If a meeting signal is detected: a popup suggests moving the lead to **Meeting** * If a won or lost signal is detected: a popup suggests moving to **Won** or **Lost** * You confirm or dismiss the suggestion. The popup auto-dismisses after 10 seconds. *** ### Add to Leads On any LinkedIn or X profile page, the extension injects an **Add to Leads** button directly into the profile actions row (next to Connect/Message/More on LinkedIn, next to Follow/Message/More on X). * Click **Add to Leads** to create a new lead from the profile * The extension automatically scrapes: name, handle/headline, company, location, bio, and profile image * If the person is already in your CRM, the button shows **In CRM** *** ### Check Connection Acceptances Click **Check Acceptances** in the CRM to manually check which connection requests have been accepted. * Fetches your recent LinkedIn connections via the API * Cross-references them with leads in the **Requested** stage * Accepted leads are automatically moved to the **Accepted** stage * Uses fuzzy name matching to handle variations *** ### Follow-Up Management The extension tracks which leads need follow-up: * Leads in Contacted or Replied stages that exceed your configured follow-up threshold are flagged * A **follow-up filter** toggle lets you see only leads that need follow-up * Set a per-lead follow-up date: Tomorrow, In 3 days, In 1 week, In 1 month, or a custom date * Generate an AI follow-up message with one click * Configure the default follow-up days (1-30) in Settings *** ### Connection Withdrawal Manage pending connection requests that have not been accepted: * Configure a withdrawal threshold (number of days before a pending request should be withdrawn) * The extension checks for leads past the threshold and shows a withdrawal modal * Withdraw connections in batch (executed with delays for safety) * Withdrawn leads are moved to the **On Hold** stage (cooldown period before they can be re-contacted) *** ### AI Messaging Assistant On both LinkedIn and X, the extension injects an AutoReach button into the message composer (LinkedIn DMs, X DMs, and X tweet reply composers). Click it to open the assistant menu with six options: #### 1. First Message Generate a personalized opening message for a new contact. * **Template mode**: Uses your saved first message template with basic personalization * **Enhanced mode**: AI strategically crafts the message using all available context * If the lead is in your CRM: shows their matched post/comment, buyer score, recent posts, and lets you generate from that context * If the lead is not in your CRM: generates from their headline * Displays lead intelligence: buyer probability, fit/intent/timing scores, suggested play, signal date #### 2. Reply Suggestion Get an AI-powered reply for an ongoing conversation. * Parses the full conversation history from the page * Sends conversation context plus lead data (CRM stage, matched post, buyer intelligence) to AI * Detects objections in the lead's messages and surfaces an objection badge #### 3. Follow Up Generate a follow-up message when the lead has not replied. * Uses full lead context: matched post, ICP match reason, buyer intelligence, recent posts * Includes your currently selected offer for context #### 4. Proofread Improve a message you have already drafted. * **Shorten**: Make your draft more concise and punchy * **Enhance**: Make your draft more compelling and persuasive * The result popup also offers Shorten and Regenerate options #### 5. Analyze Lead Opens the SDR Co-pilot with the prompt "Tell me everything you know about this person." Gives you a complete breakdown of the lead based on all available data. #### 6. SDR Co-pilot A persistent chat interface for freeform AI conversation about the current lead. * Ask anything about the lead: strategy questions, objection handling, next steps * Context includes: conversation history, CRM stage, buyer intelligence, matched post, ICP match reason, recent posts, company, bio, notes * Chat history persists per lead across sessions * Responses rendered with markdown formatting #### Suggestion Popup All AI-generated messages appear in a shared suggestion popup with these actions: * **Insert**: Types the message into LinkedIn's input with a typewriter effect * **Copy**: Copies to clipboard * **Regenerate**: Get a new variant * **Shorten**: Make it more concise * **Custom Edit**: Type a free-form instruction (e.g., "make it more casual") and AI will adjust *** ### Post / Tweet and Comment Reply Suggestions The extension injects AI assist buttons into the comment/reply composers on both platforms. * **LinkedIn:** AutoReach button next to any comment input on a feed post. If the post has highlighted comments, a picker lets you choose whether to reply to the post or to a specific comment. * **X:** AutoReach button on the tweet reply composer. Reads the tweet context and (where present) the parent thread. * AI generates a contextual comment/reply based on the post content, your offer, and the active sequence's prompt *** ### Analytics Click the analytics icon in the CRM to see a pipeline overview: * Counts per stage * Conversion funnel (Lead > Contacted > Replied > Meeting > Won rates) * Follow-up alerts with clickable links to filter your pipeline *** ### Settings The Settings tab (available on both LinkedIn and X) includes: * **Reply prompt**: Configure the AI reply prompt * **First message template**: Set your default outreach template * **Objection handling**: Add trigger/response pairs for common objections * **Follow-up days**: Set the default follow-up threshold (1-30 days) * **Withdrawal days**: Set how long before pending connections are flagged (1-30 days) * **Custom stages**: Add, rename, reorder, and delete custom pipeline stages * **Offers**: Select which offer to use for AI-generated messaging All settings are saved with a unified save bar that detects unsaved changes. *** ## Troubleshooting ### "License Key Invalid" Your license key has expired or is incorrect. 1. Go to [autoreach.tech/settings](https://autoreach.tech/settings) 2. Copy your current license key 3. Paste it into the extension 4. Click **Activate** ### "Extension Can't Access This Page" The extension only works on linkedin.com, x.com/twitter.com, and instagram.com. ### Session Expired 1. Open the **AutoReach Chrome Extension** 2. Click the **three dots** next to the account 3. Click **Reconnect** Make sure you are logged into the social account in your browser before reconnecting. ### "Connection Not Detected" LinkedIn connection acceptance detection requires: 1. The extension to be running 2. Your LinkedIn tab to stay open 3. The pending connection to still be in your CRM in the Requested stage If detection is not working, use the **Check Acceptances** button in the CRM to manually trigger a check. ## Privacy and Security * All CRM data is stored in your AutoReach account (encrypted) * License key is stored locally in your browser (not shared) * Extracted cookies are transmitted only to AutoReach servers (HTTPS encrypted) * We never log your browsing history or profile views * Keep your extracted cookies private * Cookies expire over time (2-4 weeks). Reconnect via the extension (three dots > Reconnect). Found a security issue? Email . ## Next Steps * [**Connecting Accounts**](/getting-started/connecting-accounts): Set up your X/Twitter and LinkedIn accounts * [**Creating Your First Offer**](/getting-started/create-offer): Define your target audience and outreach goals * [**Quickstart**](/getting-started/quickstart): Follow the complete setup guide # Getting Results (Day 2+) You have launched your first sequence. Now what? This guide covers what to expect in the first 24-48 hours and how to iterate toward better results. *** ## What to Expect in the First 24-48 Hours After starting a sequence, AutoReach begins executing actions according to the delays you configured in your flow. Here is a typical timeline: * **Immediately**: The first step (usually a like) starts executing for enrolled leads * **Within minutes to hours**: Subsequent steps fire according to the delays between nodes (e.g., follow after 15 minutes, DM after 1 day) * **Within 24 hours**: Your first DMs go out (assuming a 1-day delay before messaging) * **24-48 hours after DMs**: First replies start arriving Do not expect instant results. Social outreach is a slow burn - most prospects need to see your name a few times before they engage. > **Tip:** Check the Activity Feed on the **Dashboard** to see exactly what actions AutoReach has completed and when. *** ## Reading Your First Results Open your dashboard to see your pipeline metrics. The key numbers to watch early on are: * **Contacted** - confirms DMs are actually being delivered * **Replied** - the metric that matters most; aim for 5-15% reply rates on cold outreach * **Meetings Booked** - the ultimate conversion (this may take a few days to appear) If you see "Contacted" climbing but "Replied" stuck at zero after 48 hours, see the troubleshooting section below. *** ## When Replies Come In Replies appear in your **Inbox** (Conversations section). Each reply shows: * The lead's profile information * The full conversation thread * The AI's suggested response (if AI responses are enabled) ### Handling Conversations * **AI is on by default**: If you left AI responses enabled, AutoReach will automatically reply to incoming messages using your offer context and tone examples. Review the AI's responses early on to make sure the tone matches what you want. * **Take over manually**: Click on any conversation to read the thread. You can toggle AI off for that conversation and respond yourself whenever a lead shows genuine buying interest. * **Add tone examples**: If the AI's tone is not quite right, go to your sequence settings and add tone examples - these teach the AI how you actually talk. * **Meeting tracking**: When a lead books a call through your Calendly or Cal.com link, AutoReach automatically marks the lead as Meeting Booked via webhook. If you use a custom booking URL (no webhook), open the lead in the Chrome Extension CRM panel and drag it to the Meeting stage so the metrics update. > **Tip:** For your first few days, manually review every AI response before the next one goes out. This helps you tune the AI's style and catch anything off-brand. *** ## Adjusting Your Sequence if Reply Rates Are Low If your reply rate is below 3% after the first 50-100 DMs, try these adjustments. **Start at the top**: message quality and targeting account for most low-reply problems. Timing and channel are less common causes but worth checking if the first three don't help. For a deeper diagnostic flow, see [Low Reply Rates in Troubleshooting](/troubleshooting#low-reply-rates). ### Check Your Message Quality Use the **Simulation** feature to preview what your DMs look like for specific leads. If they sound generic, add more personalization to your DM template or refine your AI prompt. ### Check Your Targeting Review your scored leads. If most leads have low buyer scores, your ICP definition may be too broad. Tighten your offer's target audience, pain points, and qualifying criteria. ### Add a Warmup Before the DM If you are going straight to DM, consider adding engagement steps first (like, follow, reply to a post). Leads are more likely to respond to someone whose name they recognize. ### Adjust Timing If your DMs are going out at odd hours, review the activity window on the sending account (Accounts → \[account] → Configuration → Activity Window). Each account has its own window, so check the right one. Messages sent during business hours tend to get higher response rates. ### Try a Different Channel If X DMs are not working, try LinkedIn (or vice versa). Some audiences are more responsive on one platform than the other. *** ## When to Add More Leads Add more leads when: * Your current batch is mostly processed (most leads have reached the end of the sequence) * Your reply rate is healthy (3%+ for cold outreach) and you want more volume * You have refined your messaging and are confident in the quality You can add leads from any source: run a new search, import a CSV, or let Autopilot find them automatically. > **Warning:** Do not add thousands of leads to a brand new sequence. Start with 50-100 leads, verify your messaging works, then scale up. *** ## Setting Up Autopilot Once your manual sequences are performing well, enable Autopilot to automate the growth loop: 1. **Auto-Enrollment**: Automatically enroll newly scored leads into your active sequence. Go to the sequence settings and enable auto-enrollment with a minimum buyer score threshold. 2. **Buyer Expansion**: Enable daily recurring searches or follower extractions to continuously find new leads that match your ICP. 3. **AI Responses**: If you are comfortable with how the AI handles conversations, leave it enabled for hands-free follow-up. See [Autopilot Overview](/autopilot/overview) for full details. *** ## Enabling the Engagement Engine The Engagement Engine builds your brand presence by engaging with relevant content on X and LinkedIn. It operates separately from sequences and does not send DMs - it likes and comments on posts from people in your target audience. Why this matters: * Leads are more likely to reply to someone they have seen engaging thoughtfully in their feed * Consistent engagement builds credibility and warms up your accounts * It runs automatically in the background Enable it from your account's detail page (go to **Accounts**, click your account, and find the Engagement Engine section). Start with approval mode so you can review each engagement before it goes out. See [Engagement Engine Overview](/engagement-engine/overview) for setup instructions. *** ## Common "Why Isn't This Working?" Scenarios ### "My sequence is active but no actions are happening" * Check that your connected accounts are healthy (no paused or error status on the **Accounts** page) * Verify you have leads enrolled in the sequence (open the sequence and check the Leads tab) * Check the activity window on the sending account (Accounts → \[account] → Configuration) - actions only execute during the account's configured hours * Make sure your daily action limits have not been reached for today ### "DMs are sending but nobody is replying" * Review message quality with Simulation - personalization may be weak * Check that your leads are relevant (high buyer scores) * Verify your DMs are not landing in spam (check from the recipient's perspective if possible) * Consider adding warmup steps before the DM * Wait at least 48 hours - some people take time to respond ### "Leads are not getting scored" * Verify your offer is fully configured with ICP details, pain points, and qualifying criteria * Check the health warnings on the Dashboard - your API keys may have run out of credits * Look at the lead's profile - if enrichment has not completed, scoring cannot run yet ### "AI messages look generic or off-brand" * Add tone examples to your sequence (go to the Tone Examples section in sequence settings) * Customize the AI prompt and DM generation prompt with specific instructions * Use the Simulation tool to iterate on message quality before sending ### "Account shows a warning or paused status" * Check the account health card on the **Accounts** page for the specific error * If rate-limited, AutoReach automatically applies cooldown periods to protect your account. Wait for the cooldown to expire * If cookies expired, reconnect via the Chrome Extension (click three dots > Reconnect) * Review the [Account Safety](/settings-and-configuration/account-safety) guide for prevention tips *** ## Next Steps * [**Building Sequences**](/outreach-and-sequences/building-sequences): Refine your sequence flow and add new steps * [**DM Personalization**](/outreach-and-sequences/dm-personalization): Improve your DM templates and AI prompts * [**Simulation & A/B Testing**](/outreach-and-sequences/simulation-and-testing): Preview messages before sending * [**Autopilot Overview**](/autopilot/overview): Automate lead discovery and enrollment * [**Engagement Engine**](/engagement-engine/overview): Build brand presence through content engagement # Overview Welcome to AutoReach's core concepts guide. This section explains the fundamental building blocks that power our B2B outreach automation platform for X/Twitter, LinkedIn, and Instagram. ## The AutoReach Ecosystem AutoReach connects six core concepts into an intelligent outreach machine. Understanding how they work together is key to getting the most out of the platform. ### The Six Core Concepts **Offers** are the foundation. An offer represents the specific product or service you're selling. It is not just a description: it is the knowledge base that powers everything else on the platform, from lead scoring to personalized messaging. **Leads** are the people you're trying to reach. AutoReach automatically discovers them across X, LinkedIn, and Instagram, and can unify a lead's X and LinkedIn profiles into a single view. Each lead gets enriched with education, experience, signals, and social activity data. **Buyer Intelligence** is the AI-powered brain of AutoReach. It analyzes each lead across three dimensions (fit, intent, and timing) to surface leads with active buying signals, while ensuring strong fits are never missed. **Sequences** are your multi-step outreach campaigns. Define the actions you want to take (send DM, like post, follow), set the timing, and AutoReach executes them automatically across your enrolled leads. **Conversations** are the intelligent back-and-forth messaging with leads. AutoReach's AI can understand responses, generate contextual replies, and escalate to you when needed, all while maintaining your authentic voice. **Autopilot** is the quick-start assistant. It automatically configures searches, creates a sequence, and enrolls leads for you. You can do all of this manually, but Autopilot handles the initial setup so you can get started faster. ## The Outreach Lifecycle Here's how these concepts work together in practice: ``` ┌─────────────┐ │ 1. Discover│ ← Find leads matching your Offer via searches └──────┬──────┘ │ ▼ ┌─────────────┐ │ 2. Enrich │ ← Cross-platform profiles unified, signals detected └──────┬──────┘ │ ▼ ┌─────────────┐ │ 3. Score │ ← Buyer Intelligence calculates fit, intent, timing └──────┬──────┘ │ ▼ ┌─────────────┐ │ 4. Engage │ ← Sequences deliver personalized outreach └──────┬──────┘ │ ▼ ┌─────────────┐ │ 5. Respond │ ← Conversations use AI for intelligent replies └──────┬──────┘ │ ▼ ┌──────────────┐ │ 6. Resurface│ ← Monitor signals, re-engage at the right moment └──────┬───────┘ │ ▼ ┌──────────────┐ │ 7. Book │ ← Qualified leads become meetings └──────────────┘ ``` ## Why These Concepts Matter Each concept builds on the others: * Without **Offers**, you have no signal keywords, so scoring is generic * Without **Leads**, you have no one to contact * Without **Buyer Intelligence**, you're reaching everyone equally * Without **Sequences**, you have no consistent outreach cadence * Without **Conversations**, you can't scale personal responses * Without **Autopilot**, you configure searches and sequences yourself (which is totally fine) > **Tip:** Start by creating a detailed Offer. The richer your Offer definition (target audience, pain points, competitors), the smarter your Buyer Intelligence scores become. ## What's Next? * [**How Leads Work**](/core-concepts/leads) - Discover how AutoReach sources, enriches, and organizes leads * [**Buyer Intelligence**](/core-concepts/buyer-intelligence) - Understand the three-dimensional scoring model, buyer states, and signal detection * [**Offers & Knowledge Base**](/core-concepts/offers-and-knowledge-base) - Deep dive into how offers power everything # How Leads Work A **lead** in AutoReach is a profile of a potential buyer discovered across X, LinkedIn, or Instagram. Leads are automatically discovered, enriched, scored, and tracked throughout the entire outreach lifecycle. AutoReach can also unify a lead's X and LinkedIn profiles into a single cross-platform view. ## What Is a Lead? A lead represents one person. For X and LinkedIn, AutoReach can match the two profiles to create a unified view, giving you: * **Social presence** across both platforms * **Professional history** (education, experience, skills) * **Company and role** information * **Recent activity** and engagement patterns * **Intent and fit signals** extracted from content * **Buyer score** predicting purchase probability ## Where Do Leads Come From? AutoReach discovers leads through several different sources: ### 1. Tweet Search Search for keywords on X using natural language. AutoReach finds tweets matching your keywords and enrolls the authors as leads. **Example:** Search for "looking for workflow automation tool" finds people actively discussing this problem. ### 2. LinkedIn Search Multiple LinkedIn discovery methods covering content, people, companies, and hiring signals. * **Content search:** Find people posting about keywords (hiring, switching to competitor, etc.) * **People search (By Role):** Target by job title, location, industry, and other filters * **Company search (By Company):** Find companies matching descriptors (e.g., "B2B SaaS Sales Tooling") and add the top decision-maker from each as a lead * **Job search (Hiring Signals):** Find companies actively hiring for relevant roles, then extract the most senior decision-maker at each. Integrated into Content Search as the Hiring Signals intent category ### 3. Instagram Search Discover prospects on Instagram by mining hashtags, follower/following lists, post likers and commenters, and lookalike accounts. Instagram leads flow into the same enrichment and scoring pipeline as X and LinkedIn. ### 4. Lookalike Audiences AutoReach uses AI to find influencers, thought leaders, publications, and communities whose audiences match your ideal customer profile. It then extracts their audience (followers on X and Instagram, post commenters on LinkedIn) and scores them against your ICP. You do not upload a list. AutoReach discovers the right seed accounts automatically based on your offer definition. **Example:** If you sell to B2B SaaS leaders, AutoReach might find an industry newsletter whose followers include VPs of Sales and Directors of Revenue Operations. It extracts those followers and scores them as leads. ### 5. Follower Extraction Extract followers (or following) from specific X or Instagram accounts you choose as seed accounts. AutoReach processes the profiles and scores them against your offer. (On LinkedIn, lookalike-style extraction works off post commenters rather than followers.) ### 6. Comment Extraction AutoReach can extract leads from the comments on tweets, LinkedIn posts, or Instagram posts. When a post has high engagement from your target audience, the commenters are captured and scored as leads. ### 7. Link Extraction AutoReach extracts profile links found during searches and other discovery methods. These are processed and added as leads. ### 8. Lead Pool AutoReach maintains a shared lead pool of pre-enriched profiles. When you create an offer, the system uses vector embeddings to match existing leads against your ICP. This is the fastest way to get scored leads because they skip enrichment entirely. ### 9. Chrome Extension On LinkedIn, you can click the **Add to Leads** button on any profile page to add them directly to your CRM via the Chrome Extension. ### 10. Manual Add Add individual leads manually by URL or username. ### 11. CSV Import Bulk import leads from a spreadsheet with X handle, LinkedIn URL, Instagram profile, or email. > **Note:** Lead pool matches are instant because they are pre-enriched. Intent-based sources (Tweet Search, LinkedIn Content Search) are next fastest. CSV imports take longest because they need full enrichment from scratch. ## Lead Enrichment When a lead enters AutoReach, it goes through automatic enrichment in the background: 1. **Profile enrichment:** Enrich LinkedIn and X profiles in parallel- extract bio, headline, experience, education, skills, and company data. 2. **Activity enrichment:** Fetch recent posts, engagement patterns, and social activity 3. **Location enrichment:** *(conditional)* Resolve lead location when geographic targeting is enabled 4. **Scoring:** Run Buyer Intelligence to calculate fit, intent, and timing scores As each stage completes, the lead's profile gets richer. You can view partial data during enrichment. Scoring runs after enrichment completes, but partial enrichment failures do not block scoring- the lead is scored with whatever data is available. > **Warning:** Enrichment speed varies. X and LinkedIn rate limiting can affect enrichment speed. Large batches may take time to fully enrich. ## Lead States Every lead has a buyer state that determines where it appears in the UI and whether it is eligible for outreach: | State | UI Tab Label | Description | Where It Appears | | ------------------- | ---------------- | ------------------------------------------ | ------------------------------ | | **Active** | **Ready** | High-scoring leads ready for outreach | Buyers page (Ready tab) | | **Monitor** | **Emerging** | Promising leads not yet ready for outreach | Buyers page (Emerging tab) | | **Poor Fit** | **Low Priority** | Low-scoring leads unlikely to convert | Buyers page (Low Priority tab) | | **Disqualified** | **Disqualified** | Leads that do not match your ICP | Buyers page (Disqualified tab) | | **Not Scored** | **Analyzing** | Enrichment not yet complete | Buyers page (Analyzing tab) | | **Manual Outreach** | **Manual** | User override - treated as Active | Buyers page (Manual tab) | See [Buyer Intelligence](/core-concepts/buyer-intelligence) for full details on each state and how transitions work. ## Cross-Platform Profile Matching You can search for a lead's profile on the other platform by selecting leads, clicking Enrich, and enabling "Find LinkedIn" or "Find X Profile" under Profile Discovery. This cross-platform data makes enrichment and scoring far more accurate than single-platform data alone. See [Lead Pool](/finding-leads/lead-pool) for details. ## Lead Profile Data Each lead's profile includes: ### Basic Info * Name, headline, bio * X and LinkedIn URLs * Email (if found via web enrichment) * Location (city, state, country) * Company and job title ### Professional Background * Education (school, degree, field, dates) * Work experience (company, role, tenure, description) * Skills (extracted from profiles) ### Social Activity * Recent posts and engagement * LinkedIn post activity * Follower/following counts ### Signals and Scoring * All detected intent signals (with dates and strength) * Company-level signals * Buyer state (Active, Monitor, Poor Fit, Disqualified) * Buyer score, fit score, intent score, timing score ## Filtering Leads The All Leads view and Sequence Leads view support filtering by: * **Buyer state** (Ready, Emerging, Low Priority, etc.) * **Connection status** (Pending (Sent), Connected, Pending (Received), Not Connected) - filter by LinkedIn connection state * **Next action** - filter sequence leads by their upcoming action type Connection status badges appear inline next to lead names for quick visibility. ## Lead Lifecycle in a Sequence Once enrolled in a sequence, a lead progresses through these statuses: | Status | Meaning | | ------------------ | ---------------------------------------------------------------------- | | **Pending** | Enrolled but not yet started | | **In Progress** | Currently progressing through sequence steps (internal status: active) | | **Paused** | Manually paused mid-sequence | | **Replied** | Lead replied to a message | | **Meeting Booked** | Meeting scheduled | | **Completed** | Finished all sequence steps | | **Lost** | Lead went cold or opportunity closed without conversion | | **Failed** | Error during execution (invalid profile, rate limit) | ## Best Practices > **Tip:** 100 highly-scored leads from LinkedIn search will outperform 1,000 random CSV imports. Focus on signal-rich sources first. > **Tip:** If you are just getting started, the lead pool gives you scored leads instantly without waiting for enrichment. > **Tip:** Do not ignore Poor Fit leads entirely. Re-scoring happens when new signals are detected. A lead can jump significantly based on a job change or new social activity. ## Next Steps * [**Buyer Intelligence**](/core-concepts/buyer-intelligence): Learn how AutoReach scores leads across fit, intent, and timing * [**Finding Leads Overview**](/finding-leads/overview): Explore all the ways to discover new leads # Buyer Intelligence & Scoring AutoReach's Buyer Intelligence system is the heart of smart outreach. Instead of blasting everyone, it scores each lead across three dimensions (fit, intent, and timing), assigns a buyer state, and uses dozens of signals to predict who is most likely to buy right now. > **When to read this page:** You wonder why a specific lead scored low (or high) and want to understand which dimension is dragging the score. You're tuning your offer and want to know how each input affects scoring. Or your reply rates are low and you suspect targeting: this page tells you what AutoReach is actually evaluating, so you know what to fix in your offer. You don't need to read it before you start outreach; the system works automatically. *** ## The 30-Second Version If you only read this section: * Every lead is scored on three dimensions, each 0-100: **Fit** (do they match your ICP?), **Intent** (are they actively looking?), and **Timing** (is now the right moment?). * The three scores combine into a **buyer state**. **Active** = ready for outreach. **Monitor** = watch for new signals (auto-rechecked on a schedule). **Poor Fit** = not a match today, but still re-evaluated when new signals appear. **Disqualified** = filtered out and not auto-rescored. **Manual Outreach** = you overrode the rules. * **Your Offer drives all of this.** Sharper target audience and pain points produce sharper scoring. New leads always score against your current offer. Existing leads behave differently: Monitor and Poor Fit leads get re-evaluated automatically over time; already-enrolled Active leads keep their score until you trigger a manual rescore. * **If a score looks wrong**, open the lead and click **Should be a buyer** / **Should NOT be a buyer**. This launches the [Offer Refinement](#score-feedback) flow, which suggests offer edits to fix the misclassification. * **Where to spend your time:** Active leads convert best. Monitor leads are your near-future Active list; they'll get promoted as new signals appear. Don't manually outreach Poor Fit; let resurfacing handle them, or use Manual Outreach for specific high-value accounts you want to reach despite the filters. The rest of this page is the detail behind those five points. Read on if you want to know exactly what signals feed each score, how state transitions work, or how to tune the system. *** ## The Three Scoring Dimensions Each lead is scored independently on three dimensions, each on a 0-100 scale. ### Fit Score "Is this person in my target market?" Fit measures how well a lead matches your ideal customer profile (ICP): * Right industry? * Right company size? * Right role/seniority? * Geographic match? * Skill/background match? ### Intent Score "Is this person actively looking for a solution?" Intent measures signals that show the lead is searching for, considering, or planning to buy something: * Are they asking for recommendations? * Have they mentioned switching tools? * Are they complaining about current tools? * Engaging with competitors? * Posting about hiring/scaling challenges? ### Timing Score "Is this the right moment to reach them?" Timing measures how urgent or immediate their needs appear: * Did they just change jobs? * Recent company funding/expansion? * New hiring initiative? * Recent product launches? * Active engagement with relevant content? ### Understanding the Scores Each dimension is scored independently. You might have: * **High fit + Low intent:** Someone in your target market who is not actively looking yet * **Low fit + High intent:** Someone actively buying, but not the right persona * **High fit + High intent + High timing:** Your ideal buyer, ready to engage right now The combination of all three determines the lead's composite **Buyer Score** (0-100) and their buyer state. *** ## Buyer States Every lead has a **buyer state** that represents their current status in your pipeline. The state determines where they appear in the UI, whether they are eligible for outreach, and how AutoReach treats them. ### Active (shown as "Ready" in the UI) This person is ready to buy. They meet your ICP, show buying intent, and the timing is right. * Appears on the **Buyers** page, **Ready** tab * Fully eligible for sequence enrollment and auto-enrollment * **Next action:** Add to a sequence and start outreach ### Monitor (shown as "Emerging" in the UI) This person has potential, but is not quite ready yet. Maybe fit is strong but intent is weak, or the timing is not there yet. * Appears on the **Buyers** page, **Emerging** tab * Can be enrolled in sequences, but not recommended for aggressive outreach * AutoReach periodically re-checks these leads for new buying signals and automatically promotes them to Active if they qualify ### Poor Fit (shown as "Low Priority" in the UI) This person does not match your ideal customer profile well. Something significant does not line up - wrong industry, wrong seniority, or insufficient signals. * Appears on the **Buyers** page, **Low Priority** tab * Not recommended for outreach * Still monitored on a longer cycle for signal changes ### Disqualified This person is far outside your ICP with almost no chance of conversion. They remain in the database but are excluded from automatic processing. * Appears on the **Buyers** page, **Disqualified** tab * Must be overridden to Manual Outreach before enrollment * Skipped by the resurfacing scheduler ### Manual Outreach (shown as "Manual" in the UI) You have decided this person is worth reaching out to, overriding the scoring system. * Appears on the **Buyers** page, **Manual** tab * Eligible for sequence enrollment, treated like Active * Protected from automatic pipeline rescoring (unless you explicitly trigger a manual rescore) * **When to use:** VIP accounts, strategic partnerships, founder relationships, or intentional lower-funnel experiments ### Not Scored (shown as "Analyzing" in the UI) Lead has not completed initial enrichment/scoring. * Appears on the **Buyers** page, **Analyzing** tab * Not eligible until scoring completes * Background enrichment and scoring in progress ### State Transitions Leads move between states based on their scores. Higher scores promote to Active, moderate scores land in Monitor, low scores go to Poor Fit, and very low fit leads are Disqualified. Leads can move up or down as new signals appear. Active leads can be demoted on rescore if their signals weaken. Poor Fit leads can be promoted if they change jobs, start posting about relevant pain points, or show new timing signals. You can manually change any lead's state via the UI. Manual Outreach overrides scoring and treats the lead as eligible for enrollment. ### State Impact on Sequences | Lead State | Auto-Enrollment | Manual Enrollment | | --------------- | --------------- | ------------------------------------------------ | | Active | Yes | Immediate enrollment, outreach begins | | Manual Outreach | No | Immediate enrollment, treated as Active | | Monitor | No | Allowed, but not recommended | | Poor Fit | No | Allowed (manual override), low priority | | Not Scored | No | Enrollment queued, starts when scoring completes | | Disqualified | No | Must override to Manual Outreach first | Only **Active** leads are eligible for automatic enrollment by Autopilot. *** ## Signals Signals are data points that indicate a lead is interested, actively buying, or urgently needs your solution. They power buyer intelligence scoring and help identify who is ready to buy. ### Explicit Intent Signals * **Asked for Recommendation** - Lead explicitly asked for product/service recommendations * **Switching From Competitor** - Lead mentioned switching away from a competitor (one of the strongest signals) * **Looking for Alternative** - Lead is actively searching for an alternative * **Complained About Current Tool** - Lead complained about their current tool or process * **Tool Mention** - Lead mentioned using or considering a competitor or similar tool ### Company-Level Signals * **Funding** - Company announced a new funding round * **Product Launch** - Company launched a new product or major feature * **Mergers and Acquisitions** - Company acquired or was acquired * **Cost Cutting** - Company announced layoffs, consolidation, or cost reduction * **IPO Filing** - Company filed for IPO * **Geographic Expansion** - Company expanding to a new geographic market ### Engagement Signals * **Competitor Engagement** - Lead engaged with content from competitor companies * **Engagement Pattern** - Lead has high engagement on posts related to your category * **Own Post Engagement** - Lead engaged with YOUR posts (inbound signal). AutoReach detects likes, retweets, new follows (X) and reactions, comments, mentions of you, and new connections (LinkedIn) * **Competitor Customer** - Lead is a confirmed user of a competing product * **Interaction Orbit** - Lead suddenly starts engaging with multiple competitor or adjacent vendor accounts in a short time window, indicating active evaluation. See [Account Signals](/settings-and-configuration/account-signals) for details ### Hiring Signals * **Hiring Activity** - Company is actively hiring * **Hiring Roles** - Specific roles being hired for * **Job Change Detected** - Lead recently changed jobs (strong timing signal) * **Role Tenure** - How long the lead has been in their current role * **New Hire** - Lead started their current role recently ### Custom Intent Signals Beyond built-in signals, your **Offer** can define custom intent signals specific to your product. Specify keywords and phrases that your buyers use, and AutoReach monitors for mentions and factors them into scoring. ### Offer Pain Match AutoReach automatically checks if a lead's content matches your offer's pain points. If a lead posts about challenges you solve, it is detected as a pain match signal and boosts their intent score. ### Industry and Classification * **Industry Classification** - Compared against your offer's target industries. A match increases fit * **Competitor Employee** - Lead works at a competing company. Scored as Poor Fit ### Location Match Lead's location compared against your preferred locations. The check method depends on your offer's **location filter type**: "lead" matches the lead's own location; "company\_hq" matches the company's headquarters location. ### How Signals Feed Scoring Signals feed into the three scoring dimensions: * **Fit:** Industry match, role match, company size, location * **Intent:** Asking for recommendations, switching tools, complaints, pain point matches, competitor engagement, competitor customer status, interaction orbit clusters, custom intent signals * **Timing:** Job changes, new hires, funding, IPO filings, product launches, hiring activity The more signals a lead has, and the stronger those signals are, the higher their scores. ### Signal Recency AutoReach continuously monitors for new signals. It periodically rechecks leads for job changes, new posts, and company activity. Recent signals carry more weight than older ones. ### Viewing Lead Signals In the lead profile, you can see: * All detected signals in the lead's signal summary * Intent strength (Extremely High, High, Moderate, Low) - an overall AI-assessed rating of buying intent * Account-level signals with individual strength ratings and dates * Raw activity (posts, comments, job changes) *** ## Scoring Rules AutoReach applies rules that override normal scoring: * **Industry mismatch:** If a lead is in the wrong industry, their fit score is reduced * **Competitors:** If a lead works at a competitor company, they are scored as Poor Fit with zeroed scores * **Internal builders:** If a lead is actively building an in-house solution that overlaps with what you sell (e.g., a CTO building their own AI platform when you sell AI tools), they're flagged as not a buyer (right company, wrong persona) and land in Poor Fit > **Note:** You can manually override scoring rules via the UI (set "Manual Outreach" state) if you want to reach someone despite these filters. ## Lead Source Context AutoReach automatically considers how a lead was discovered when scoring. A lead found by searching for your keywords carries inherent intent signal, while a CSV import or follower extraction starts with no intent signal. This happens automatically - no configuration needed. ## Rescoring You can rescore leads at any time by selecting them and running a rescore. This performs a full AI-powered analysis that examines the lead's entire profile, signals, and context. ## Score Feedback If you disagree with a lead's score, you can correct it directly. Open a lead's detail view and click: * **Should be a buyer** (thumbs up) - if the lead scored low but you believe they are a good fit * **Should NOT be a buyer** (thumbs down) - if the lead scored high but is not actually relevant This opens the **Offer Refinement** flow. AutoReach analyzes why the score is off and suggests changes to your offer definition - adjustments to your target audience description, pain points, or ICP criteria. You can review and edit the suggestions before applying them. Once applied, the offer is updated and the lead is rescored. This feedback loop improves scoring accuracy over time by aligning your offer definition with your real-world judgment. ## Score Changes Over Time Lead scores are not static. They update when: 1. **New signals detected** - Background activity fetch finds new posts/engagement 2. **Job changes detected** - Career moves update experience and timing 3. **Offer redefined** - You update your ICP, keywords, or target criteria 4. **Manual rescore triggered** - You explicitly request re-analysis 5. **Re-engagement** - Lead responds to your message ## Best Practices > **Tip:** Your conversion rates will be highest with Active leads. Spend 80% of your effort on this list. > **Tip:** Your next batch of Active leads comes from the Emerging (Monitor) state. Check in weekly to see who got promoted. > **Warning:** A lead might be Poor Fit today and Active tomorrow if they change jobs or companies. Keep monitoring. > **Tip:** Use Manual Outreach strategically. It is powerful for high-value accounts, but do not overuse it. > **Tip:** Leads actively asking for recommendations or evaluating tools are significantly more likely to reply than those with no intent signal. > **Tip:** If you notice a certain phrase or pattern in conversations with buyers, add it as a custom signal. AutoReach will find more people with that pattern. ## Next Steps * [**Account Signals**](/settings-and-configuration/account-signals): Interaction Orbit and company-level heat scoring * [**Auto-Enrollment**](/autopilot/continuous-operations): How Active leads get enrolled into sequences automatically * [**Finding Leads**](/finding-leads/overview): Discover leads showing the signals that matter most # Offers & Knowledge Base An **offer** is the foundation of everything in AutoReach. It is not just a description of your product: it is the intelligence engine that powers lead discovery, scoring, personalization, and AI responses. ## What Is an Offer? One offer = one specific product or service you're selling. **Examples:** * "Workflow automation for mid-market SaaS companies" * "Executive coaching for first-time VPs" * "API compliance auditing for financial services firms" You can create and manage multiple offers, each with its own lead pipeline, sequences, and knowledge base. ## Offer Fields When you create an offer, you define: ### Creation Mode When creating a new offer, you can choose between two modes: * **Website mode:** Paste a URL and click **Extract** to auto-populate all offer fields from your site * **Manual mode:** Fill in each field yourself This toggle only appears during creation. Editing an existing offer always shows the standard form. ### Core Definition * **Name:** What you are calling this offer (required) * **Description:** Detailed description of what you are selling and why (required). Use the **Enhance** button to have AI rewrite your description to be clearer and more specific. * **Website:** Your website URL ### Target Audience * **Target audience:** Free-form ICP definition (required). You can include who to target, who to avoid, and any context that helps AutoReach understand your ideal buyer. For new offers, click **Generate from description** to have AI create this from your description- you can then edit it. Use **Regenerate** to get a fresh suggestion. * **Locations:** Target locations, with a **location filter type** toggle that controls whether leads are matched by their own location or their company's HQ location * **Industries:** LinkedIn industries used to filter people searches. Use the **Generate** button to have AI suggest relevant industries. * **Language:** The language for your outreach messages. Choose from a searchable language picker that supports a wide range of languages (defaults to English). ### Market Positioning * **Pain points:** Problems your offer solves (used for intent matching and message personalization). Add manually or use the **Generate** button. * **Known competitors:** Companies or tools that compete with you (helps detect leads engaging with competitors- these leads are scored as Poor Fit with zeroed scores, not Disqualified). Add manually or use the **Generate** button. ### Value * **Deal value:** Expected contract size ($). Used to estimate pipeline value. ### Other Settings * **Active toggle:** Only active offers are used by Autopilot and sequences for lead scoring and outreach. * **Signal Likelihood:** Displayed as a badge on your offer card (High / Medium / Low signal). This is auto-generated by AI based on your description and pain points - it estimates how likely your target buyers are to post about their needs on social media. You cannot set this manually; it updates when you generate or regenerate your target audience or pain points. ## How Offers Power Everything Your offer definition cascades through AutoReach's entire system: ### 1. Lead Discovery AutoReach automatically generates search keywords from your offer's description and pain points, then uses them to find relevant leads on X and LinkedIn. More specific descriptions and pain points produce more targeted discovery. ### 2. Lead Scoring Your offer's **target audience** and **pain points** guide the buyer intelligence system: * People matching your target industry get higher fit scores * People posting about your pain points get higher intent scores * People in your preferred locations get a fit boost * People at competitor companies are marked as Poor Fit with zeroed scores ### 3. DM Personalization When sequences send DMs, AutoReach uses your offer's **description** and **pain points** to personalize messages: ``` "Hi [name], I noticed you're in [industry] and posted about [pain_point]. We help teams like yours solve [pain_point] with [benefit]. Worth a quick call?" ``` ### 4. AI Responses When leads reply, AutoReach's conversation AI uses your offer to: * Understand context of the discussion * Generate on-brand replies * Know when to escalate to a human * Suggest next steps > **Tip:** Spend time defining your offer well. A great offer definition is a force multiplier that makes every downstream feature smarter. ## The Knowledge Base System The **Knowledge Base** lets you upload your strategic docs, case studies, pricing, qualification criteria, and more. AutoReach uses these documents to: * Make AI responses more accurate and on-brand * Understand your business deeper than just the offer description * Qualify leads better with custom criteria * Generate better personalization in DMs ### Supported File Types * PDF * DOCX (Word) * TXT * Markdown (MD) Limit: 10 documents per offer, 10MB per file. ### How It Works When you upload a document, AutoReach processes it through a full RAG (Retrieval-Augmented Generation) pipeline: 1. The file is stored in cloud storage 2. A background worker parses the document, splits it into chunks, and generates vector embeddings for each chunk 3. Embeddings are stored for similarity search When a prospect replies to your outreach, AutoReach performs a vector similarity search against your knowledge base chunks to find the most relevant content, then feeds it to the AI alongside the conversation to generate accurate, on-brand responses. > **Warning:** Upload your best docs: case studies, pricing, qualification criteria, objection handling, and key talking points. A few high-quality documents will perform better than a large volume of generic content. ## Tone Examples **Tone examples** are conversation samples that define your outreach voice. They are configured **per sequence**, not at the offer level. Tone examples show AutoReach how you talk to prospects. They cover different conversation stages like openings, value propositions, objection handling, social proof, and closing. You add tone examples manually to your sequence, and can edit or remove them at any time. Once added, AutoReach auto-generates a tone summary in the background that the AI uses alongside the examples. During conversations, relevant tone example content is retrieved via vector similarity search to keep AI responses on-voice. > **Tip:** Review and edit your tone examples in your sequence's Advanced Settings. If your samples sound corporate but you are naturally casual, your AI replies will not match your real voice. ## Conversation Analyzer The **Conversation Analyzer** is an AI system that reviews your outreach conversations and suggests improvements. ### What It Does 1. **Samples conversations** from your sequence- comparing positive outcomes (meeting booked) against negative ones (lost, graceful exit) 2. **Identifies patterns** in what worked vs. what didn't 3. **Suggests specific actions**: * **Add tone example**- new conversation samples to guide AI voice * **Modify tone example**- adjust existing examples based on what worked * **Remove tone example**- drop examples that are hurting performance * **Modify prompt**- changes to one of your sequence's AI prompts (how AI replies to DM responses, how AI replies to post comments, or how AI writes the initial cold DM) The analyzer focuses on tone examples and prompts, not offer-level fields like pain points or personalization settings. ### How to Use It Access the Conversation Analyzer from your sequence page. It reviews your conversations and returns actionable suggestions. You can apply or dismiss each suggestion- applying writes the changes directly to your sequence's tone examples or prompts. > **Tip:** The Conversation Analyzer works best with sufficient data. Run it monthly to see patterns emerge. ## Managing Multiple Offers You can create multiple offers in AutoReach. Each offer: * Has its own lead pipeline * Can run independent sequences * Has a separate knowledge base * Has isolated scoring logic **Common setup:** * Offer 1: "Core product for mid-market" * Offer 2: "Enterprise premium service" * Offer 3: "International expansion service" Each has different ICPs, signals, and messaging. > **Warning:** If a lead fits both Offer 1 and Offer 2, they'll appear in both pipelines. This is intentional, as different offers might have different sales approaches to the same person. ## Best Practices > **Tip:** A narrow, well-defined offer (yoga for busy parents aged 30-45) scores better than broad offers (fitness for everyone). Specific signals lead to specific discovery, which produces higher-quality leads. > **Tip:** As you close deals, update your offer's "average deal value". As you learn what types of companies buy, refine your target audience. > **Tip:** You can run AutoReach without a knowledge base, but uploading your best docs will make AI responses more sophisticated and on-brand. > **Warning:** Uploading 200-page documents doesn't help. Upload 5-10 high-quality docs instead. Longer isn't smarter; relevance is. ## Next Steps * [**Creating Your First Offer**](/getting-started/create-offer): Walk through the offer creation process step by step * [**Tone & Knowledge**](/ai-and-conversations/tone-and-knowledge): Learn more about uploading and managing knowledge base documents # Overview AutoReach provides multiple lead discovery methods across X, LinkedIn, and Instagram. Access them from the **Leads** page, which has four tabs: * **All**- Every lead discovered across all your sources * **Signals**- Leads found from intent signal searches, scored and ranked in real time * **Sources**- Extractions from lookalike accounts, link extractions, and CSV imports * **Lookalikes**- Manage your saved lookalike accounts To add new leads, click **Add Lead** on the Leads page. You'll choose a platform (X, LinkedIn, or Instagram), select an account and offer, then pick a discovery method. ## X Discovery Methods When you select X, four options appear: ### From a Lookalike Extract followers or following from a specific X account. You provide a target username, choose followers or following, and set a max count. AutoReach paginates through the list and scores each person against your offer. Enable **Buyer Expansion** to automatically re-extract from seed accounts daily to capture new followers. **Best for:** Mining the audience of industry accounts, newsletters, or competitors on X. ### From Signals (Intent Stream) Search X for people discussing problems your offer solves. AutoReach generates targeted search queries from your offer, or you can provide your own keywords. Both tweet authors and commenters are captured as leads. Advanced settings let you configure max tweets, days back, whether to include replies, and max replies per tweet. Enable **Buyer Expansion** to automatically re-run this search every 24 hours with fresh keywords. **Best for:** Finding leads showing active buying signals on X. [Learn more about X Tweet Search →](/finding-leads/tweet-search) ### From a Link Extract leads from a specific X post or search URL. Two modes: * **Post Comments**- Extract people who commented on a specific tweet * **Search Profiles**- Extract profiles from a search results URL **Best for:** Capturing engaged audiences from specific high-signal posts or threads. ### Add Manually Add a lead by their X username. The lead is queued for enrichment and scoring. ## LinkedIn Discovery Methods When you select LinkedIn, six options appear: ### From a Lookalike Find people similar to a lookalike account on LinkedIn by mining the commenters on that account's posts. (LinkedIn lookalike extraction works off post engagement, not the follower list.) Each run processes a capped number of the seed account's posts and captures the people who commented on them. **Best for:** Scaling outreach by tapping into relevant professional communities. ### By Role Search LinkedIn for people by job title in your offer's target locations. Uses your offer's preferred locations and industries to filter results. Select a role from AI-generated role suggestions, and set a max result count. Enable **Buyer Expansion** to automatically rotate through different roles daily and continuously find new leads. **Best for:** Systematic targeting by role, location, and industry. To target specific companies, use **By Company** instead. [Learn more about LinkedIn People Search →](/finding-leads/linkedin-people-search) ### By Company Search for companies matching descriptors like "B2B SaaS Sales Tooling" or "Lead Generation Agency", then add the top decision-maker from each as a lead. AI can generate descriptors from your offer, or you can enter your own. Enable **Buyer Expansion** to re-run each descriptor daily and capture new companies as they appear. **Best for:** When you know the *type* of company you want to reach but not the people yet. Returns one focused lead per company instead of many. [Learn more about LinkedIn Company Search →](/finding-leads/linkedin-company-search) ### From Signals (Intent Stream) Search LinkedIn posts and comments for intent signals. AutoReach shows which buying signals are active for your offer and generates queries accordingly. Both post authors and commenters are captured by default. Advanced settings include: * **Posts per search query**- How many posts to analyze per query * **Search period**- Past 24 hours, past week, past month, or any time * **Max comments per post**- How many commenters to extract per post * **Include commenters**- Toggle commenter extraction on/off * **Feed search**- Also search your LinkedIn feed for signals * **Hiring signals**- Configure Max Jobs per Role and Max Companies to find decision-makers at companies that are actively hiring Enable **Buyer Expansion** to automatically re-run this search every 24 hours. **Best for:** Finding engaged LinkedIn users with professional signal strength. [Learn more about LinkedIn Content Search →](/finding-leads/linkedin-content-search) ### From a Link Extract leads from a specific LinkedIn post or search URL. Two modes: * **Post Comments**- Extract people who commented on a specific post * **Search Profiles**- Extract profiles from a search results URL **Best for:** Capturing engaged audiences from specific high-signal LinkedIn posts. ### Add Manually Add a lead by their LinkedIn profile URL or vanity name. The lead is queued for enrichment and scoring. ## Instagram Discovery Methods When you select Instagram, these options appear: ### From a Hashtag Extract the authors of top and recent posts under one or more hashtags. Enter a single hashtag, or let AutoReach generate a niche-relevant hashtag pool from your offer. Enable **Buyer Expansion** to rotate through the pool daily and capture fresh authors. **Best for:** Finding people actively posting about topics tied to your offer. ### From a Lookalike Extract the followers or following list of a target Instagram account. Provide the target username, choose followers or following, and set a max count. Enable **Buyer Expansion** to re-extract daily. **Best for:** Mining the audience of an influencer, competitor, or community account. ### From a Link Extract the people who engaged with a specific Instagram post: **likers**, **commenters**, or both. **Best for:** Capturing the engaged audience of a high-signal post or reel. ### Add Manually Add a lead by Instagram username. The lead is queued for enrichment and scoring. [Learn more about Instagram Lead Discovery →](/finding-leads/instagram-search) ## Lookalike Account Discovery From the **Lookalikes** tab, click **New Lookalike** to discover relevant accounts. Two search methods: * **By Offer**- AutoReach uses your offer's target audience to find influencers, thought leaders, and communities whose audiences match your ICP * **By Profile URL**- Provide a specific influencer or competitor profile URL The search returns up to 10 candidate accounts with follower counts and audience categories. Select which to save, then use them as seed accounts for extraction via the "From a Lookalike" option. Works on X, LinkedIn, and Instagram. [Learn more about Lookalike Audiences →](/finding-leads/lookalike-audiences) ## Lead Pool (Instant Matching) AutoReach maintains a shared pool of pre-enriched lead profiles with vector embeddings. When you create an offer, the system automatically matches existing pool entries against your ICP. Matching leads appear in your pipeline instantly- no waiting for enrichment. The Lead Pool is fully automatic. You can see pool-sourced leads by filtering the All tab by "Pool DB" source. **Best for:** Getting scored leads instantly without waiting for enrichment. [Learn more about Lead Pool →](/finding-leads/lead-pool) ## CSV Import Click **Import Leads** on the Leads page to bulk-import leads from a CSV file. Upload your file, preview the results (including duplicate detection), select accounts for enrichment, and confirm. AutoReach supports X handles, LinkedIn URLs, Instagram profiles, and email addresses. Imported leads are automatically queued for enrichment and scoring. ## CSV Export Select leads from the All tab and export them as a CSV file. The export includes profile information, scores, contact details, source data, and enrichment results. Use filters to narrow down which leads to include before exporting. ## Cross-Platform Profile Matching Cross-platform matching (finding X profiles for LinkedIn leads and vice versa) is a **manual action**, not an automatic pipeline step. Select leads, click Enrich, and enable "Find LinkedIn" or "Find X Profile" under Profile Discovery. See [Lead Pool](/finding-leads/lead-pool) for details. ## Choosing the Right Method | Method | Platform | Best For | | ------------------- | ----------------------------- | ---------------------------------------------- | | From a Lookalike | X, LinkedIn, Instagram | Scaling via relevant community audiences | | By Role | LinkedIn | Systematic targeting by role/location/industry | | By Company | LinkedIn | One decision-maker per matching company | | From Signals | X, LinkedIn | Intent signals from active discussions | | From a Hashtag | Instagram | People posting about your offer's topics | | From a Link | X, LinkedIn, Instagram | Capturing engagement on specific posts | | Add Manually | X, LinkedIn, Instagram | Adding individual prospects | | Lookalike Discovery | X, LinkedIn, Instagram | Finding relevant accounts to extract from | | Lead Pool | X, LinkedIn, Instagram | Instant matches from pre-enriched profiles | | CSV Import | X, LinkedIn, Instagram, Email | Bringing in external prospect lists | ## Quick Start 1. **Starting fresh?** Click **Add Lead**, choose a platform, and select **From Signals** to find intent-driven prospects. 2. **Have specific role/location targets?** Use **By Role** (LinkedIn) for precise filtering. 3. **Want to scale quickly?** Discover **Lookalike** accounts, then extract their followers. 4. **Need fast results?** Create an offer and let the **Lead Pool** instantly match pre-enriched leads. 5. **Want continuous growth?** Enable **Buyer Expansion** on your searches and extractions for daily autopilot discovery. > **Note:** All lead discovery methods automatically score new prospects against your offer for fit, intent, and timing before they appear in your pipeline. > **Warning:** Lead quality depends on accurate ICP definition. Spend time setting up your offer with clear pain points and target audience for best results. # X/Twitter Tweet Search Find high-intent prospects by searching X for tweets that match your offer. AutoReach uses AI to generate targeted search queries organized by intent category, then extracts both tweet authors and commenters as leads. ## How It Works ### 1. Query Generation When you start a search, you can either: * **Let AI generate queries:** Provide your offer and AutoReach generates targeted keywords plus structured **intent clusters**- groups of queries organized by intent type. The AI selects the most relevant intent categories for your offer. Competitor-specific keywords are generated and merged separately. * **Provide your own keywords:** Enter your own keywords and search query directly to skip AI generation entirely. #### Intent Categories The AI organizes queries across these intent types (selecting the most relevant for your offer): * **Operational Pain** - day-to-day frustrations and inefficiencies * **Budget Pressure** - cost concerns and budget constraints * **Hiring & Scaling** - growth challenges and team scaling * **Buying Evaluation** - actively comparing or evaluating solutions * **Competitor Switch** - switching away from or frustrated with competitors * **Growth Signals** - expansion, new markets, scaling needs * **Security & Compliance** - regulatory, security, or compliance concerns * **Vertical Specific** - industry-specific pain points * **Niche Jargon** - insider terminology, tools, certifications, and community names specific to your industry ### 2. Search Execution AutoReach searches X for tweets matching your queries. Results are automatically deduplicated across all your searches so the same person never creates duplicate leads. #### Search Parameters | Parameter | Default | Description | | ------------------ | -------------------------- | --------------------------------------------------------------- | | Max tweets | 250 | Total tweets to collect (range 10-300) | | Days back | 60 | How far back to search | | Include replies | enabled | Extract commenters from matching tweets | | Max replies | 100 | Max replies to check per tweet | | Exclusions | giveaway, retweet, airdrop | Terms to exclude from results | | Daily recurring | disabled | Enable automatic daily re-runs | | Enrichment options | - | Whether to run enrichment and deep analysis on discovered leads | ### 3. Lead Extraction AutoReach pulls leads from two sources within each matching tweet: * **Tweet authors:** People posting directly about relevant topics, challenges, or needs * **Commenters:** Professionals engaging in the conversation (when reply extraction is enabled) All extracted prospects are converted to leads and queued for enrichment and scoring against your offer. ### 4. Recurring Daily Searches (Buyer Expansion) Enable **Buyer Expansion** to run your search automatically every 24 hours. Each recurring run **regenerates fresh keywords** based on your offer, using the previous keywords as a reference to force variation and capture new prospects. ### 5. Cost Estimation Before running a search, use the cost estimation feature to preview estimated AI and API costs based on your selected offer and settings. ## Best Practices for Keyword Selection The quality of your search results depends heavily on the keywords and phrases in your queries: * **Be specific.** Multi-word phrases like "API rate limiting frustration" or "AWS cost optimization" produce far better results than single generic words like "API" or "cloud." * **Use your audience's language.** Include industry jargon, product names, and terminology your target buyers actually use when discussing their problems. * **Start broad, then refine.** Run an initial search with a range of terms, review the results, and narrow your queries based on what performs well. * **Experiment with variations.** Different keyword combinations surface different prospect segments. Test multiple approaches to find the best fit for your offer. * **Customize exclusion terms.** The default exclusions are minimal (`giveaway`, `retweet`, `airdrop`). Add terms relevant to your space to filter out noise, or clear the list entirely if defaults are too aggressive. ## Example Workflow **Offer:** B2B SaaS sales automation platform **AI-generated queries might include:** Keywords: * "manual sales process too slow" * "need sales pipeline visibility" * "CRM data quality issues" Intent clusters: * **Operational Pain:** "my team spends hours on data entry", "sales admin is killing productivity" * **Buying Evaluation:** "evaluating sales tools", "comparing CRM platforms" * **Competitor Switch:** "moving away from Salesforce", "HubSpot isn't scaling" **Result:** Each query surfaces relevant tweets from CTOs, VPs of Sales, and Operations Managers. Both tweet authors and commenters are extracted, converted to leads, and queued for enrichment and scoring. ## Troubleshooting **Getting too many irrelevant results?** * Review your offer description for clarity and specificity * Add more exclusion terms to filter out common false positives * Use more niche, multi-word search phrases **Not finding enough prospects?** * Broaden your search terms or include common synonyms * Increase the max tweets setting to collect more results * Increase the days back setting to search further into the past * Check whether your target audience actively discusses these topics on X **Seeing duplicate prospects across searches?** * This is expected. AutoReach deduplicates prospects at both the per-search and global level, so the same person appearing in multiple searches will not create duplicate leads. ## Next Steps * [**Enrichment Pipeline**](/enrichment/pipeline): Learn how discovered leads get enriched with profile and company data * [**Building Sequences**](/outreach-and-sequences/building-sequences): Set up outreach sequences for your newly discovered leads # LinkedIn Content Search Discover high-intent decision-makers through LinkedIn posts and comments. AutoReach generates intent-organized search queries, searches LinkedIn's content feed, and extracts both post authors and commenters as leads. ## When to Use This vs. People or Company Search Content Search is the highest-intent way to find leads on LinkedIn. Instead of asking "who has this title?" it asks "who is *right now* posting or commenting about the problems my offer solves?" Authors are warm. Commenters are sometimes even warmer: they took the time to engage publicly. Use Content Search when: * You want your first few replies fast: these leads have inherent intent signal and tend to score and convert higher than role-only matches. * Your offer solves a problem people complain about publicly (most B2B pain points qualify). * You want to mine engagement on a specific influencer's posts (use **From a Link** in [Finding Leads Overview](/finding-leads/overview#from-a-link) for that; Content Search casts a wider net). Use [People Search](/finding-leads/linkedin-people-search) instead when you need volume by role, or [Company Search](/finding-leads/linkedin-company-search) when you're targeting a specific company type. Most pipelines benefit from running Content Search alongside one of the others rather than choosing one: Content for intent, the others for steady volume. ## How It Works ### 1. Intent-Based Query Generation AutoReach's AI generates search queries based on your offer, organized by intent category: | Intent Category | What It Finds | | -------------------- | --------------------------------------------------------------------------------- | | **Pain Points** | Professionals discussing operational challenges, inefficiencies, and frustrations | | **Hiring Signals** | Recruiting efforts, team expansion, skill gaps | | **Solution Seeking** | People evaluating tools, comparing options, asking for recommendations | | **Buying Intent** | Budget allocation, purchase decisions, vendor selection | | **Growth Signals** | Revenue growth, market expansion, new initiatives | All intent categories are searched automatically to maximize coverage. Queries are always generated server-side. Unlike X Tweet Search, you cannot provide your own queries directly. ### 2. Content Discovery AutoReach searches LinkedIn's content feed to find posts matching your queries. Posts are discovered and their full details fetched- including post text, author profile, engagement metrics, and creation date. ### 3. Commenter Extraction When commenter extraction is enabled (the default), AutoReach extracts commenters from each discovered post, capturing their name, headline, profile URL, comment text, and the parent post context. > **Note:** LinkedIn commenters are often decision-makers actively engaging with relevant content, which is a strong buying signal. ### 4. ICP Matching Extracted prospects are matched against your offer's target audience. All discovered prospects are added as leads- scoring happens during the enrichment pipeline. ### 5. Recurring Daily Searches (Buyer Expansion) Enable **Buyer Expansion** to run the search automatically every 24 hours. Each recurring run **regenerates fresh queries** to avoid repeating previous searches and capture new content. ## Search Parameters | Parameter | Default | Description | | ---------------------- | ---------- | ----------------------------------------------------------------------- | | Posts per search query | 25 | Max posts to collect per query | | Search period | Past month | How far back to search (Past 24 hours, Past week, Past month, Any time) | | Include commenters | enabled | Extract commenters from discovered posts | | Max comments per post | 50 | Max comments to fetch per post | | Feed search | disabled | Also search your LinkedIn feed for signals | | Buyer Expansion | disabled | Enable automatic daily re-runs | | Enrichment options | - | Whether to run enrichment and deep analysis on discovered leads | ### Hiring Signals Additional settings for hiring signal detection: | Parameter | Default | Description | | ----------------- | ------- | ------------------------------------------------------------ | | Max Jobs per Role | 100 | Max job postings to collect per role target (slider: 25–200) | | Max Companies | 30 | Max companies to process (slider: 10–50) | See [LinkedIn Job Search](/finding-leads/linkedin-job-search) for details on how hiring signals work. ## Cost Estimation Before running a search, use the cost estimation feature to preview estimated costs based on your settings. ## Progress Tracking While a content search runs, progress is tracked and updated in real time. You can monitor: * **Posts found:** Total LinkedIn posts matching your queries * **Comments fetched:** Total comments retrieved * **People found:** Unique prospects identified * **People after ICP match:** Prospects matching your ICP * **Current operation:** What phase the search is in * **Current query:** Which query is currently being searched ## Best Practices 1. **Be specific.** More specific search terms yield higher-quality prospects than broad keywords. 2. **Include pain points.** Search for the exact pain points your offer solves (from your offer description). 3. **Keep commenters on.** LinkedIn commenters are often highly engaged decision-makers. Leave commenter extraction enabled. 4. **Run recurring searches.** Set up daily LinkedIn content searches to capture new prospects continuously with fresh query rotation. ## Example Workflow **Offer:** "People analytics and HR software platform" **Intent categories searched:** Pain Points, Hiring Signals, Solution Seeking, Buying Intent, Growth Signals **AI-generated queries might include:** * "struggling with employee retention" * "need visibility into team performance" * "talent management challenges" * "hiring pipeline needs improvement" **Result:** Content search finds posts discussing HR challenges, hiring, and performance management. Post authors and commenters are extracted- CHROs, VPs of People, and Talent Acquisition leaders discussing your exact problem space. All are added as leads and queued for enrichment and scoring. ## Troubleshooting **Seeing too many generic results?** * Review your offer description and pain points for clarity **Not finding enough prospects?** * Increase posts per search query to collect more results * Add more specific pain points to your offer * Check if your target audience is active on LinkedIn **Getting low ICP match rates?** * Verify your offer's target audience and pain points are well-defined **High duplicate rate across searches?** * This is normal. The same prospect may comment on multiple relevant posts. Deduplication happens before leads are added to your database. ## Next Steps * [**LinkedIn People Search**](/finding-leads/linkedin-people-search): Target prospects by job title, company, and location * [**Enrichment Pipeline**](/enrichment/pipeline): See how discovered leads get enriched with full profile data # LinkedIn People Search Find decision-makers using LinkedIn's search. Systematically target specific roles, locations, and industries to build highly filtered prospect lists. (To target by company, use [LinkedIn Company Search](/finding-leads/linkedin-company-search) instead.) ## When to Use This vs. Other LinkedIn Searches LinkedIn has three search methods and they answer different questions: * **People Search (this page):** Use when you know the *role* you want to reach ("VP of Sales", "Head of Demand Gen") and want a steady stream of people matching that role across companies in your target industries and locations. Best for systematic, high-volume role-based prospecting. * [**Company Search**](/finding-leads/linkedin-company-search)**:** Use when you know the *type of company* you want to reach ("B2B SaaS sales tooling", "lead gen agencies") but don't know who specifically. Returns one decision-maker per matching company. Lower volume, more focused. * [**Content Search**](/finding-leads/linkedin-content-search)**:** Use when you want *high-intent leads* who are actively discussing problems your offer solves. Highest signal strength, lower volume than role-based search. A common pattern: start with Content Search to get a few warm replies fast, then layer People Search on top for steady volume once you've validated the offer is landing. ## How It Works LinkedIn People Search lets you directly query LinkedIn's search with precise targeting criteria. Rather than inferring prospects from content, you explicitly define who you're looking for. There are two ways to start a search: 1. **By role + filters:** Provide a job title/role keyword along with location, industry, and other filters. This is the most common approach for systematic prospecting. 2. **By lookalike seed:** Select a seed from a lookalike account discovery. LinkedIn lookalike extraction mines the commenters on that account's posts (a capped number of posts per run), since LinkedIn no longer exposes follower lists for this purpose. ### Available Filters | Filter | Description | | -------------- | ---------------------------------------------------------------------------------------------------------------------------- | | **Title/Role** | Job title keyword (e.g., "CTO", "VP Sales"). Not used in lookalike-seeded searches. | | **Location** | Resolved from your offer's preferred locations. Supports countries, regions, cities, and abstract concepts like "Worldwide." | | **Industry** | Resolved from your offer or overridden directly. | Note: **Company size** and **network distance** are not search filters - they are part of your offer's ICP definition and checked during scoring, not at search time. Lookalike-seeded searches do not use a title or follower filter. They mine the commenters on the seed account's posts instead (see [Lookalike Audiences](/finding-leads/lookalike-audiences)). ### Search Parameters | Parameter | Default | Description | | ---------------------- | ------------ | --------------------------------------------------------------- | | **LinkedIn Account** | *(required)* | LinkedIn account to search from | | **Role** | - | Job title keyword (required for direct role searches) | | **Offer** | - | Resolves location + industry filters from the offer | | **Industries** | from offer | Override offer's industry IDs | | **Max Results** | 250 | Max results to collect (slider: 50–5,000) | | **Enrichment Options** | - | Whether to run enrichment and deep analysis on discovered leads | | **Buyer Expansion** | disabled | Enable daily autopilot with role rotation | ## Progressive Background Search Searches run in the background- results are saved progressively as they arrive, rather than waiting for the entire search to complete. The UI shows progress as results come in. If a search is interrupted, it automatically resumes from where it left off. Non-retryable errors (expired session, automation detected) stop the search. ## Buyer Expansion (Daily Autopilot) When buyer expansion is enabled, AutoReach automatically re-runs the search daily with role rotation to continuously discover new leads. **How it works:** 1. Each day, the system rotates to the next role in the expansion queue 2. Runs a new search for that role 3. If a role returns no results, it rotates to the next 4. When all roles are exhausted, buyer expansion is automatically disabled You can select a role from AI-generated suggestions based on your offer, or provide custom roles. ## Account Safety AutoReach automatically manages LinkedIn account safety with rate limiting, request throttling, and concurrency controls. You do not need to configure any of this- it is fully automatic. ## Result Data Each search result includes: * Name and headline * LinkedIn profile URL * Location * Profile image URL Results are queued for enrichment based on your enrichment settings. ## Building Effective Filter Combinations ### Example 1: CTOs in a Specific Region * **Role:** CTO, VP Engineering * **Location:** San Francisco Bay Area (resolved from offer) * **Industry:** Software, Technology * **Result:** Technical decision-makers in your target metro ### Example 2: Financial Services Decision-Makers * **Role:** VP Finance, CFO, Treasurer * **Location:** New York, London, Singapore * **Industry:** Financial Services, Banking * **Result:** Finance leaders across major financial hubs ### Example 3: Continuous Prospecting with Buyer Expansion * **Role:** VP Sales (initial) * **Buyer expansion:** Enabled with default roles * **Result:** Daily automated discovery cycling through CEO, CTO, VP, Director, Manager roles at matching companies ## Best Practices 1. **Use multiple searches for different personas.** Don't try to find everyone in one search. Create separate searches for different roles. 2. **Combine filters strategically.** Location + Title + Industry yields higher-quality results than any single filter. 3. **Enable buyer expansion for ongoing discovery.** Set up a search with buyer expansion enabled to continuously find new decision-makers without manual effort. 4. **Start with a focused search, then scale.** Run a search with tight filters first, review results, then broaden if needed. 5. **Monitor quality downstream.** As leads are scored and enrolled in sequences, track response rates. Adjust filters if quality is low. ## Troubleshooting **Getting too many results?** * Add more filter criteria (industry, location) * Use more specific title keywords * Narrow geographic scope **Not getting enough results?** * Broaden your filters (wider geography, related titles) * Try alternative job title variations * Remove industry filter to cast a wider net **Seeing irrelevant profiles?** * Be more specific with title keywords * Add industry filter if not set * Results are scored by buyer intelligence after enrichment- irrelevant profiles will score low automatically **Search seems slow?** * This is normal for large result sets. Searches are paced conservatively to protect your account. * Searches run in the background - you can continue using the app. * Large searches (1,500+ results) may take a while to complete. **Search stuck or failed?** * Auto-recovery retries stuck searches automatically on server restart * If the LinkedIn session expired, reconnect your account and the search will resume * Temporary pauses due to rate limiting clear automatically ## Next Steps * [**Lookalike Audiences**](/finding-leads/lookalike-audiences): Discover relevant seed accounts, then extract their audience (post commenters on LinkedIn, followers on X and Instagram) * [**LinkedIn Content Search**](/finding-leads/linkedin-content-search): Find leads via content engagement instead of profile filters * [**Enrichment Pipeline**](/enrichment/pipeline): See how discovered leads get enriched with full profile and company data # LinkedIn Company Search (By Company) Find decision-makers by searching for companies that match a description, then extracting one top-ranked decision-maker from each. Use **By Company** when you know the *type* of company you want to reach, but not the specific people yet. ## How It Works LinkedIn Company Search is a two-step pipeline: 1. **Find companies** matching your descriptors (e.g., "Lead Generation Agency", "B2B SaaS Sales Tooling", "Boutique Recruiting Firm") 2. **Find the top decision-maker at each company** based on seniority and offer-relevance The result is a focused list of one prospect per company, not dozens of irrelevant employees. ## Where to Start a Search 1. Go to **Leads** and click **Add Lead** 2. Choose **LinkedIn** as the platform 3. Select your account and offer 4. Click **By Company** ## Configuring the Search ### Company Descriptors Descriptors are short phrases that describe the *type* of company you want to reach. Examples: * "Lead Generation Agency" * "B2B SaaS Sales Tooling" * "Outbound Sales Consultancy" * "Mid-Market Recruiting Firm" Two ways to add descriptors: * **Generate from offer**: Click **Regenerate from offer** to have AI suggest descriptors based on your offer's description, target audience, and pain points. You can refine the suggestions before running the search. * **Add manually**: Type a descriptor and press Enter (or click the + button) to add it. > **Tip:** Mix broad descriptors ("B2B SaaS") with niche ones ("PLG analytics startup") to balance volume and precision. ### Max Companies Use the slider to set how many companies to process. Each company contributes one decision-maker, so this is also the maximum number of leads you'll add from the search. | Setting | Range | Default | | ------------- | ------ | ------- | | Max Companies | 10-100 | 50 | ### Buyer Expansion (Daily Autopilot) Toggle **Buyer Expansion** to re-run each descriptor daily. New companies that appear over time (newly-listed, recently-funded, expanded into the territory) are automatically discovered and turned into leads. When enabled, AutoReach: 1. Re-runs each descriptor every 24 hours 2. Adds the top decision-maker from each newly-discovered company 3. Marks descriptors as exhausted when they stop returning fresh companies 4. Auto-disables Buyer Expansion once all descriptors are exhausted ### Cost Estimation Before running, the cost estimator shows expected AI and search costs for your settings. ## What Gets Returned For each matched company, the top-ranked decision-maker is added as a lead with: * Name, headline, profile URL, location * Company name and basic company data * The descriptor that matched the company (visible on the lead's source data) Leads are queued for enrichment and scoring like any other source. ## How "Top Decision-Maker" Is Picked For each company, AutoReach pulls the employee list and ranks people by seniority. Founders, co-founders, and C-suite rank highest, followed by VPs, Directors, Heads of, and Managers. The single highest-ranked match is added as the lead. This is fully automatic. You do not configure roles directly for By Company search. To target specific roles at the same companies later, layer a [LinkedIn People Search](/finding-leads/linkedin-people-search) on top. ## Best Practices 1. **Use specific descriptors.** "B2B SaaS sales analytics platform vendor" produces sharper matches than "software company." 2. **Mix tiers.** A handful of broad descriptors plus a handful of niche ones gives you both reach and precision. 3. **Write a detailed offer.** AI descriptor generation pulls from your offer's description, target audience, and pain points. Vague offers produce vague descriptors. 4. **Enable Buyer Expansion for ongoing discovery.** New companies pop up constantly. Daily re-runs catch them without manual effort. 5. **Combine with By Role.** By Company finds one person per company. By Role finds many people across many companies. Use them together to cover both axes. ## Example Workflow **Offer:** "Outbound infrastructure for B2B SaaS RevOps teams" **AI-generated descriptors:** * "B2B SaaS RevOps consultancy" * "Outbound sales agency" * "Mid-market sales tooling vendor" * "PLG analytics startup" **Result:** AutoReach finds 30 companies across these descriptors, picks the top RevOps or Sales leader at each, and adds them as leads. Buyer Expansion re-runs daily, surfacing newly-active firms. ## Hiring Signals vs By Company By Company and **Hiring Signals** (inside [LinkedIn Content Search](/finding-leads/linkedin-content-search)) both produce one decision-maker per company, but they use different selection methods: | Feature | Selection Method | When to Use | | ------------------ | ---------------------------------------------------------------------- | ------------------------------------------------------ | | **By Company** | Match company by descriptor, pick best-fit decision-maker | You know the type of company you want | | **Hiring Signals** | Find companies hiring for relevant roles, pick best-fit decision-maker | You want timing-driven leads at growth-stage companies | Use them together. By Company gives you a steady base; Hiring Signals adds time-sensitive opportunities. ## Troubleshooting **Few or no companies returned?** * Descriptors may be too narrow. Add broader synonyms. * Your geo filter may be limiting matches. Try widening the offer's preferred locations. * LinkedIn may have throttled the account briefly. Wait and retry. **Decision-makers don't look right?** * AutoReach picks the most senior person at each company. If you need a specific role (e.g., RevOps Manager), run [LinkedIn People Search](/finding-leads/linkedin-people-search) on the discovered companies instead. **Companies look duplicated within a single search?** * Each search execution deduplicates companies as it processes them. The same company will not be added twice within one run. ## Next Steps * [**LinkedIn People Search**](/finding-leads/linkedin-people-search): Find multiple roles per company once you've identified the right targets * [**LinkedIn Job Search (Hiring Signals)**](/finding-leads/linkedin-job-search): Time-sensitive company discovery from job postings * [**Enrichment Pipeline**](/enrichment/pipeline): How discovered leads get enriched and scored # LinkedIn Job Search Find decision-makers at companies that are actively hiring. LinkedIn Job Search identifies growth-stage companies through job postings, resolves company profiles, and extracts the most senior decision-maker at each company. ## How It Works Job search is not a standalone search type - it is integrated into LinkedIn Content Search as the **Hiring Signals** intent category. When a content search runs, hiring signal detection is included automatically. Additional settings let you configure the job search parameters. ### 1. AI-Generated Role Targets AutoReach generates job title keywords tailored to your offer using AI, based on your offer description, name, and target audience. Additional niche titles are generated from industry-specific terminology. **Example for a Sales Automation Platform:** * "Sales operations manager" * "Sales pipeline builder" * "CRM implementation consultant" * "VP of sales operations" **Example for HR Analytics:** * "People analytics lead" * "HR metrics manager" * "Head of people operations" ### 2. Job Listing Discovery For each role target, AutoReach searches LinkedIn Jobs filtered by your offer's preferred locations. Job data includes the job title, company name, location, workplace type (Remote/Hybrid/On-site), and posting date. ### 3. Company Resolution After collecting jobs across all role targets, AutoReach deduplicates by company and fetches full company profiles- including industry, staff count, headquarters location, website, and specialties. The list is capped at your configured **Max Companies** limit. ### 4. Decision-Maker Discovery For each company, AutoReach finds employees and selects the **most senior decision-maker**- prioritizing C-suite, founders, and VPs. One decision-maker per company is extracted to ensure focused outreach. ### 5. Lead Enrichment and Scoring Each discovered decision-maker is added to your lead database and queued for enrichment and scoring based on your enrichment settings. ## Competitor Customer Detection When your offer has competitors configured, LinkedIn Job Search automatically discovers **competitor customers**- companies posting jobs that require experience with competing products. Decision-makers at these companies are extracted and tagged with elevated priority. This runs automatically as part of the content search pipeline when **Hiring Signals** is enabled and competitors are defined. ## Search Parameters These are configured within the LinkedIn content search settings: | Parameter | Default | Description | | --------------------- | ------- | ------------------------------------------------------------ | | **Max Jobs Per Role** | 100 | Max job postings to collect per role target (slider: 25–200) | | **Max Companies** | 30 | Max companies to process (slider: 10–50) | ## Best Practices 1. **Rely on AI-generated titles.** The AI generates domain-specific role targets based on your offer. These yield higher-quality matches than generic titles. 2. **Combine with other intent categories.** Run **Hiring Signals** alongside **Pain Points** or **Solution Seeking** in a single content search to maximize coverage. 3. **Monitor company profiles.** The resolved company data (industry, staff count, funding) helps validate whether a company is in your ICP before outreach. 4. **Enable Buyer Expansion.** Turn on daily recurring to catch new job postings and hiring activity continuously. 5. **Leverage competitor customer detection.** If you have competitors defined in your offer, this feature automatically finds companies already paying for solutions like yours. ## Example Workflow **Offer:** "People analytics and HR intelligence platform" **AI generates role targets:** * "Head of People Operations" * "People Analytics Manager" * "HR Technology Lead" **Result:** Job search finds job postings across the roles, identifies the hiring companies, resolves their profiles (industry, size, HQ), and extracts the most senior decision-maker at each. Competitor customer detection also finds companies hiring for roles requiring experience with competing HR platforms. ## Troubleshooting **Getting too many irrelevant job postings?** * Reduce the Max Jobs per Role slider to limit volume * Reduce the Max Companies slider to focus on the best matches **Decision-makers don't seem relevant?** * AutoReach prioritizes C-suite and VPs. Combine with [LinkedIn People Search](/finding-leads/linkedin-people-search) to find specific roles at companies discovered through job search. **Not finding enough companies?** * Increase the Max Jobs per Role slider * Check if your target industry actively posts on LinkedIn Jobs ## Next Steps * [**LinkedIn People Search**](/finding-leads/linkedin-people-search): Find additional decision-makers at companies discovered through job search * [**LinkedIn Content Search**](/finding-leads/linkedin-content-search): Discover professionals discussing challenges related to hiring activity * [**Enrichment Pipeline**](/enrichment/pipeline): See how discovered leads get enriched with full profile and company data # Instagram Lead Discovery Find prospects on Instagram by mining hashtags, follower and following lists, post engagers, and lookalike accounts. Instagram leads flow into the same pipeline as X and LinkedIn: they are enriched, scored against your offer, and become eligible for sequences and Autopilot. > **Prerequisite:** Connect an Instagram account through the Chrome Extension first. See [Connecting Your Accounts](/getting-started/connecting-accounts). ## Starting an Instagram Search From the **Leads** page, click **Add Lead**, choose **Instagram**, then select the Instagram account and offer to use. Pick a discovery method from the options below. Each search runs in the background and converts discovered profiles into leads automatically. ## Discovery Methods ### From a Hashtag Extract the authors of top and recent posts under one or more hashtags. Enter a single hashtag, or, if your offer is linked, let AutoReach generate a pool of niche-relevant hashtags for you. Enable **Buyer Expansion** to keep the search running daily. AutoReach rotates through the hashtag pool one tag at a time, capturing fresh authors each run so the source never goes stale. **Best for:** Finding people actively posting about topics tied to your offer. ### From a Lookalike (Followers / Following) Extract the **followers** or the **following** list of a target Instagram account. You provide the target username, choose followers or following, and set a maximum number of leads. AutoReach paginates through the list and scores each person against your offer. Enable **Buyer Expansion** to re-extract from the seed account daily and capture new followers over time. **Best for:** Mining the audience of an influencer, competitor, or community account in your niche. ### From a Link (Post Engagers) Extract the people who engaged with a specific Instagram post. Provide the post URL and choose which engagers to capture: * **Likers** - people who liked the post * **Commenters** - people who commented on the post **Best for:** Capturing the engaged audience of a specific high-signal post or reel. ### Add Manually Add a single lead by Instagram username. The lead is queued for enrichment and scoring like any other. ## What Gets Extracted For every discovered profile, AutoReach captures the username, full name, verified and private flags, and (during enrichment) follower count, following count, post count, bio, external link, and business-account status. Results are deduplicated against your existing leads so the same person never creates a duplicate. Private profiles and profiles that cannot be resolved are filtered out before they reach your pipeline. ## Enrichment and Scoring Instagram leads run through the standard enrichment and [Buyer Intelligence](/core-concepts/buyer-intelligence) scoring pipeline. A pre-enrichment fit gate screens out clear non-fits before deeper enrichment runs, which keeps your pipeline focused on relevant prospects. ## Buyer Expansion (Recurring Discovery) Searches marked with **Buyer Expansion** resume automatically each day. For hashtag searches, AutoReach advances through the hashtag pool; for follower and following searches, it continues from where the last run left off. This is the engine behind continuous Instagram discovery in [Autopilot](/autopilot/overview). ## Next Steps * [**Finding Leads Overview**](/finding-leads/overview): All discovery methods across platforms * [**Supported Actions by Platform**](/outreach-and-sequences/supported-actions): What you can do with Instagram leads in a sequence * [**Autopilot**](/autopilot/overview): Run Instagram discovery and outreach hands-free # Lookalike Audience Discovery Find high-quality prospects by identifying influencers and accounts whose audiences match your ICP. AutoReach uses AI-powered web search to discover seed accounts, then extracts and scores their audience as leads (followers on X and Instagram, post commenters on LinkedIn). ## How It Works Lookalike Audience Discovery follows three steps: ### 1. Discover Seed Accounts AutoReach uses AI-powered web search to find accounts whose audiences overlap with your target market. Based on your offer, it searches for relevant accounts, verifies them against live platform APIs, and returns a curated list of seed accounts. Discovery works on X, LinkedIn, and Instagram from the same interface. **What it searches for:** The AI searches across multiple audience categories: * **Industry publications**- newsletters, media outlets, trade publications * **Professional communities**- groups, forums, community accounts * **Vendor/competitors**- competitor accounts and adjacent vendors * **Thought leaders**- executives, analysts, and domain experts * **Vertical-specific**- niche accounts specific to your industry If your offer specifies preferred locations, the AI prioritizes accounts with audiences in your target regions. ### 2. Extract the Audience Once seed accounts are saved, their audience is extracted through the standard extraction pipelines: * **X seeds:** Followers are extracted and queued for enrichment * **LinkedIn seeds:** The commenters on the seed account's posts are mined (a capped number of posts per run) and queued for enrichment * **Instagram seeds:** Followers (or following) are extracted and queued for enrichment In the **manual flow**, you discover seeds from the Lookalike Audiences page, save them to your library, then separately trigger extraction via the "From a Lookalike" option. Under **Autopilot**, extraction is fully automatic after discovery. ### 3. Score and Add Leads Every extracted profile goes through the standard enrichment and scoring pipeline. All extracted profiles are added as leads and scored by buyer intelligence like any other source. ## Search Options When starting a lookalike search, you configure: * **Offer**- select the offer that provides ICP context, location, and competitors * **Channel**- choose X, LinkedIn, or Instagram * **Platform account**- select your connected X, LinkedIn, or Instagram account * **Profile URL** (optional)- enter a profile URL for profile-based discovery instead of offer-based **Two discovery modes:** * **By Offer:** Select an offer- finds seeds matching your offer's ICP, domain, and audience * **By Profile URL:** Enter a profile URL- finds accounts similar to a specific person ## Available Actions From the Lookalike Audiences page, you can: * **Estimate cost**- preview the estimated cost before running a search * **Search for seeds**- discover seed accounts with real-time progress updates * **Save accounts**- save one or more discovered accounts to your library * **Browse saved accounts**- view your saved accounts with pagination and filtering * **Delete accounts**- remove individual or multiple saved accounts ## Automatic Rotation Under Autopilot When Autopilot is enabled, seed accounts rotate automatically: 1. When a seed is exhausted, Autopilot discovers **new** seed accounts via web search 2. New seeds are saved and extraction begins automatically 3. Previously used accounts are tracked to avoid duplicates 4. If the AI can't find new relevant accounts, it broadens its search to find lesser-known accounts 5. If discovery stalls, you receive a notification This means Autopilot actively discovers fresh seed accounts- it does not just re-extract from existing ones. ## Cost Estimation Before running a search, use the cost estimation feature to preview estimated costs for the web search AI calls. ## Best Practices 1. **Mix account types.** Combining publications, communities, and individual thought leaders produces more diverse and well-rounded audiences. 2. **Prioritize engaged audiences.** Accounts with high engagement rates tend to have more active, reachable followers. 3. **Use multiple platforms.** Run lookalike discovery across X, LinkedIn, and Instagram to maximize coverage. The extraction pipelines differ but all produce scored leads. 4. **Monitor lead quality.** Track response rates from lookalike-sourced leads. If quality drops, your seed accounts may need adjustment. 5. **Combine with other methods.** Lookalike audiences complement content search and people search. Use them together for best coverage. 6. **Let Autopilot handle rotation.** Manual seed management works, but Autopilot's automatic rotation with escalating diversity prompts keeps the pipeline flowing without intervention. ## Example Workflow **Offer:** "API platform for payment processing" **Step 1:** AutoReach web-searches for fintech publications, developer community accounts, API-focused thought leaders, and competitor ecosystems on X. Returns 10 verified seed accounts with 1,000+ followers each. **Step 2:** You save the 6 most relevant seeds. Follower extraction begins- 200 followers per seed, totaling \~1,200 profiles. **Step 3:** Each follower is enriched and scored. CTOs, engineering leaders, and technical founders who match your ICP surface as active leads. The rest land in Monitor or Poor Fit. **Under Autopilot:** When the 6 seeds are exhausted, the scheduler automatically discovers 10 new seed accounts and creates fresh extractions. This repeats continuously. ## Troubleshooting **Not finding relevant seed accounts?** * Make sure your offer description includes relevant domain keywords * Check that your target audience is active on the platform you are searching * Broaden your offer's pain points or audience definition to give the AI more to work with * Try profile-based discovery with a known relevant account **Low ICP match rate on extracted followers?** * Review whether the seed accounts truly align with your ICP * Refine your ICP definition, particularly pain points and target audience * Try targeting more niche, specialized accounts rather than very broad ones **Seed accounts getting exhausted?** * This is expected. Previously extracted followers are not re-extracted * Under Autopilot, new seed accounts are discovered automatically * You can also add seed accounts manually to increase flow **Getting prospects with missing profiles?** * Some followers do not maintain active profiles. These are filtered out during enrichment and are not counted against your lead totals **Autopilot stall notification?** * After 3+ consecutive misses finding new seeds, you'll receive a notification * Review and update your offer description to give the AI fresh context * Consider switching channels (X, LinkedIn, or Instagram) if one platform is exhausted ## Next Steps * [**Continuous Operations**](/autopilot/continuous-operations): Automate lookalike discovery with Autopilot's daily expansion * [**Enrichment Pipeline**](/enrichment/pipeline): Understand how extracted followers get enriched and scored # Lead Pool (Instant Matching) ## Lead Pool (Instant Matching) Discover qualified prospects by matching your offer's ICP against a shared pool of pre-enriched lead profiles. Lead Pool delivers results in seconds because the profiles are already enriched - only scoring needs to run. ### What Is the Lead Pool? The Lead Pool is a shared database of enriched lead profiles. Every lead enriched through AutoReach is added to the pool. When you create an offer, AutoReach automatically matches existing pool entries against your ICP to find the best matches. Unlike keyword-based filtering, the matching understands the meaning behind your offer description. A lead whose profile is conceptually aligned with your target audience will surface even if they do not share exact keywords with your search criteria. ### How It Works **Profile Storage** - When any lead completes the enrichment pipeline, their professional profile is stored in the pool, including role, bio, industry, company, location, and skills. **ICP Matching** - When matching triggers, your offer's ICP is compared against stored profiles. The matching is based on your target audience definition, industries, and preferred locations - focusing on who the buyer is, not what you are selling. **Lead Creation** - Matched profiles are added as new leads with source marked as "Lead Pool," fresh scoring against your specific offer, and enrichment skipped since profile data is already available. Existing leads are excluded to prevent duplicates. ### When Does It Trigger? | Trigger | When It Fires | | ------------------------------ | --------------------------------- | | **New offer created** | When you create a new offer | | **Search finishes** | When any search completes | | **Follower extraction starts** | When a follower extraction begins | Updating an existing offer does not automatically trigger a pool match - only offer creation does. ### Cost Lead Pool matching is **not zero-cost**. While platform API costs are zero (leads are already enriched), AI scoring costs apply since every matched lead is queued for scoring using your configured AI models. ### Viewing Pool Leads See pool-sourced leads by filtering the **All** tab on the Leads page by "Pool DB" source. ### Progress During LinkedIn Searches When a LinkedIn search runs, pool matches triggered by that search are counted alongside the profiles fetched during the search. The processing banner shows the combined count so you can see the full result set as it is built. ### Best Practices 1. **Build the database first.** Lead Pool is most effective with a large base of enriched leads. Start with discovery searches before relying on pool matching. 2. **Write detailed target audience definitions.** The matching is built from your target audience, industries, and locations. Be specific in those fields. 3. **Monitor match quality.** Track response rates from Lead Pool matches and compare them to other discovery methods. 4. **Combine with other methods.** Use Lead Pool for rapid scaling while continuing X and LinkedIn searches for fresh intent signals. *** ## Cross-Platform Profile Matching AutoReach can find the matching profile on the other platform for your leads. If you have a prospect on LinkedIn, it can locate their X profile. If you have them on X, it can locate their LinkedIn profile. ### How to Trigger It Cross-platform matching is a **manual action**: 1. Select the leads you want to match on the Leads page 2. Click the **Enrich** button in the floating selection bar 3. In the Enrich modal, look under **Profile Discovery** 4. Toggle **Find LinkedIn** (for X leads) or **Find X Profile** (for LinkedIn leads) 5. Confirm to start the search The options are platform-aware. "Find LinkedIn" only appears for X leads, and "Find X Profile" only appears for LinkedIn leads. When you select a profile finder, the corresponding profile enrichment is automatically enabled so the found profile gets enriched immediately. ### How Matching Works AutoReach uses AI-powered web search to find matching profiles across platforms. Candidates are evaluated on name similarity, company match, source credibility, and corroboration from multiple sources. ### What Happens After a Match * **If found:** the lead record is updated with the matched profile URL, and enrichment proceeds for both platforms * **If not found:** enrichment continues with data from the source platform only LinkedIn and X profile enrichment run in parallel for leads that have URLs on both platforms. Failure on one platform does not block the other. ### Example Workflows **LinkedIn-First with X Follow-up** - Run a LinkedIn search, then use Find X Profile to locate their X accounts. Send LinkedIn messages first, then follow up on X for multi-channel outreach. **X-First with Full Professional Context** - Run a tweet search, then use Find LinkedIn to get full professional data (company, role, tenure, education, skills) for richer personalization. **Account-Based Outreach** - Run a LinkedIn people search targeting specific companies, then use Find X Profile for coordinated outreach across both platforms. ### Troubleshooting **Not finding X profiles for some LinkedIn leads?** This is expected. Not every professional maintains an active X account. **A match looks incorrect?** Low-confidence matches are filtered out automatically, but occasional false positives can occur with common names. You can manually update the lead record. ## Next Steps * [**Enrichment Pipeline**](/enrichment/pipeline): How matched profiles flow through enrichment and scoring * [**How Leads Work**](/core-concepts/leads): Unified lead profiles across platforms # Keyword Generation AutoReach uses AI to generate targeted search queries from your offer definition. Keywords power both one-time and recurring daily searches across X and LinkedIn, keeping your prospect pipeline full without manual effort. ## How It Works Keyword generation is a multi-layered system. For X, it produces simple conversational keywords alongside structured intent-based queries, competitor keywords, and niche jargon. For LinkedIn, it produces professional search phrases organized by buyer intent. Each layer is optimized for the platform's search capabilities. ### Where Keywords Come From AutoReach derives keywords from multiple elements of your offer: * **Target audience** details like titles, roles, and industries * **Pain points** described in your offer * **Competitors**- competitor-specific keywords are generated and merged separately * **Industry jargon**- niche terminology, insider terms, and community names ## X Keyword Generation X keyword generation has four components: ### Simple Keywords Short conversational phrases in first-person style ("my X keeps breaking", "need a better X") designed to match how real people talk about their problems on X. ### Intent Query Clusters Structured search queries organized by buyer intent. The AI selects categories relevant to your offer, covering common B2B signals like operational pain, evaluation and buying activity, growth and hiring signals, and industry-specific concerns. ### Competitor Keywords For each competitor in your offer, AutoReach generates keywords like "switching from \[competitor]", "\[competitor] alternative", and similar variations. ### Niche Jargon Industry-specific insider terminology- tools, certifications, acronyms, conferences, job titles, and community names that your target audience uses. ## LinkedIn Keyword Generation LinkedIn queries are short professional phrases without search operators- optimized for LinkedIn's content search. ### LinkedIn Intent Coverage The AI generates queries spanning a range of buyer intents (operational pain, solution-seeking, buying signals, growth and hiring signals) so you capture prospects at different stages. All relevant categories are searched automatically when you start a LinkedIn content search. ### Key Differences from X * Queries are professional/B2B-oriented phrases without search operators * All intent categories are searched automatically ## Previewing Keywords You can preview generated keywords before running a search. The keyword generator is available when configuring X tweet searches and LinkedIn content searches. Previewing lets you review the AI-generated keywords and queries without starting a search, so you can adjust your offer description or pain points if needed. ## Keyword Overrides For X searches, you can provide your own keywords and search query at search start time to skip AI generation entirely. LinkedIn searches always generate queries server-side. You cannot provide custom queries directly. ## Recurring Keyword Regeneration When Buyer Expansion is enabled, keywords are **regenerated fresh** on each daily run- not reused. The AI references previous keywords to ensure variation and avoid repeating the same searches. If regeneration fails, existing keywords are used as a fallback. ## Best Practices 1. **Review AI-generated keywords.** AutoReach generates solid starting points, but you know your market best. Override where it makes sense. 2. **Include industry jargon in your offer.** Niche terminology in your offer description and pain points produces better AI-generated keywords. 3. **Vary intent signals.** Mix pain-focused ("struggling with"), growth-focused ("scaling"), and evaluation-focused ("comparing") to cover different buyer stages. 4. **Let recurring searches regenerate.** Keyword rotation prevents stale results and captures different prospect segments over time. 5. **Combine with other discovery methods.** Keywords complement people search, lookalike audiences, and lead pool matching. Use them together for broader coverage. ## Example Workflow **Your Offer:** "Financial planning and analysis (FP\&A) software for growing SaaS companies" **X Keywords might include:** * "financial planning SaaS" * "FP\&A forecasting" * "budget cycle manual" * "revenue forecast inaccurate" * "switching from Anaplan" * "Adaptive Insights alternative" **X Intent Clusters might include:** * Queries about manual spreadsheet processes * Queries about evaluating FP\&A tools * Queries about slow budget cycles **LinkedIn Queries might include:** * "financial planning manual process" * "FP\&A software recommendation" * "scaling finance team" * "hiring FP\&A analyst" **Result:** Multi-layered queries across both platforms, running daily with fresh keyword rotation to surface CFOs, FP\&A leaders, and finance operators. ## Troubleshooting **Getting too many irrelevant results?** * Your keywords may be too broad. Override with more specific terms on X. * Add exclusion terms to the search. * Focus on intent clusters rather than simple keywords for better precision. **Not getting enough results?** * Broaden search terms or include synonym variations. * Increase the days back setting on the search to look further into the past. * Verify that your target audience actively discusses these topics online. **Keywords seem too generic?** * Make sure your offer includes enough domain context- industry jargon, specific pain points, and competitor names help the AI generate more targeted keywords. * Override with your own validated keywords at search start (X only). ## Next Steps * [**X Tweet Search**](/finding-leads/tweet-search): Use your generated keywords to find high-intent prospects on X * [**LinkedIn Content Search**](/finding-leads/linkedin-content-search): Apply keywords to discover decision-makers on LinkedIn # The Enrichment Pipeline The enrichment pipeline transforms raw leads into rich, actionable profiles. Starting with just a name and platform URL, AutoReach progressively layers on social profiles, work history, recent activity, and scoring to build a complete picture of every lead. ## Before and After A typical lead enters the pipeline like this: ``` Name: Jane Doe LinkedIn URL: linkedin.com/in/janedoe ``` After the pipeline runs, the same lead looks like this: ``` Jane Doe Role: VP of Demand Generation at Acme SaaS Location: San Francisco, CA Company: 250 employees, $30M revenue, B2B software Recent activity: 4 posts in last 30 days, including 2 about outbound automation and 1 mentioning a competitor of yours Fit score: 87 / Intent score: 72 / Timing score: 65 Buyer state: Active ``` That transformation (name + URL into a scored, contextualized profile) is what enrichment does. The phases below explain how it gets there. **You don't trigger this manually.** Every lead AutoReach discovers runs through the pipeline automatically. ## Pipeline Architecture The pipeline runs all enrichment phases sequentially in a single job per lead. Each lead progresses through the phases below in order. ## Pipeline Phases ### Phase 1: Profile Enrichment LinkedIn and X profile enrichment run **in parallel**: **X enrichment** (when an X profile URL exists): * Fetches profile data: user ID, DM availability, follower/following counts, verified status, profile image, bio, and name **LinkedIn enrichment** (when a LinkedIn profile URL exists): * Extracts full profile data: job title, company, work history, education, skills, certifications, languages, and network size Failure on one platform does not block the other- the pipeline continues with whatever data is available. ### Phase 2: Activity Enrichment Fetches recent posts and engagement from LinkedIn (preferred) or X. **What gets collected:** * Recent posts (structured data and text for AI consumption) * Interaction orbit- who the lead interacts with on social media (powers Dark Funnel / orbit cluster signals) * Signal date- updated if any post is more recent than the current value Activity failure is non-fatal- the pipeline continues. ### Phase 3: Scoring A full AI-powered analysis that evaluates fit, intent, and timing against your offer. Post-scoring, several operations trigger: * **Account signals:** Dark Funnel aggregation * **Autopilot enrollment:** Checks if the lead qualifies for automatic sequence enrollment ### Phase 4: Finalize * Lead status is set to ready * Lead is added to the Lead Pool (for instant matching on future offers) * Source search completion is checked ## Cross-Platform Profile Finding Profile finding (searching for the lead's profile on the other platform) runs in **separate dedicated processes**, not as part of the main pipeline. Profile finding only triggers when explicitly enabled- this is not the default flow. See [Lead Pool](/finding-leads/lead-pool) for details. ## Web Enrichment (Optional) A **separate, optional process** that is NOT part of the standard pipeline. Must be triggered independently. Gathers company info, social profiles, and additional context from the web. See [Web Enrichment](/enrichment/web-enrichment) for details. ## Pipeline Modes | Mode | What Runs | Use Case | | -------------------- | ------------------------------------------------------------ | ----------------------------------------- | | **Full pipeline** | All phases (enrich + score) | Default for new leads | | **Scoring-only** | Scoring phase only (profile and activity enrichment skipped) | Lead Pool matches | | **Enrich-only** | Profile and activity enrichment only (no scoring) | Manual re-enrichment | | **Manual re-enrich** | Selected enrichment types only | Targeted refresh of specific data | | **No-op** | Leads immediately set to ready | When neither enrich nor scoring is needed | ## Pipeline Status Values | Status | Meaning | | -------------- | --------------------------------------- | | Pending | Pipeline job queued but not yet started | | Processing | Pipeline job is actively running | | Ready | All phases complete | | Billing Paused | Paused due to billing/credits issue | While a lead is Processing, it can be temporarily waiting on a deferred job (account paused, daily budget reset, rate limit). It stays in Processing during the wait and resumes automatically when the blocker clears. ## Pipeline Progress Tracking Progress is tracked in real time and includes: * Overall status: running, completed, or paused * Counters: total, completed, enriching, scoring ## Recovery and Resilience The pipeline automatically recovers from various failure scenarios- stuck leads, interrupted jobs, and billing pauses are all handled without manual intervention. Leads that get stuck are automatically retried and eventually marked as ready so they don't block the pipeline. ## Intent-Based Prioritization Leads sourced from intent signals (people actively posting about problems your product solves) are enriched and scored ahead of standard imports. High-priority leads enter the pipeline with higher priority scores and are processed first. ## From Enrichment to Outreach Enriched data powers two core systems: **Buyer Scoring** uses the full profile, activity history, and company context to calculate how likely a lead is to convert. Leads with richer enrichment data receive more accurate scores. **Personalized Outreach** draws on enriched profiles to generate relevant messages. AI references recent posts, shared interests, role context, and company details to craft outreach that feels personal rather than templated. **Enrichment quality matters**: Leads with incomplete social profiles or limited recent activity will receive lower scores and less personalized messaging. The more data available on a lead's public profiles, the better AutoReach can serve you. ## Next Steps * [**Buyer Intelligence & Scoring**](/core-concepts/buyer-intelligence): Learn how enriched data powers fit, intent, and timing scores * [**Outreach Overview**](/outreach-and-sequences/overview): See how enriched profiles drive personalized messaging # LinkedIn Profile Data LinkedIn enrichment extracts comprehensive professional profile data from LinkedIn profiles. This data powers buyer scoring, ICP matching, and AI-generated personalized outreach. ## How It Works LinkedIn profile enrichment fetches comprehensive profile data using your connected LinkedIn account- including experience, education, skills, languages, certifications, contact information, and company data. If your LinkedIn account is temporarily unavailable, the job is automatically deferred and retried. ## What Gets Extracted ### Basic Profile Information | Field | Description | | ----------------- | ---------------------------------------------------- | | Name | Full name | | Headline | Professional headline (e.g., "VP Sales @ Acme Corp") | | Summary | About/bio section text | | Location | Geographic location | | Industry | Primary industry classification | | Profile Image | Profile picture URL | | Background Image | Banner/cover image URL | | Profile ID | LinkedIn profile identifier | | Public Identifier | Vanity URL slug | | Bio | Built from headline + summary | ### Contact Information Extracted from the lead's LinkedIn contact information: | Field | Description | | -------------- | ----------------------- | | Email | Primary email address | | Phone Numbers | Array of phone numbers | | Websites | Array of website URLs | | Twitter Handle | Linked Twitter/X handle | The first website is promoted to the lead's website URL if the lead doesn't already have one. Email is promoted to the lead's primary email if not already set. ### Work Experience Each position includes: | Field | Description | | ------------------ | -------------------------------------------- | | Title | Job title | | Company Name | Employer name | | Company URL | Company LinkedIn page URL | | Company Logo | Company logo image URL | | Company Size | Employee count range (e.g., start and end) | | Company Industries | Industry classifications | | Location | Position-specific location | | Start Date | Month and year | | End Date | Month and year, or null for current position | | Description | Position description text | | Employment Type | Full-time, Part-time, Contract, etc. | Positions are deduplicated by title, company, and year, and sorted newest-first. ### Derived Role Data From the work history, the enrichment derives current role information: | Field | Description | | --------------- | -------------------------------- | | Title | Current job title | | Company Name | Current employer | | Is Current | Whether the position is current | | Tenure (months) | Months in current role | | Tenure (days) | Days in current role (precision) | ### Education Each entry includes: | Field | Description | | -------------- | ----------------------------------- | | School Name | University, college, or institution | | School Logo | School logo URL | | Degree | Bachelor's, Master's, PhD, etc. | | Field of Study | Major or specialization | | Start Date | Month and year | | End Date | Month and year | | Description | Additional details | | Activities | Activities and societies | ### Skills Sorted by endorsement count (highest first): | Field | Description | | ----------------- | ----------------------------------------------- | | Name | Skill name (e.g., "Python", "Machine Learning") | | Endorsement Count | Number of endorsements from connections | ### Languages | Field | Description | | ----------- | ------------------------------------------------------------------------------------ | | Name | Language name | | Proficiency | Level (e.g., Native or Bilingual, Professional Working, Limited Working, Elementary) | ### Certifications | Field | Description | | --------------- | ---------------------------------------------------- | | Name | Certification name | | Authority | Issuing organization (e.g., AWS, Google, Salesforce) | | License Number | Official credential ID | | Issue Date | Month and year | | Expiration Date | Month and year | | URL | Credential verification URL | ### Network Strength | Field | Description | | ---------------- | -------------------------- | | Connection Count | Total LinkedIn connections | | Follower Count | LinkedIn followers | ### Company Data The enrichment also fetches the full company profile for the lead's current employer: | Field | Description | | ----------------- | -------------------------------- | | Company ID | LinkedIn company ID | | Name | Company name | | Description | Company description | | Tagline | Company tagline | | Industry | Company industry | | Company Type | Type of organization | | Headquarters | HQ location | | Staff Count | Estimated employee count | | Staff Count Range | Range string (e.g., "1001-5000") | | Website | Company website URL | | Founded Year | Year founded | | Specialities | Array of specialties | | Logo URL | Company logo URL | | Follower Count | Company page followers | The staff count is promoted to the lead's employee count field for lead filtering. **Company Jobs**: Open job postings at the company, each with title, location, and posting date. Used as a hiring signal. **Company Posts**: Recent company page posts with text, date, and URL. Used for signal detection. ### Recent Activity When running as part of the pipeline, LinkedIn enrichment also fetches the lead's recent posts and engagement (if not already fetched): | Field | Description | | ----------------- | ----------------------------------------------------- | | Recent Posts | Array of structured post objects | | Posts Text | Concatenated text with date labels for AI consumption | | Fetch Timestamp | When posts were fetched | | Empty Flag | True if no posts found | | Interaction Orbit | Who the lead interacts with on social media | | Signal Date | Updated if a newer post is found | ## Error Handling Transient errors (rate limits, account busy, budget exceeded) are automatically deferred and retried. Permanent errors (no LinkedIn URL, profile doesn't exist) are logged and the pipeline continues- enrichment failure does not block scoring or other enrichment steps. If your LinkedIn session expires, you'll need to reconnect your account. ## Data Quality LinkedIn profiles vary in completeness. Missing fields are simply not populated- scoring accounts for incomplete data. A lead with only a headline and current company still gets enriched and scored, just with less context for personalization. Company data and company posts are fetched on a best-effort basis. If either fails, the lead continues through the pipeline with profile-level data only. ## Next Steps * [**X Profile Data**](/enrichment/twitter-data): See what data is extracted from X profiles * [**Web Enrichment**](/enrichment/web-enrichment): Discover how company-level data is gathered from the web * [**Enrichment Pipeline**](/enrichment/pipeline): See how LinkedIn enrichment fits into the full pipeline # X/Twitter Profile Data X enrichment extracts profile information and recent activity from X (Twitter) profiles. This data powers buyer scoring, outreach personalization, and - most importantly - determines whether you can send DMs to a lead without following them first. ## How It Works X profile enrichment fetches the full profile data using your connected X account. If no X account is available, enrichment skips and the pipeline advances. ## What Gets Extracted ### Profile Information | Field | Description | | --------------- | ------------------------------------------------ | | Twitter User ID | Unique account identifier | | Name | Display name | | Username | Handle (derived from profile URL) | | Bio | Profile description | | Follower Count | Number of followers | | Following Count | Number of accounts followed | | Verified | Blue checkmark or verification status | | Profile Image | Profile picture URL | | Can DM | Whether you can send direct messages (see below) | ### The Can-DM Flag The most important piece of X data is the **can-DM flag**. This boolean controls whether your outreach can land directly in someone's inbox. * **Can-DM = True** - You can send direct messages without following the account first. Your initial message goes straight to their inbox. * **Can-DM = False** - Your first message will appear in their "Message Requests" folder, which has much lower engagement. ### Recent Activity After profile data is fetched, recent posts are collected: | Field | Description | | ----------------- | ----------------------------------------------------- | | Recent Posts | Array of structured post objects | | Posts Text | Concatenated text with date labels for AI consumption | | Interaction Orbit | Who the lead interacts with on social media | | Signal Date | Updated if a newer post is found | Older posts are filtered out to keep activity data relevant. Activity data is not required for the pipeline to advance - if it can't be fetched, scoring continues without it. ## Error Handling Transient errors (rate limits, network issues) are automatically retried. Permanent errors (profile not found, no account available) are logged and the pipeline continues- X enrichment failure does not block LinkedIn enrichment, scoring, or other pipeline phases. ## Data Quality * **Suspended accounts** - profile data cannot be retrieved (404) * **Private/protected accounts** - limited data available; recent posts may not be accessible * **New accounts** - may have minimal activity for personalization * **Deleted accounts** - return 404 and enrichment skips ## How X Data Complements LinkedIn | Dimension | LinkedIn | X | | ------------------- | ------------------------ | -------------------------------------- | | **Profile Type** | Professional credentials | Interest signals and personality | | **Data Freshness** | Updated infrequently | Updated daily or hourly | | **What's Shown** | Formal work history | Real-time thoughts and interests | | **Outreach Value** | Context and credibility | Conversation starters and intent | | **DM Reachability** | Not applicable | Can-DM flag critical for first contact | ## Next Steps * [**LinkedIn Profile Data**](/enrichment/linkedin-data): See what professional data is extracted from LinkedIn * [**Web Enrichment**](/enrichment/web-enrichment): Learn how company-level context is gathered from the web * [**Enrichment Pipeline**](/enrichment/pipeline): See how X enrichment fits into the full pipeline # Web Enrichment Web Intelligence uses AI-powered web search to gather additional context about leads - company details, social profiles, contact information, recent news, and professional achievements. It is a **separate, optional process** that is not part of the standard enrichment pipeline. ## When to Use Web Enrichment Web Enrichment is **opt-in and costs extra AI usage**, so most users don't run it on every lead. Reach for it when: * You're working a small list of **high-value accounts** and the default LinkedIn/X enrichment data feels too thin to write a great message. * A specific lead's **company isn't well-known** and you need funding stage, tech stack, or recent news for context. * You want to find **additional contact channels** (GitHub, Substack, podcast appearances) for a lead you can't reach via DM. Skip it when you're running high-volume outreach where standard enrichment is good enough. The cost adds up across hundreds of leads, and the lift in personalization quality is marginal for most B2B offers. ## How It Works Web Intelligence uses AI web search to gather three categories of data: * **Identity & Company**- current company, role, website, size, industry, funding, tech stack, headquarters * **Social & Contact**- additional social profiles (GitHub, YouTube, Medium, Substack, personal website) and contact information (phone, additional emails, scheduling links) * **Activity & Insights**- recent news mentions, speaking engagements, publications, podcast appearances, and achievements (last 6 months only) ### Preconditions A lead must have **at least one** of: * A LinkedIn profile URL * A website URL * A bio longer than 20 characters Leads without any of these are skipped. ## What Gets Extracted ### Company Information | Field | Description | | --------------- | ------------------------------------------------------------- | | Company Name | Company name | | Company Website | Company domain URL | | Company Size | Employee range ("1-10", "11-50", "51-200", "201-500", "500+") | | Industry | Primary industry classification | | Description | Brief company description | | Founded Year | Founding year | | Headquarters | Primary office location | | Funding Stage | Funding stage (Seed, Series A, B, C, etc.) | | Funding Amount | Total raised to date | | Investors | Array of known investor names | | Technologies | Array of tech stack tools and frameworks | If the lead does not already have an employee count value, the company size is promoted for filtering. ### Social Profiles | Field | Description | | ---------------- | ----------------------------------- | | Twitter | X/Twitter profile URL | | GitHub | GitHub profile | | YouTube | YouTube channel | | Medium | Medium blog | | Substack | Substack newsletter | | Personal Website | Portfolio, consulting site, or blog | | Other | Array of other platform profiles | ### Contact Information | Field | Description | | ----------------- | ----------------------------------------------- | | Phone | Phone number | | Additional Emails | Array of work or personal emails beyond primary | | Calendly | Scheduling page URL | ### News Mentions Array of recent news items (last 6 months only): | Field | Description | | ------- | --------------------------------------- | | Title | Headline | | Source | Publication name | | URL | Link to the article | | Date | Publication date (YYYY-MM-DD, required) | | Summary | Brief description | ### Insights | Field | Description | | -------------------- | ------------------------------------------- | | Notable Achievements | Array of professional accomplishments | | Speaking Engagements | Conferences and events | | Publications | Articles, whitepapers, or research authored | | Podcast Appearances | Interviews and podcast guest slots | ### Confidence & Sources | Field | Description | | --------------------- | ---------------------------------- | | Enrichment Confidence | Overall data reliability indicator | | Enrichment Sources | List of web sources used | | Enrichment Summary | AI-generated overview of the lead | Web enrichment data is only included in scoring context when overall confidence is sufficient. ## How It's Triggered Web Intelligence is **not part of the standard pipeline** and does **not run automatically** when leads are added. It must be triggered explicitly. You can trigger Web Intelligence manually from the Leads page by selecting leads and enabling the Web Intelligence option. You can also preview costs before enriching, retry failed or skipped leads, and check enrichment progress - all from the same interface. ## Cost Cost estimation is available before starting enrichment. Cost varies by AI model configuration. ## Error Handling Transient errors (rate limits) are automatically retried. Other errors are recorded and the job completes- web enrichment failure does not block other enrichment steps. ## Next Steps * [**Email and Website Finding**](/enrichment/optional-enrichment): Discover work email addresses and company websites * [**Enrichment Pipeline**](/enrichment/pipeline): See how the standard enrichment pipeline works # Optional Enrichment (Email & Website) Email finding and website finding are **optional enrichment steps** that are not part of the standard pipeline. You trigger them explicitly from the Leads page. ## When You Need These Run these only if you want to **add email as an outreach channel** for leads you found on LinkedIn or X. Without an email address, you can only reach a lead on the platform where they were discovered. With one, email becomes available as a sequence step. If you're staying with LinkedIn/X-only outreach, you can skip both steps entirely. If you do want email outreach, **order matters**: run website finding first, then email finding. Email finding searches against a known company domain, so if the website hasn't been discovered yet, email finding has much lower hit rates. *** ## Website Finding Website finding discovers company and personal websites for leads using AI-powered web search. Found websites improve email finding accuracy (providing a domain to search against) and give AI more context for outreach personalization. ### How It Works Website finding uses AI web search to find a lead's company or personal website. It tries multiple approaches and returns the best match. A lead must have at least one of: * A LinkedIn profile URL * A bio longer than 20 characters Leads without either are skipped. ### What Gets Stored | Field | Description | | -------------- | ------------------------------------------------------- | | Website URL | The discovered website URL | | Website Source | How it was found (LinkedIn company, bio, direct search) | ### Domain Filtering Website finding filters out social media platforms, link aggregators, URL shorteners, scheduling tools, and other non-website domains. Only actual company or personal websites are accepted. ### How to Trigger Select leads from the Leads page and enable the website finding option when triggering enrichment. You can also retry failed or skipped leads from the same interface. *** ## Email Finding Email finding discovers work email addresses for leads using the Findymail API. It requires your own Findymail API key. ### How It Works Email finding makes a single Findymail API call per lead, passing the lead's name and company domain (or company name as fallback). Findymail handles verification on their servers and returns a verified email address or null. AutoReach validates the returned email's format and filters out generic/role-based addresses (info@, support@, sales@, etc.) before storing it. ### Preconditions For email finding to run: 1. **User has a Findymail API key** configured in Settings 2. **Lead has a name** (first name is extracted and required) 3. **Lead has at least one of:** * A website URL (domain is extracted) * A LinkedIn profile URL (LinkedIn company website used) * A company name from LinkedIn data * A bio longer than 20 characters ### What Gets Stored | Field | Description | | ----------------- | ---------------------------------------------------- | | Email | The discovered email address | | Email Verified At | Timestamp if Findymail reports the email as verified | ### Setup Go to **Settings > AI & Models** and enter your Findymail API key. Once saved, AutoReach will use Findymail's database to look up verified work email addresses for your leads. If you do not have a Findymail account, you can sign up at findymail.com and add your key afterward. ### Cost Email finding uses credits from your **Findymail account**, not from AutoReach. If credits are exhausted, Findymail returns an error and email finding stops until credits are replenished. ### How to Trigger Select leads from the Leads page and enable the email finding option when triggering enrichment. You can also retry failed or skipped leads from the same interface. *** ## Running Website Finding Before Email Finding Website finding feeds into email finding. The email finder uses the lead's website URL to extract a company domain, which is the preferred input for Findymail lookups. Running website finding before email finding improves email discovery rates by providing a verified domain. This is not a strict dependency. Email finding can proceed without a website URL by falling back to LinkedIn company data or the lead's bio. ## Error Handling For both services, transient errors (rate limits, network issues) are automatically retried. Permanent errors are recorded on the lead. Leads missing required data are skipped. ## Troubleshooting **Email finding not working?** * Verify your Findymail API key is configured correctly in Settings * Check your Findymail credit balance **No emails found for some leads?** * Not every lead has a discoverable email. Success rates vary by industry and company size * Ensure leads have a website URL or company name * Run website finding first to populate website URLs for better domain coverage ## Next Steps * [**Web Enrichment**](/enrichment/web-enrichment): Gather additional company context from the web * [**Enrichment Pipeline**](/enrichment/pipeline): See how the standard enrichment pipeline works # Overview Sequences are automated multi-step campaigns that execute engagement actions across X, LinkedIn, and Instagram, plus email. Each sequence walks leads through a configured flow of actions (likes, follows, DMs, comments, connection requests, emails) with AI-personalized messaging and natural timing. ## What Is a Sequence? A sequence is a series of steps that execute automatically against enrolled leads. Each step performs an action (like a post, send a DM, follow the account) with configurable delays between steps. Sequences support branching logic via condition steps that route leads based on whether they replied. A typical sequence might look like: 1. Day 1: Like their recent post 2. Day 2: Follow them 3. Day 3: Send a personalized DM 4. Day 5: Condition- if replied, stop. If not, send a follow-up DM. ### Sequence Statuses | Status | Description | | ---------- | ---------------------------------------------- | | Draft | Initial state- can be configured and edited | | Scheduling | Actions are being scheduled for enrolled leads | | Active | Running- actions are scheduled and executing | | Paused | Temporarily stopped (can be resumed) | | Completed | All leads have finished the sequence | | Failed | Sequence encountered a critical error | ### Platform Support When creating a sequence, you select which accounts to use: an X account, a LinkedIn account, an Instagram account, or a combination. At least one social account is required. Email steps become available when a mailbox is connected to the parent X or LinkedIn account. Individual steps specify which platform they target, so a single sequence can mix steps that execute on different platforms. ## Supported Actions Each step in a sequence performs one of these action types: | Action | Platforms | Description | | ------------------ | ----------------------------- | ----------------------------------------------------- | | DM | X, LinkedIn, Instagram | Send a direct message (AI-personalized) | | Email | Email (Gmail/Outlook) | Send an email from a connected mailbox | | Like | X, LinkedIn, Instagram | Like/react to a lead's recent post | | Comment | LinkedIn, Instagram | Comment on a lead's post | | Follow | X, Instagram | Follow the account | | Watch Story | Instagram | View a lead's active Instagram stories | | Connection Request | LinkedIn | Send a LinkedIn connection request | | Condition | X, LinkedIn, Instagram, Email | Branch based on lead status (replied vs. not replied) | Each step has a defined order, platform-specific settings, positioning data for the visual flow editor, and a reference to the next step for linear flow progression. ## AI Message Personalization Every message sent through a sequence is personalized using the lead's enriched data. The system supports two layers: **Direct substitution** - template variables replaced with lead data (identity, network stats, company info, enrichment insights, contact info, and more) **AI-inferred placeholders** - filled by the AI model from lead context (`{{role}}`, `{{pain_point}}`, `{{question}}`, and any custom placeholder you define) AutoReach provides 30+ built-in template variables across categories like identity, company data, professional profile, web enrichment insights, and contact info. The AI draws on the lead's profile data, recent activity, enrichment intelligence (tech stack, funding, company context), the offer's knowledge base (RAG), and tone examples configured on the sequence. See [**DM Personalization**](/outreach-and-sequences/dm-personalization) for the full variable reference. ## Sequence Execution ### Scheduling Actions are scheduled automatically and executed one at a time to avoid rate limit issues. **Lead selection order:** Higher-priority leads (based on buyer score and signal recency) get earlier time slots. **Step progression:** Each step's connection determines what happens next. Condition steps evaluate lead status and branch accordingly. After each action completes, the next step is queued with the configured delay. ### Natural Timing Sequence execution uses natural pacing - actions are spaced across the day, grouped into human-like sessions, and include realistic delays for reading and typing. Activity varies by time of day and day of week to match natural patterns. ### Activity Window Outreach only executes during the configured activity window. Activity windows are **per-account**: each connected LinkedIn, X, and Instagram account has its own schedule and timezone. A window is defined as one or more time blocks per day (for example a morning block plus an afternoon block), capped at **8 hours of total active time per day**. The default for a new account is 09:00-13:00 plus 14:00-18:00 (8h), Monday to Friday, weekends off. When an account acts on a lead, the acting account's window governs the schedule. Actions outside the window are deferred to the next open slot. Configure per-account in **Accounts → \[account] → Configuration → Activity window** card. ## Lead Statuses in a Sequence | Status | Description | | -------------- | ------------------------------------------------ | | Pending | Enrolled but not yet contacted | | Active | First action executed, progressing through steps | | Paused | Temporarily paused (sequence paused or manual) | | Replied | Lead has replied to at least one message | | Meeting Booked | Meeting scheduled | | Completed | Sequence flow completed for this lead | | Failed | Action execution failed | | Lost | Manually removed from sequence | ## Daily Send Limits Each sequence has a configurable daily action limit (per-sequence) that protects account health. Limits reset at midnight in the acting account's timezone. If a limit is reached, remaining actions are deferred to the next day. LinkedIn connection requests are also tracked under a separate per-account daily limit. See [Scheduling](/outreach-and-sequences/scheduling) for the full limits table, enforcement details, and daily connection limits. ## Conversation Follow-Ups Sequences support automatic follow-ups for stale conversations: | Setting | Default | Range | | ------------------------------ | ------- | ----- | | Conversation follow-up enabled | false | - | | Follow-up wait days | 3 | 1–30 | | Max follow-up count | 2 | 1–10 | The system finds conversations with no recent activity and schedules follow-up messages automatically. ## Adding Leads to Sequences **Validation before enrollment:** * Lead must have platform data matching the sequence (X profile for X sequences, LinkedIn for LinkedIn sequences) * Lead must not already be enrolled in this sequence * Lead must not be blacklisted * If "Skip contacted leads" is enabled: lead must have no previous contact history * For X sequences: lead must be able to receive DMs Leads failing validation are skipped with a reason. If the sequence is active, scheduling begins immediately for new leads. ## Reply Detection and Opt-Out **Reply detection** is passive- incoming messages in a conversation update the lead's status to "Replied" and increment the sequence's reply counter. **Opt-out handling** uses a blacklist system: * Blacklist accounts manually from the Leads page * Blacklisted leads are automatically removed from all sequences * Blacklisted accounts cannot be re-enrolled * The system does not auto-blacklist based on replies- opt-out decisions are left to the user ## Sequence Metrics | Metric | Description | | --------------- | --------------------------------------------------- | | Leads Added | Total leads enrolled | | Contacted | Leads that received at least one action | | Replied | Leads with status "Replied" or "Meeting Booked" | | Meetings | Leads with a meeting scheduled | | Reply Rate | Replied / Contacted | | Conversion Rate | Meetings / Replied | | Acceptance Rate | Connection requests accepted / sent (LinkedIn only) | Metrics are updated automatically when lead statuses change. Step-level execution stats are available via the activity log, which tracks every action with status, timestamps, and execution data. ## Key Sequence Settings | Setting | Description | | --------------------------------- | ----------------------------------------------------------------------------------------------------------------------- | | Daily Actions | Max total actions per day across all channels (likes, comments, follows, DMs, connection requests, emails). Default 20. | | Auto-Enroll Active Buyers | Automatically add leads scored as active for this offer | | AI Prompt | Custom AI instructions for message generation | | DM Generation Prompt | Specific prompt for cold DM generation | | How AI Comments on Posts | Instructions for AI comment generation on LinkedIn posts (LinkedIn only) | | Skip Contacted Leads | Skip leads with prior contact history | | Skip Negative Posts | Skip leads with negative content in recent posts (LinkedIn only) | | Skip Old Posts (days) | Skip leads whose recent posts are older than N days (LinkedIn only) | | Prefer Original Posts | Prefer original posts over reposts for engagement | | AI Replies Off by Default | Disable AI auto-responses for new conversations | | Max AI Responses per Conversation | Cap on AI responses per conversation (default 0 = unlimited) | ## Next Steps * [**Building Sequences (Flow Editor)**](/outreach-and-sequences/building-sequences): Learn the sequence builder interface * [**Supported Actions by Platform**](/outreach-and-sequences/supported-actions): Detailed reference for each action type on X vs LinkedIn * [**DM Personalization**](/outreach-and-sequences/dm-personalization): How AI generates personalized first messages * [**Scheduling**](/outreach-and-sequences/scheduling): Activity windows, timing, and send cadence * [**Simulation and A/B Testing**](/outreach-and-sequences/simulation-and-testing): Preview messages and compare variants # Building Sequences (Flow Editor) A sequence is your outreach playbook: an ordered list of actions (likes, follows, connection requests, DMs, emails) that AutoReach runs against every lead enrolled in it. You build one sequence, and it runs the same flow for hundreds of leads on your behalf. That is how outreach scales. The flow editor is a visual sequence builder. You design sequences as connected nodes on a canvas, configure each step's action and timing, then save and start the sequence. > **You usually don't start here.** If you're new, let Autopilot create your first sequence: it picks the right default flow for your platform. Open the flow editor when you want to customize the steps, change timing, add branching, or build a sequence that doesn't match the default templates below. ## Creating a New Sequence 1. Navigate to **Sequences** in the sidebar 2. Click **Create Sequence** 3. Enter a sequence name 4. Associate an offer (required- the create button is disabled without one) 5. Select platform accounts: at least one of Twitter, LinkedIn, or Instagram is required 6. Click **Create** The sequence is created in Draft status and opens in the flow editor with a default template applied based on the selected platform(s). Once set, platform accounts **cannot be changed** on a sequence- this prevents workflow disruption mid-execution. ## Default Templates When a new sequence is created, a default flow template is applied based on the configured accounts: **X (Twitter) template:** ``` Like → Follow → DM ``` **LinkedIn template:** ``` Connection Request → Condition: Connection Accepted? (retry 3d × 3) → DM (if accepted) ``` **Instagram template:** ``` Like → Comment → Watch Story → Follow → Condition: Followed Back? → DM (if followed back) ``` **Combined X + LinkedIn template (both accounts selected):** ``` Like X → Like LinkedIn → Follow (X) → Connection Request (LinkedIn) → Like X → Condition: Connection Accepted? → TRUE: DM LinkedIn → FALSE: DM X ``` Each template includes pre-configured delays between steps. Templates can be modified after creation - adjust the timing and steps to match your outreach style. Use the "Reset to Default" dialog to restore the original template. ## The Flow Editor Interface The flow editor displays your sequence as nodes connected by edges on a canvas: * **Nodes** represent actions (like, DM, follow, etc.) or logic (conditions) * **Edges** (arrows) show the execution flow between steps * **Handles** (small circles on nodes) are connection points- drag from a source handle to a target handle to create edges * **Drag nodes** to reposition them on the canvas There is no explicit "Start" node- the entry point is the topmost node (lowest Y position on the canvas). ### Adding Steps **Using the "+" button:** Click between nodes or at the end of the flow to open the step popover and select an action type. **Drag-to-connect:** Drag from a node's source handle and drop on empty canvas to open a creation menu where you select the new step type. ### Deleting Steps Hover over a node to reveal the delete button. When a step is deleted, the chain is repaired- any step that pointed to the deleted step is automatically updated to point to the deleted step's next step, preserving flow continuity. ### Saving Saving is **manual**- click the "Save Sequence" button. An "Unsaved changes" indicator appears when the flow has been modified. The save operation validates the flow structure before persisting. **What gets saved:** * **Visual layout**- node positions, edges, and viewport state for the flow editor * **Execution model**- action configurations, step order, and step connections ## Action Types | Action | Platforms | Description | | ------------------- | ----------------------------- | ---------------------------------------------------------------- | | Like | X, LinkedIn, Instagram | Like a lead's recent post | | Comment | LinkedIn, Instagram | Comment on a lead's post (AI-generated from post content) | | Follow | X, Instagram | Follow the account | | Watch Story | Instagram | View a lead's active Instagram stories | | DM | X, LinkedIn, Instagram | Send a direct message | | Email | Gmail, Outlook | Send an email from a connected mailbox | | Condition | X, LinkedIn, Instagram, Email | Branch based on lead status | | Connection Request | LinkedIn | Send a connection request | | Withdraw Connection | LinkedIn | Auto-scheduled by the auto-withdraw setting (not added manually) | ## Configuring Steps Click any node to open its configuration panel on the right side. ### Common Settings (All Action Types) | Setting | Description | | ------------- | ------------------------------------------------------------------ | | Platform | X, LinkedIn, or Instagram (for actions that support more than one) | | Delay Days | Days to wait before executing (0 = immediate) | | Delay Minutes | Additional minutes (0–59) | | Label | Optional display name for the step | Total delay = (days x 24 hours) + minutes. Displayed on the node as "Delay: 1d 15m" or "Immediate" for zero delay. ### DM Configuration | Setting | Description | | ---------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Message template | Optional. Leaving it empty uses prompt-only mode (recommended): the AI writes the full message from your DM generation prompt, the lead's data, and offer context. Adding a template with `{{variable}}` placeholders switches the step into legacy slot-filling mode, where the AI only fills the placeholders. | | AI prompt | Optional custom AI instructions for personalization | Use the variable picker to browse available placeholders. In prompt-only mode the AI personalizes the whole message from the lead's profile data, recent activity, and offer context. ### Email Configuration | Setting | Description | | ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Subject | Email subject line, supports `{{variable}}` placeholders. Falls back to a default if left empty. | | Body | Email body, supports `{{variable}}` placeholders. With AI personalization on, the body can be left as a brief prompt and the AI writes the full email from the sequence's email generation prompt and lead context. | | AI personalization | Optional. When on, AI rewrites the body per lead using profile and offer context | Email steps require at least one connected mailbox (Gmail or Outlook). Recipient domain determines which mailbox sends; neutral domains use your Primary mailbox. See [Email Channel](/outreach-and-sequences/email-channel) for full details. ### Comment Configuration | Setting | Description | | --------- | ----------------------------------------------------- | | AI prompt | Optional - custom instructions for comment generation | Comments are AI-generated from the lead's most recent LinkedIn post content. If no AI prompt is set, the sequence's comment prompt ("How AI comments on posts") is used as default instructions. ### Connection Request Configuration | Setting | Description | | ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------ | | Connection note | Optional personalized note. Notes are capped at 200 characters at send time (LinkedIn's limit); longer notes are trimmed to the last whole word. | | Auto-withdraw days | Auto-withdraw if not accepted within N days (default 14, range 1–30) | The connection note supports `{{variable}}` placeholders for personalization. It can also be AI-generated from a note prompt instead of a fixed template. ### Condition Configuration | Setting | Description | | --------------------- | ------------------------------------------- | | Condition type | What to evaluate (see table below) | | Retry interval (days) | Days between rechecks | | Max retries | Max rechecks before falling to FALSE branch | **Condition types:** | Type | Description | | ------------------- | --------------------------------------------------------------------------- | | Replied | Lead replied to a previous message | | Followed back | Lead followed back (X and Instagram) | | Has profile | Lead has a filled-out profile on the target platform | | Has email address | Lead has a usable email on file (valid on any platform and for Email steps) | | Connection accepted | LinkedIn connection request was accepted | Condition nodes have two outgoing branches: **TRUE** (left handle, green) and **FALSE** (right handle, red). Each branch can connect to a different next step. The condition is evaluated at the scheduled time- if false and retries remain, it rechecks after the retry interval. After exhausting retries, the FALSE branch executes. Example: "Recheck every 3 days, up to 3 times" = 9 days max before falling to the FALSE branch. ## Flow Validation ### Editor Validation (Before Save) * At least one node must exist * All action nodes must be reachable from the entry node (topmost) * No orphaned nodes * Condition nodes must have at least one branch edge (true or false) ### Pre-Start Validation (Before Activation) * Sequence must not already be active * At least one configured account must exist and not be paused for health issues * At least one step must be configured * At least one pending or active lead must be enrolled ## Sequence Settings Beyond the flow itself, configure overall sequence behavior: ### Enrollment & Lead Filters | Setting | Description | | --------------------- | ------------------------------------------------------------ | | Skip contacted leads | Don't re-reach leads contacted in other sequences | | Skip negative content | Skip leads with controversial or toxic posts (LinkedIn only) | | Skip old posts (days) | Only reach leads with activity within N days (LinkedIn only) | | Prefer original posts | Prefer original posts over reposts for engagement | ### Daily Limits Each sequence has a single unified daily action limit covering all channels (likes, comments, follows, DMs, connection requests, and emails). LinkedIn connection requests are also tracked per-account separately. See [Scheduling](/outreach-and-sequences/scheduling) for full enforcement details. ### AI Temperature Controls how strictly the AI follows your prompt rules vs. allowing creative variation. Range 0 (strict) to 1 (default). Lower values produce more predictable, rule-following responses. This setting is only available when your content writing model supports temperature control. ### AI & Responses | Setting | Default | Description | | --------------------------------- | ------------- | ------------------------------------------- | | Max AI responses per conversation | 0 (unlimited) | Max auto-replies per conversation (max 100) | | AI replies off by default | false | New conversations start with AI off | ### Follow-Up | Setting | Default | Range | | ------------------------------ | ------- | ----- | | Conversation follow-up enabled | false | - | | Follow-up wait (days) | 3 | 1–30 | | Max follow-ups | 2 | 1–10 | ### AI Prompts | Setting | Description | | ------------------------ | ------------------------------------------------------------------------ | | AI prompt | General AI instructions for message generation | | DM generation prompt | Specific prompt for cold DM generation | | How AI comments on posts | Instructions for AI comment generation on LinkedIn posts (LinkedIn only) | ### Outreach Persona Configure the sender identity and sales flow used by AI when generating messages: | Setting | Description | | --------------------- | ---------------------------------------------------------- | | Sender name | Name used in messages (defaults to your profile name) | | Sender role | Your role/title for context | | Sender is the founder | Toggle if you are both the sender and the company founder | | Founder name | Company founder's name (hidden when sender is the founder) | **Sales Flow** defines the steps in your closing flow (e.g., video call, free tool, website). Each step has a type, label, link, and an optional description explaining when to use that step. The AI references these steps when guiding leads toward a booking. ## Starting and Pausing | Action | Effect | | --------- | ----------------------------------------------------------------------------------- | | **Start** | Sets status to Active, triggers background scheduling of actions for enrolled leads | | **Pause** | Temporarily stops all pending actions without losing progress. Can be resumed. | ## Common Sequence Patterns ### Soft Engagement First ``` Like → Follow (1 day) → DM (2 days) ``` Builds familiarity before direct contact. The lead sees engagement before receiving a message. ### Direct with Conditional Follow-Up ``` DM → Condition: Replied? (2 days) → TRUE: Stop → FALSE: Follow-Up DM ``` ### LinkedIn Connection-First ``` Connection Request → Condition: Accepted? (retry 3d × 3) → TRUE: DM → FALSE: End ``` ## Managing Sequences ### Duplicating a Sequence To reuse an existing sequence design, click **Duplicate** from the sequence menu. You will be asked to provide a new name and select which accounts to use (X, LinkedIn, or both). The duplicate copies all steps, flow layout, AI prompts, and settings - but starts with no enrolled leads, so you can use it for a different audience or offer. ### Viewing the Lead Timeline Click on any lead within a sequence to see their individual action timeline. The timeline shows every step that has been completed, is currently pending, or has failed - along with scheduled times, execution timestamps, and any error details. This is useful for understanding exactly where a specific lead is in the sequence flow. ### Retrying Failed Actions When an action fails (due to rate limits, network errors, or platform issues), you can retry it by clicking the retry button on the failed action in the lead timeline. Retried actions are rescheduled and re-enter the execution queue. ### Filtering Sequence Leads The sequence leads table supports filters to help you focus on specific subsets of the enrolled audience: | Filter | Description | | ------------------------------------- | ----------------------------------------------------------------------------------- | | Status | Lead lifecycle status (Pending, Active, Replied, Meeting Booked, etc.) | | Current step | Which step the lead is currently on | | Has replied / Has meeting / Contacted | Boolean filters for common outcomes | | Followers range | Filter by follower count (min / max) | | Enrolled date | Filter by enrollment timestamp (after / before) | | Next action type | What action is scheduled next (DM, Like, Connection Request, etc.) | | Scheduled | When the next action is due (e.g., today, next 7 days) | | Connection | LinkedIn connection state: Pending (Sent), Pending (Received), Connected, Withdrawn | Active filters appear as removable chips above the table. Use the sort dropdown alongside to reorder by last action time, score, name, or enrollment date. ## Next Steps * [**Supported Actions**](/outreach-and-sequences/supported-actions): Detailed reference for each action type * [**DM Personalization and Template Variables**](/outreach-and-sequences/dm-personalization): How AI generates personalized first messages * [**Scheduling**](/outreach-and-sequences/scheduling): Activity windows, timing, and send cadence * [**Simulation and A/B Testing**](/outreach-and-sequences/simulation-and-testing): Preview messages and compare variants # Supported Actions by Platform Reference for all sequence action types, how they execute on each platform, and the checks and behaviors that govern them. ## How to Use This Page This is a reference, not a tutorial. You'll come here to answer specific questions like *"can I send a comment on X?"* (no, it is LinkedIn and Instagram only) or *"what does Withdraw Connection actually do?"* (cleans up unanswered connection requests so you can retry later). If you're building your first sequence, start with the default templates in [Building Sequences](/outreach-and-sequences/building-sequences). They already combine these actions in proven flows (e.g., Like → Follow → DM on X, or Connection Request → DM on LinkedIn). Come back here when you want to add or replace a step and need to know what's available, what platform supports it, and what guardrails apply. The platform support matrix is the table below. Anything marked with `-` isn't available on that platform: for example, you can't send a Comment on X or a Connection Request on email. ## Actions Reference | Action | X | LinkedIn | Instagram | Email | Description | | ------------------- | - | -------- | --------- | ----- | -------------------------------------------- | | Like | Y | Y | Y | - | Like/react to a lead's recent post | | Comment | - | Y | Y | - | Comment on a lead's post | | Follow | Y | - | Y | - | Follow the account | | Watch Story | - | - | Y | - | View a lead's active Instagram stories | | DM | Y | Y | Y | - | Send a direct message | | Email | - | - | - | Y | Send an email through your connected mailbox | | Connection Request | - | Y | - | - | Send a LinkedIn connection request | | Withdraw Connection | - | Y | - | - | Withdraw a pending or accepted connection | | Condition | Y | Y | Y | Y | Branch based on lead status | ## Action Details ### Like **Platforms:** X, LinkedIn, Instagram Likes a lead's recent post. The post is selected automatically. **Post selection:** * Fetches the lead's recent posts * Filters out posts already interacted with across **all sequences** (not just the current one) * Respects the "prefer original posts" and "skip old posts" settings * Selects the most recent qualifying post **Execution:** * **X:** Likes the tweet via the X API * **LinkedIn:** Likes the post via the LinkedIn API * **Instagram:** Likes one of the lead's recent posts, skipping any already liked If no qualifying posts are found, the action is skipped and the sequence advances. *** ### Comment **Platforms:** LinkedIn, Instagram Posts an AI-generated comment on a lead's recent post. **Post selection:** * Same logic as Like but also filters out posts already commented on (global check across all sequences) * If the skip negative content setting is enabled, posts with tragic or negative content are skipped (checked via AI) * Attempts to coordinate with a prior like action - commenting on the same post that was recently liked **Comment generation:** * AI generates the comment using the post text as context * Uses the step-level AI prompt if configured, otherwise falls back to the sequence's comment prompt ("How AI comments on posts") * Supports multilingual comments via the offer's language setting **Execution:** * **LinkedIn:** Posts a comment via the LinkedIn API. If commenting is disabled on the first post, tries the next candidate post. * **Instagram:** Posts a comment on one of the lead's recent posts, skipping posts already commented on and ones where comments are disabled. *** ### Follow **Platforms:** X, Instagram Follows the lead's account. LinkedIn uses Connection Request instead. **Execution:** * Checks if already following- skips if already following * Simulates natural browsing behavior before following * Follows the account *** ### Watch Story **Platforms:** Instagram only Views the lead's active Instagram stories. A light-touch action that puts you on the lead's story-viewer list, often used as a warm-up step before a follow or DM. **Execution:** * Fetches the lead's active stories * Marks them as viewed * If the lead has no active stories, the action is skipped and the sequence advances *** ### Send DM **Platforms:** X, LinkedIn, Instagram Sends a direct message to the lead. The message template is personalized using lead data and AI. **Platform routing:** Each DM step is configured for a specific platform. The system validates the lead has the required profile data for that platform before sending. **Message personalization:** * Template variables are replaced with lead data: `{{name}}`, `{{first_name}}`, `{{bio}}`, `{{company}}`, `{{company_size}}`, `{{location}}`, `{{followers}}`, etc. * Special placeholders: `{{post}}` (lead's latest post), `{{replied_post}}` (post from a prior comment action) * AI fills remaining smart placeholders (`{{role}}`, `{{pain_point}}`, custom placeholders) using enrichment data, web enrichment, and knowledge base context * Booking URL is injected with platform-specific tracking if configured **X DM execution:** * Checks whether the lead can receive DMs. In an X-only sequence, a lead that cannot be DM'd is removed (its remaining actions are skipped) so it does not endlessly reschedule. * Creates or retrieves the conversation * Sends via the X API **LinkedIn DM execution:** * Checks connection status- LinkedIn requires an accepted connection to send DMs * If an existing conversation already has messages, skips the template DM and queues an AI response instead (avoids duplicate cold outreach) * Sends via the LinkedIn API **Instagram DM execution:** * Resolves the lead's Instagram account from cached profile data * Creates or retrieves the conversation thread * Sends via Instagram, with natural composition timing (profile dwell, thinking, typing, review) **Duplicate prevention:** The system detects crash/retry scenarios and prevents double-sending. *** ### Send Email **Platforms:** Email (Gmail and/or Outlook) Sends an email to the lead from one of your connected mailboxes. Subject and body support the same template variables and AI personalization as DMs. **Mailbox routing:** * Recipients on `@gmail.com` send from your Gmail account * Recipients on `@outlook.com` / `@hotmail.com` / `@live.com` send from your Outlook account * Everyone else sends from the **Primary** mailbox (you set this from the **Email** card on the parent account's detail page) **Execution:** * Validates the lead has an email address. If missing, the action is skipped. * Confirms the relevant mailbox is connected and active. If not, the action is skipped. * Skips leads with a recorded bounce on a previous send. * Renders the subject and body templates, applies AI personalization if enabled, then sends. **Reply handling:** Inbound replies are detected on the next mailbox poll and matched to the conversation. The reply appears in the unified Inbox. See [Email Channel](/outreach-and-sequences/email-channel) and [Connecting Email](/getting-started/connecting-email) for setup and configuration details. *** ### Connection Request **Platforms:** LinkedIn only Sends a LinkedIn connection request, optionally with a personalized note. **Daily limits:** You can set a daily connection request limit per account to any whole number from 1 to 100 (default 15, which is the recommended starting point). Choose a limit that matches your account type and risk tolerance. If the daily limit is reached, the action enters Deferred status and remaining connection request actions are deferred to the next day. **Execution:** * Checks daily limit * Simulates natural browsing behavior before sending * Personalizes the connection note if configured (supports `{{variable}}` placeholders; capped at 200 characters at send time, trimmed to the last whole word) * Sends the connection request * If auto-withdraw is configured, schedules a withdraw action after the specified number of days *** ### Withdraw Connection **Platforms:** LinkedIn only Withdraws a pending LinkedIn connection. This action is **automatically scheduled** by a connection request step's auto-withdraw setting - it is not available in the flow builder palette and cannot be manually added to sequences. Does **not** count toward sequence progress or advance the lead. *** ### Condition **Platforms:** X, LinkedIn, Instagram, Email (condition types vary by platform) Evaluates a condition and routes the lead to the TRUE or FALSE branch. **Condition types:** | Type | Platform | How It's Checked | | ------------------- | ----------------------------- | ----------------------------------------------------------------------------------------- | | Replied | X, LinkedIn, Instagram, Email | Checks if the lead has replied or sent inbound messages | | Followed Back | X, Instagram | Checks if the lead followed you back | | Connection Accepted | LinkedIn | Checks if the LinkedIn connection request was accepted; if so, cancels pending withdrawal | | Has Profile | X, LinkedIn, Instagram | Checks if lead has a filled-out profile on the target platform | | Has Email Address | Any platform, Email | Checks if the lead has a usable email on file | **Retry logic:** If the condition is not met and retries are configured (retry interval and max attempts): * A new condition action is scheduled for the same step after the retry interval * The retry attempt counter is incremented * Current action is marked as completed (not failed) * The FALSE branch only executes after all retries are exhausted ## Pre-Action Checks Before every action executes, the system runs these checks in order: | Check | Behavior if Failed | | ---------------------------------------------------------------------- | ------------------------- | | Action status is Pending or Queued | Skip | | Lead not in terminal status (Replied, Meeting Booked, Lost, Completed) | Skip | | Lead not marked as competitor | Skip | | Lead not blacklisted | Skip | | For X-only sequences: lead can receive DMs | Remove lead from sequence | | Daily action limit not exceeded | Reschedule for tomorrow | | Lead has required platform data for action type | Skip | Each sequence has a single unified daily action counter that covers all channels (X, LinkedIn, Instagram, email). If the limit is reached, the action is rescheduled for the next day at a random time within the activity window. ## Action Lifecycle Every action moves through these states: | Status | Description | | ----------- | ------------------------------------------------------------------------------------------------------ | | Pending | Waiting for scheduled time | | Queued | Scheduled time reached, in the processing queue | | In Progress | Being executed (includes human delay simulation) | | Completed | Successfully executed | | Skipped | Bypassed (no content, missing platform data, blacklisted, error) | | Failed | Could not complete (rarely used- most failures become skipped) | | Cancelled | Action cancelled before execution (e.g., sequence paused, lead removed, superseded by a manual action) | | Deferred | Rate-limited connection requests; auto-resumes the next day | After every completed or skipped action, the next step is queued based on the flow graph. ## Error Handling Errors are classified by severity: * **Minor errors** (no content found, duplicate action)- logged, sequence continues normally * **Critical errors** (rate limit, bot detection, auth failure)- may pause the account and notify you by email * **Budget exhaustion**- action rescheduled for the next day, preserving the sequence flow ## Human Delay Simulation All actions include natural timing that simulates browsing, reading, and typing behavior. Delays vary by action type and context to create realistic patterns. ## Next Steps * [**Building Sequences**](/outreach-and-sequences/building-sequences): Combine actions into multi-step campaigns * [**DM Personalization**](/outreach-and-sequences/dm-personalization): How AI generates personalized first messages * [**Scheduling & Send Limits**](/outreach-and-sequences/scheduling): Activity windows and rate limit details # DM Personalization When a DM or Email step executes in a sequence, AutoReach generates a personalized message for each lead. It takes your template with `{{variable}}` placeholders, enriches it with lead data, and uses AI to produce a natural, personalized message. The variable system documented here applies to **both DMs (X / LinkedIn) and Email steps** (subject and body). Email-specific routing details are covered in [Email Channel](/outreach-and-sequences/email-channel). ## When You Need This Page You don't have to read this whole page to get started. Autopilot generates a working DM template for you, and the defaults are fine for the first run. Come back here when: * **Your reply rate is low** and you want to add more relevant references to your messages (a recent post, a specific role, a pain point). * **You're writing your own template from scratch** and need to know which `{{variables}}` are available. * **Messages are coming out generic** because you're missing the variables that drive personalization (typically `{{post}}` for X, `{{current_role}}` + `{{company_name}}` for LinkedIn, or a custom `{{pain_point}}` placeholder the AI fills from your offer). * **A variable isn't substituting** and you need to know whether it's a known variable (direct fill from lead data) or a custom one (AI-inferred from context). If your first sequence is sending and getting replies, you can skip this page entirely. ## Two modes: prompt-only and template There are two ways a DM or email step can generate its message: * **Prompt-only (default, recommended):** leave the message template empty. The AI writes the whole message from your DM generation prompt, the lead's data, and offer context. There are no `{{placeholders}}` to fill; the prompt owns the rules. * **Template (legacy slot-filling):** provide a message template with `{{variable}}` placeholders. The AI only fills the placeholders, and the prompt becomes style guidance. This path exists for older sequences and for cases where you want strict control over wording. The variable reference below applies to template mode. In prompt-only mode the same lead data is available to the AI, just without explicit placeholders. ## Personalization Flow (template mode) When a template is present, message personalization follows three passes: 1. **Direct substitution** - replaces known variables with data from the lead's profile 2. **Post fetch** - fetches `{{post}}`/`{{tweet}}` content if referenced in the template 3. **AI personalization** - fills remaining placeholders using lead profile, enrichment data, offer context, knowledge base content, and tone examples Messages with unresolved variables are blocked and never sent. *** ## Known Variables (Direct Substitution) These variables are replaced directly from lead data in the first pass. ### Identity | Variable | Source | | ---------------- | --------------------------------- | | `{{name}}` | Full name, falls back to username | | `{{first_name}}` | Extracted from full name | | `{{username}}` | Platform handle | | `{{bio}}` | Lead bio | | `{{location}}` | Lead location | | `{{email}}` | Lead email | ### Network | Variable | Source | | --------------- | --------------- | | `{{followers}}` | Follower count | | `{{following}}` | Following count | ### Profile Links | Variable | Source | | ------------------- | -------------------- | | `{{website}}` | Lead website URL | | `{{linkedin_url}}` | LinkedIn profile URL | | `{{x_profile_url}}` | X profile URL | ### Professional Profile | Variable | Source | | ------------------ | ------------------------------------------ | | `{{headline}}` | LinkedIn headline, falls back to bio | | `{{summary}}` | Profile summary | | `{{current_role}}` | Current job title from LinkedIn experience | | `{{skills}}` | Top skills, comma-separated | ### Company Data | Variable | Source | | ---------------------- | ------------------------------------------- | | `{{company_name}}` | Company name from enrichment or profile | | `{{company_size}}` | Employee count range (e.g., "51-200") | | `{{company_industry}}` | Company industry from enrichment or profile | | `{{funding_stage}}` | Company funding stage | | `{{tech_stack}}` | Top technologies, comma-separated | ### Web Enrichment Insights | Variable | Source | | ------------------------ | -------------------------------------------------- | | `{{achievements}}` | Notable achievements (filtered to current company) | | `{{speaking}}` | Speaking engagements | | `{{podcasts}}` | Podcast appearances | | `{{enrichment_summary}}` | Web enrichment summary | Achievements, speaking engagements, and podcasts are filtered to the lead's current company. Data from previous employers is excluded. ### Sender and Booking | Variable | Source | | ------------------ | ---------------------------------------- | | `{{user_name}}` | Your name (sender) | | `{{booking_link}}` | Calendar URL with lead-specific tracking | *** ## Special Variables (Fetched at Send-Time) These variables trigger data fetching when used in a template. ### `{{post}}` / `{{tweet}}` Both map to the same logic (`{{tweet}}` is a backwards-compatible alias). Fetches the lead's recent post content using this priority: 1. **Stored content** - if the lead was sourced from a search with original post content 2. **Live fetch** - fetches the latest post from the lead's profile 3. **Bio fallback** - AI uses bio for personalization instead For commenter-sourced leads, the context includes both the original post and the lead's reply. ### `{{replied_post}}` References a post that was commented on in a **prior sequence step**. If a comment action executed earlier in the sequence, this variable provides the original post text and your comment text so the DM can reference prior engagement. If no prior comment happened (step was skipped), the AI uses a generic opener instead. *** ## Custom Variables (AI-Inferred) Any `{{placeholder}}` that is not in the known variables list is treated as a custom variable. The AI fills it using explicit information from the lead's profile, enrichment data, and offer context. If the information is not available, the placeholder and surrounding text are removed. The AI never invents or guesses information. Common AI-inferred variables: | Variable | What AI looks for | | ---------------- | -------------------------------------------------- | | `{{role}}` | Job title from bio, headline, or enrichment | | `{{company}}` | Company name from bio or enrichment | | `{{industry}}` | Industry from bio or enrichment | | `{{pain_point}}` | Relevant challenge based on offer and lead context | | `{{question}}` | Contextual question based on lead and offer | You can use any custom placeholder name. *** ## Example Template ``` hey {{first_name}}, saw you're {{current_role}} at {{company_name}} which probably means {{pain_point}} I help with {{tech_stack}} teams - think {{enrichment_summary}}. quick 15-min review, no cost. interested? {{user_name}} {{booking_link}} ``` **Pass 1** replaces `{{first_name}}`, `{{current_role}}`, `{{company_name}}`, `{{tech_stack}}`, `{{enrichment_summary}}`, `{{user_name}}`, `{{booking_link}}` with lead data. **Pass 2** sends `{{pain_point}}` to AI, which infers it from the lead's profile and offer context or removes it with surrounding text. *** ## Booking Link Tracking The `{{booking_link}}` variable injects your calendar URL with a lead-specific tracking parameter so a resulting booking can be matched back to the lead. The identifier is prefixed by platform (`li:` for LinkedIn, `ig:` for Instagram, bare for X): | Provider | Parameter | Notes | | -------- | ------------- | ----------------------------------------------------------------- | | Calendly | `utm_content` | Captured invisibly and returned in the booking webhook | | Cal.com | `username` | Prefills (and can hide) the username booking question | | Custom | none | Custom calendar links are passed through without a tracking param | *** ## DM Quality Standards AutoReach enforces quality standards on generated messages to ensure they read naturally and avoid common spam patterns: * **Concise and direct** - short messages with simple language * **Ends with a question** - invites a natural reply * **No generic openers** - avoids overused phrases * **Specific personalization** - references actual lead data (company, role, recent post) rather than generic compliments * **No hype language** - avoids buzzwords and sales jargon ## Custom Prompts Sequences have two separate prompt fields for DM control: | Field | Purpose | | -------------------- | ------------------------------------------------------------ | | DM Generation Prompt | Overrides the system prompt for message personalization only | | AI Prompt | Controls AI conversation replies across all stages | The DM Generation Prompt is optional. If not provided, the default system prompt is used. It allows DM-specific instructions separate from general conversation AI behavior. ## Template Generation AutoReach can generate DM templates for you using AI. Provide a campaign name, tone preference (professional/casual/friendly), and length, and the AI creates a template with `{{variable}}` placeholders ready for personalization. ## Negative Content Checking Before replying to a lead's post, the system checks for sensitive content. If the check cannot be completed, the message proceeds rather than blocking. **Skipped topics:** Death, obituaries, serious illness, natural disasters, violence, terrorism, suicide, personal tragedies. **Allowed topics:** Work complaints, business failures, layoffs, market downturns, sarcasm, controversial opinions, career setbacks. ## Multi-Language Support Message generation can run in any language you set on the offer. The language picker covers the full ISO 639-1 set (English, Italian, Spanish, French, German, Portuguese, and well over a hundred others). For non-English offers, all AI output (keywords, messages, templates) is generated in the selected language. ## Platform Behavior All known variables work identically on both X and LinkedIn. Platform differences affect the AI's tone and context, not variable availability: | Aspect | X | LinkedIn | | ---------------- | ----------------- | -------------------------- | | Username prefix | `@username` | `username` (no prefix) | | Tone instruction | Casual and direct | Slightly more professional | | `{{post}}` label | "on X" | "on LinkedIn" | ## Next Steps * [**Simulation and A/B Testing**](/outreach-and-sequences/simulation-and-testing): Preview DMs with real leads before sending * [**Building Sequences**](/outreach-and-sequences/building-sequences): Integrate DM steps into multi-step campaigns * [**Supported Actions**](/outreach-and-sequences/supported-actions): When to use DM vs. other engagement types # Email Channel Email is a first-class outreach channel in AutoReach alongside X and LinkedIn. You can add email steps to sequences, personalize subject and body with the same template variables and AI personalization used for DMs, and handle replies in the same unified Inbox. > **Setup required:** Connect Gmail and/or Outlook before email steps are available. See [Connecting Email](/getting-started/connecting-email). ## When to Use Email vs. DMs Email isn't a replacement for X or LinkedIn DMs, it's a complement. Each channel has different strengths: | Channel | Best for | Requires | | --------------- | -------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------- | | **LinkedIn DM** | Professional credibility, decision-makers in B2B | Connection accepted | | **X DM** | Faster, more casual, high-volume engagement-first outreach | Lead has DMs open | | **Email** | Long-form messages, follow-ups outside the platform, leads who aren't active on social | A verified email address (run [Email Finding](/enrichment/optional-enrichment) first) | The most effective sequences often combine channels rather than picking one. A common pattern: open with a LinkedIn engagement step (like + follow) → connection request → if accepted, send a LinkedIn DM → if no reply after several days, follow up via email. You don't need email to use AutoReach. Many users run LinkedIn-only or X-only outreach successfully. Add email when you want a third channel for follow-ups or to reach leads who don't respond on social. ## Email as a Sequence Step When building a sequence, **Email** appears as an action type when at least one email mailbox is connected. ### Configurable Fields | Field | Description | | -------------------- | ---------------------------------------------------------------- | | Subject | Subject line template (supports `{{variable}}` placeholders) | | Body | Body template (supports `{{variable}}` placeholders, multi-line) | | AI Personalization | Toggle to have AI rewrite the body per lead using their context | | Delay days / minutes | Wait time before this step executes | ### Personalization Email steps use the **same template variable system** as DMs. All identity, network, profile, company, web enrichment, and special variables described in [DM Personalization](/outreach-and-sequences/dm-personalization) work identically in email subject and body. When **AI Personalization** is enabled, AutoReach: 1. Replaces known variables with lead data (`{{first_name}}`, `{{company_name}}`, etc.) 2. Sends the result plus the lead's profile and offer context to the AI 3. AI rewrites the email so it reads naturally and references specifics from the lead's profile When AI Personalization is **off**, only direct variable substitution runs. Unknown placeholders are left literal. ### Booking Link The `{{booking_link}}` variable works in email subject and body just like in DMs, with lead-specific tracking parameters appended to your calendar URL. *** ## Mailbox Routing Each email step is sent from one of your connected mailboxes. The routing logic is automatic: * `@gmail.com` and `@googlemail.com` recipients send from your Gmail account * `@outlook.com`, `@hotmail.com`, `@live.com`, `@msn.com` (and regional variants) send from your Outlook account * For other domains, AutoReach also checks the domain's MX records and routes Google Workspace and Microsoft 365 hosted domains accordingly * Domains it cannot classify use your **Primary** mailbox You set the primary mailbox from the **Email** card on the parent X or LinkedIn account's detail page by clicking **Make Primary** on the mailbox you want as default. *** ## Reply Detection Connected mailboxes are polled every 15 minutes for new inbound messages. When a reply arrives: * AutoReach matches it to the original outbound thread (by thread ID or `In-Reply-To` header) * The lead's status is updated to **Replied** * The full thread appears in the **Inbox** alongside X and LinkedIn conversations * If AI auto-replies are enabled on the sequence, an AI response is queued Email replies count toward your sequence's reply rate the same way DM replies do. *** ## Bounce Detection Bounces are detected automatically on inbound traffic: * Sender patterns (mailer-daemon, postmaster) * Standardized bounce report formats * Subject line indicators When a bounce is detected: 1. The lead is marked with a bounced status 2. **All future email sends to that lead are skipped** automatically 3. The conversation card in the Inbox shows a bounce indicator This prevents repeat sends to dead addresses and protects your sender reputation. *** ## Daily Limits Email sends count toward the same per-sequence unified daily action limit that covers DMs, likes, comments, follows, and connection requests (default 20). When the limit is reached, remaining email sends are deferred to the next day along with any other queued actions. See [Scheduling & Send Limits](/outreach-and-sequences/scheduling) for details. In addition to the per-sequence limit, each mailbox is rate-paced at the provider level to keep delivery clean: * **One in-flight send per mailbox** at a time (sends are serialized, not blasted in parallel) * **Up to 10 sends per minute per mailbox** (token bucket) Sends are spread across the sending account's activity window (per-account) with natural gaps to avoid spam patterns. *** ## Inbound Auto-Replies When a lead replies by email: * An AI response is queued into the same auto-response system used for DMs * Replies are sent with natural timing (delayed to feel human) * The same **Max AI responses per conversation** limit applies * The same per-conversation **AI ON / AI OFF** toggle in the Inbox controls whether AI replies for that thread *** ## Best Practices 1. **Start the sequence with engagement, not email.** A like or comment on LinkedIn before the email gives the lead context for who you are. 2. **Use AI personalization for the first email.** A first cold email referencing the lead's specific role and challenge outperforms a templated blast. 3. **Keep subject lines short and specific.** Avoid clickbait. Reference something concrete. 4. **Set a follow-up step.** Use a wait + email step to follow up if the first email goes unanswered. 5. **Check the Inbox.** Email replies often come hours later than DM replies. Glance at the Inbox each morning so you don't miss warm threads. 6. **Watch your bounce rate.** A spike in bounces usually means lead data quality is low. Run [website finding](/enrichment/optional-enrichment) before [email finding](/enrichment/optional-enrichment) to improve email accuracy. *** ## Email Channel vs Email Finding Two related but separate features: | Feature | Purpose | | ----------------------------- | --------------------------------------------------------------------------- | | **Email Channel** (this page) | Sends and receives email through your connected mailboxes | | **Email Finding** | Discovers email addresses for leads using a third-party service (Findymail) | Email Finding prepares the data. The Email Channel delivers messages. See [Email and Website Finding](/enrichment/optional-enrichment) for the discovery side. *** ## Next Steps * [**Connecting Email**](/getting-started/connecting-email): Set up Gmail or Outlook * [**DM Personalization**](/outreach-and-sequences/dm-personalization): Variables and AI personalization that also apply to email * [**Inbox**](/ai-and-conversations/inbox): Where email replies show up * [**Email Finding**](/enrichment/optional-enrichment): Find email addresses for leads who don't have one yet # Scheduling & Send Limits AutoReach schedules sequence actions to look like natural human behavior- spacing actions across the day, clustering them into sessions with breaks, and enforcing daily limits to keep accounts healthy. ## Activity Window The activity window defines when outreach actions execute. Actions scheduled outside the window are deferred until the window opens. Activity windows are **per-account**, not per-user. Each connected account (LinkedIn, X, Instagram) has its own schedule and timezone. A window is one or more time blocks per day (for example a morning block plus an afternoon block), capped at 8 hours of total active time per day. This lets you run accounts on independent schedules: for example, a US LinkedIn account on EST business hours and an EU X account on CET business hours. Configure an account's window from the **Accounts** page → open the account → **Configuration** tab → **Activity window** card. | Setting | Default | | --------------- | -------------------------------------------------------------- | | Schedule | 09:00-13:00 + 14:00-18:00 (8h), Monday to Friday, weekends off | | Timezone | Account-configured | | Max active time | 8 hours per day (summed across all blocks) | A window can be a single block or several blocks per day, as long as the total stays within the 8-hour daily cap. If an account has not been customized, AutoReach falls back to the user-level window for that user. Once you save a per-account window, that account uses its own schedule independently. Actions scheduled outside the window are automatically deferred until the window opens. ## Action Scheduling ### How Actions Are Scheduled The scheduler creates action records for each lead and processes them when their scheduled time arrives. ### Lead Priority Higher-priority leads (based on buyer score and signal recency) get earlier time slots within the day. ### Timing Distribution Actions are distributed across the activity window to simulate natural human behavior: * **Natural spacing**- actions are spread throughout the day with minimum gaps between them * **Time-of-day variation**- higher activity during peak hours, lower during early morning and evening * **Session clustering**- actions are grouped into human-like sessions with breaks between them * **Day-of-week variation**- weekday and weekend activity levels differ naturally ## Daily Send Limits Daily limits are **per-sequence** (not per-account). Each sequence has its own counters. ### Action Limits Each sequence has a single unified daily action limit covering **all** outreach actions: likes, comments, follows, DMs, connection requests, and emails. Condition steps do not count. In addition to the per-sequence daily limit, email sends are paced per mailbox at one in-flight send at a time and up to 10 sends per minute to keep delivery clean. LinkedIn connection requests are tracked separately per-account. See [Daily Connection Limits](#daily-connection-limits-linkedin) below. ### Enforcement Limits reset at **midnight in the acting account's timezone**. If a limit is reached, remaining actions are deferred to the next day. ## Daily Connection Limits (LinkedIn) LinkedIn connection requests have a separate daily limit tracked per-account (not per-sequence). When the limit is reached, pending connection requests are deferred to the next day. Other action types continue normally. Set the daily connection limit on each LinkedIn account to match your account type. See [Supported Actions](/outreach-and-sequences/supported-actions#connection-request) for the limits by account type. ## Natural Timing All action execution includes natural pacing that simulates human behavior - browsing, reading, typing, and thinking pauses. Delays vary by action type, time of day, and day of week to create realistic patterns unique to each account. Occasional random pauses are inserted between actions to simulate natural breaks and distractions. ## Gradual Resumption When a paused sequence is resumed or actions become overdue, they are spread across time instead of firing all at once. This keeps activity patterns looking natural. ## Max AI Responses Per Conversation Limits how many times AI auto-replies in a single conversation: | Setting | Default | Maximum | | --------------------------------- | ------------- | ------- | | Max AI responses per conversation | 0 (unlimited) | 100 | When set to 0, AI auto-replies are unlimited- the AI will continue responding as long as the conversation is active. When set to a positive number, AI responds up to that many times per conversation, then stops. ## Next Steps * [**Building Sequences**](/outreach-and-sequences/building-sequences): Configure limits and timing per sequence * [**Supported Actions**](/outreach-and-sequences/supported-actions): Action-specific rate limits and behaviors * [**Outreach Overview**](/outreach-and-sequences/overview): How sequences work end-to-end # Simulation & A/B Testing Test your outreach sequences in a safe environment before going live. The Conversation Simulator lets you run realistic multi-turn conversations with AI-powered lead personas, debug every aspect of AI decision-making, and score your messaging quality, all without sending anything to real people. ## Getting Started Navigate to **Simulation** in the sidebar. The simulator opens with a three-step setup wizard. ### Step 1: Select Sequence Choose which sequence to simulate. Active, paused, and draft sequences are all available. The simulator loads the sequence's DM template, AI prompt, and all configuration. ### Step 2: Select Lead Two options for choosing who to simulate with: **All Leads tab** - Search your existing leads by name, username, company, or email. Filter by buyer state (Ready, Emerging, Low Priority, Not Scored). The simulator loads their full profile, enrichment data, and recent activity. **Custom Lead tab** - Create a synthetic lead on the fly. Only a name is required. Optionally add company, role, and bio. Useful for testing edge cases or specific personas without needing a real lead in your pipeline. > **Tip:** Simulate with 3-5 diverse leads across different roles, industries, and company sizes. Include at least one lead with minimal data to ensure messages still read well when enrichment is sparse. ### Step 3: Choose Mode Three simulation modes are available: | Mode | Description | | ------------- | -------------------------------------------------------------------------------------- | | **Manual** | You play the lead - type responses and see how AI replies in real time | | **Automatic** | AI plays both sides - watch a full conversation unfold with configurable lead personas | | **Batch** | Run multiple automatic simulations at once with different persona combinations | **Persona presets** - In Automatic and Batch modes, select from built-in lead personalities: | Persona | Behavior | | ---------------------- | --------------------------------------------- | | Hard to convince | Skeptical, needs proof, pushes back | | Easy & enthusiastic | Genuinely interested, moves fast | | Asks lots of questions | Curious, wants to understand everything | | Not a fit | Misunderstood the DM, looking for work | | Price sensitive | Budget-conscious, needs ROI justification | | Uses a competitor | Already has a solution, wants differentiation | | Skeptical | Suspects spam or bot, cautious | | Terse responder | Replies in 1-5 words | You can also write a **custom persona** description for specific scenarios (e.g., "A VP of Engineering at a 500-person company who just switched from a competitor last month"). **Language support** - If your offer targets a non-English audience, lead personas automatically respond in the offer's configured language, so you can test how conversations flow in the target language. ## Manual Mode In Manual mode, the AI sends the opening DM based on your sequence template. You then type responses as the lead and watch how the AI handles each reply. Use this to stress-test specific scenarios: * **Positive reply** - "This looks great!" - does AI offer clear next steps? * **Objection** - "Too expensive." - does AI handle it gracefully? * **Question** - "How is this different from X?" - does AI provide a credible answer? * **Skepticism** - "Is this automated?" - does AI respond naturally? ## Automatic Mode In Automatic mode, the AI plays both sides of the conversation. The selected persona determines how the simulated lead behaves. You see the conversation unfold turn by turn with a live turn counter. Controls during an automatic simulation: * **Stop** - End the simulation early After the conversation ends, you can continue it: * **Lead replies** - Add another lead message to extend the conversation * **AI sends follow-up** - See how the AI follows up after silence The conversation ends naturally when one of these occurs: | Exit Reason | Description | | ---------------- | -------------------------------------- | | No Reply | The lead persona stops responding | | End Conversation | The lead explicitly ends the chat | | Exit | The AI reaches a natural conclusion | | Max Turns | The conversation hit the turn limit | | Error | Something went wrong during generation | ## Batch Mode Run 2-25 simulations at once, each with a different combination of persona presets. Batch mode shows all conversations in a live grid with real-time progress. Each run shows: * Persona labels * Live conversation turns * Exit reason * Turn count * Scorecard scores After all runs complete, you get a **batch summary** with: * Average scores across all runs * Goal completion rate * Breakdown by exit reason * **Weak spot analysis** - identifies which conversation stage has the highest drop-off rate, showing you exactly where your messaging loses leads ## Scorecard Every completed simulation (automatic or batch) generates a scorecard that grades your AI's performance: | Metric | Scale | What It Measures | | ------------------ | ----- | ----------------------------------------------------- | | Objection Handling | 1-10 | How well the AI addressed concerns and pushback | | Tone Consistency | 1-10 | Whether the AI maintained a natural, consistent voice | | Relevance | 1-10 | Whether responses addressed the lead's actual points | | Naturalness | 1-10 | How real vs. scripted the conversation felt | The scorecard also reports: * **Goal completed** - whether the AI reached a soft close or meeting booking * **Turns to close** - how many turns it took to reach the goal * **Repetition detected** - whether the AI repeated itself * **Stage progression** - which conversation stages were reached ## Debug Panel Click any message in a simulation to inspect the AI's decision-making. The debug panel has four tabs: **Stage and Signals** - Shows the classified conversation stage, confidence level, stage reasoning, detected buying signals, and objection subtype. **System Prompt** - Reveals the full prompt sent to the AI, broken into layers: base instructions, conversation context, lead data, knowledge base content, and tone guidance. **RAG Context** - Shows which knowledge base sections and tone examples were retrieved for this response. **Lead Analysis** - Displays the lead's profile data, communication style, response temperature, and detected intent signals. *** ## A/B Testing The simulator includes a built-in **A/B Compare** mode that runs the same scenario with two different AI prompts side by side. This lets you see exactly how prompt changes affect conversation quality before applying them to live sequences. ### How A/B Compare Works 1. Open the simulator and select a sequence, lead, and **Manual** or **Automatic** mode 2. Toggle **A/B Compare** on in the toolbar 3. The simulator creates two conversation slots: **Prompt A** (your current AI prompt) and **Prompt B** (an editable copy) 4. Edit Prompt B with your changes 5. Both conversations run simultaneously against the same lead and persona 6. Each slot gets its own scorecard so you can compare results directly Both conversations use the same lead data, the same persona presets, and the same DM template. Only the AI prompt differs. This isolates the impact of your prompt changes. ### What to Test **Tone shifts** - Compare professional vs. casual prompts and check which produces better scorecard results. **Objection handling** - Test different approaches to pushback (e.g., case studies vs. ROI reframing). Use the "Price Sensitive" or "Hard to Convince" persona to trigger objections. **Conversation strategy** - Compare discovery-first approaches ("Ask questions before pitching") vs. value-led approaches ("Lead with an insight and suggest a call"). **Length and detail** - Test a detailed prompt with stage-specific instructions vs. a minimal prompt that gives the AI more freedom. ### Tips for Effective A/B Testing 1. **Change one thing at a time.** If you change tone, strategy, and length all at once, you cannot tell which change made the difference. 2. **Run multiple comparisons.** Test with at least 3-5 different leads and personas. 3. **Use challenging personas.** Easy leads do not differentiate prompts well. Use "Hard to Convince", "Skeptical", or "Uses Competitor" to stress-test. 4. **Check the debug panel.** Click individual messages to see how the AI interpreted the stage, what knowledge base content it pulled, and what signals it detected. 5. **Follow up with batch mode.** Once you have a winning prompt, run a batch of 10-25 simulations across all personas to validate it at scale. 6. **Apply and iterate.** When you find a better prompt, update your sequence settings and test again. *** ## Conversation Analysis Click **Analyze** in the toolbar to run an AI-powered review of the entire conversation. The analyzer examines the exchange and generates actionable suggestions for improving your prompts, tone examples, and messaging strategy. Suggestions can include modifications to your AI prompt, new tone examples to add, existing tone examples to adjust, and other improvements. You can preview exactly what each suggestion would change before applying it. ## Capturing Tone Examples When the AI produces a particularly good response during simulation, you can save it as a tone example. Click the bookmark icon on any AI message to capture that exchange as a reusable reference for future conversations. ## Toolbar The simulation toolbar provides quick access to: * **New** - Start a fresh simulation * **Restart** - Re-run the current simulation from scratch * **Analyze** - Run conversation analysis * **Debug** - Toggle the debug panel * **Settings** - Edit the DM template and AI prompt (changes are for testing only) * **Lead Profile** - View the lead's full profile * **A/B** - Toggle A/B prompt comparison ## Common Issues **AI responses feel generic:** Add more context to your knowledge base. Upload documents about your product, case studies, and competitive positioning. Add tone examples that demonstrate your preferred voice. **Simulation does not feel representative:** Use real leads from your pipeline rather than synthetic ones. Try different persona presets to cover a range of buyer behaviors. Run a batch of 10+ simulations to see patterns. **Template variables showing as blank:** Check that the lead has the relevant data in their profile. Try a different lead with more complete enrichment. See [DM Personalization](/outreach-and-sequences/dm-personalization) for the full list of available variables. ## Next Steps * [**DM Personalization and Template Variables**](/outreach-and-sequences/dm-personalization): All available personalization variables * [**Building Sequences**](/outreach-and-sequences/building-sequences): Apply what you learned to real campaigns * [**Tone and Knowledge Base**](/ai-and-conversations/tone-and-knowledge): Train the AI's voice with real conversation samples # Inbox & Real-Time Messaging The Inbox is your unified view of all conversations across X, LinkedIn, Instagram, and email. Every DM, email, and reply flows through here, with real-time message detection and AI status indicators. You always have full control to step in, override AI responses, and handle conversations that need a human touch. ## Conversation List Your Inbox displays all conversations in a single, paginated list. Each conversation shows: * Lead name and profile * Timestamp of most recent activity * AI status indicator (bot icon overlay when AI is disabled) * Platform icon and status badges ### Starring Conversations Click the star icon on any conversation to keep it easy to find. Starred conversations work as a personal shortlist for high-priority leads or deals in progress. ### Unread Indicators New inbound messages are tracked automatically. The Inbox shows an unread count that updates periodically as new messages arrive. ### Filtering Filter conversations by: * **Status** (All Statuses, Replied, Meeting Booked) * **Sequence** (specific sequence) * **Search query** (lead name) Combine filters to narrow your view. For example, show only "Replied" conversations to focus on warm leads. ## Message Detection New messages on X are detected via Twitter's DM connection. LinkedIn and Instagram messages are detected through periodic polling, which adapts its cadence to how active a conversation is. Email replies are detected by polling your connected Gmail or Outlook mailboxes; bounces are flagged automatically and future sends to bounced addresses are skipped. > **Note:** End-to-end encrypted Instagram DM threads cannot be read by AutoReach and are skipped. Reply detection and AI responses work on standard Instagram threads. ## AI Status Indicators * **AI ON toggle**: Visible at the top of each conversation, indicates whether AI can auto-reply * **Bot icon**: A red Bot icon overlay appears on conversations where AI is disabled ## Sending Manual Messages Send a message directly from the Inbox without AI involvement: 1. Open a conversation 2. Type your message 3. Click Send Manual messages are: * **Sent instantly** with no scheduling delay * **Not subject to daily limits**, so manual DMs do not count against your action limits * **Auto-cancel pending AI responses**. If the AI had a response queued, it is immediately cancelled when you send a manual message ## AI ON/OFF Toggle At the top of any conversation, toggle AI responses: * **AI ON**: AI generates and sends responses automatically based on your prompts and tone examples * **AI OFF**: Only your manual messages are sent The toggle applies to that single conversation only. You can have AI enabled for most conversations while handling your hottest prospects manually. ## Call Briefs Click the **Call Brief** button in any conversation to generate a pre-call preparation document. The brief pulls data from the lead's profile, conversation history, and buyer intelligence to give you talking points before a meeting. See [**Meetings, Booking, and Call Briefs**](/meetings-and-crm/meetings) for details. ## Next Steps * [**AI Response Engine**](/ai-and-conversations/ai-response-engine): How AI generates responses * [**Tone and Knowledge Base**](/ai-and-conversations/tone-and-knowledge): Customize the AI's voice and provide sales context # AI Response Engine The AI Response Engine generates contextual, stage-aware replies to incoming messages across X, LinkedIn, Instagram, and email. ## When AI Should Reply vs. When You Should The AI is built to handle the routine middle of a conversation, not the moments that make or break a deal. A good rule of thumb: **Let the AI handle:** * The first 2-3 replies after someone responds to your cold DM (acknowledging, asking discovery questions, handling common objections). * Mid-funnel leads who went silent (the AI's automatic follow-ups have a fresh angle each time). * High-volume, low-stakes back-and-forth where speed of response matters more than perfect wording. **Take over manually:** * A lead who asks detailed product questions or shows strong buying intent. These conversations deserve your attention. * Conversations where the lead pushes back on something nuanced (pricing negotiation, custom terms, technical depth the AI doesn't have). * Anything where getting the next sentence wrong loses the deal. Toggle AI on/off per conversation from the Inbox. AutoReach will keep tracking the conversation either way: turning AI off just stops it from sending replies on your behalf. > **Tip:** You don't have to choose all-or-nothing. Let AI handle openers and discovery, then take over once a lead is clearly qualified and ready to talk specifics. ## Response Flow When a new inbound message is detected, AutoReach: 1. **Classifies the conversation stage** (Opener Reply, Discovery, Objection Handling, etc.) 2. **Retrieves relevant context** from your knowledge base and tone examples 3. **Generates a response** tailored to the stage, lead context, and your voice 4. **Sends with natural timing** to avoid instant robotic replies ## Conversation Stages AutoReach classifies conversations into eight stages. Each stage has its own tone and goals. The AI detects the current stage and adapts its responses accordingly. ### 1. Opener Reply **Triggered**: Lead just responded to your cold DM. **Goal**: Acknowledge their message, show genuine interest, keep it light. No pitching. ### 2. Discovery **Triggered**: You are learning about their situation. **Goal**: Ask targeted questions about their pain points, timeline, and constraints. ### 3. Value Prop **Triggered**: Lead is asking about your solution or you are introducing it. **Goal**: Connect your value directly to what the lead mentioned, using relevant proof points from your knowledge base. ### 4. Objection Handling **Triggered**: Lead raises a concern, budget question, or pushback. **Goal**: Validate the concern first, then address it with a specific counter-point and clear next step. ### 5. Soft Close **Triggered**: Lead seems ready to move forward or book a call. **Goal**: Ask if they are open to a call first. Only share a booking link after they agree - never send it unsolicited. ### 6. Follow Up **Triggered**: Lead went silent for a configurable period. **Goal**: Re-engage with a fresh angle or new information that gives the lead a reason to respond. ### 7. Offer Pivot **Triggered**: The lead closed the specific hook the conversation opened with, but the underlying offer could still be relevant (they solved the one-off version of a recurring problem your offer addresses). **Goal**: One low-pressure attempt to name the recurring version of the pain and frame the offer as the structural fix. If the offer is not clearly relevant, the AI closes gracefully instead. This stage can be turned off per sequence. ### 8. Graceful Exit **Triggered**: Lead declines or the conversation has reached a dead end. **Goal**: Exit respectfully and leave the door open for the future. ### Stage Detection The AI classifies the current stage based on the full conversation history. Conversations can move between stages non-linearly. For example, from Soft Close back to Objection Handling if the lead raises new concerns. The classifier re-evaluates on every new message. ## Prompt Priority Your custom AI prompt (set in sequence settings) takes the **highest priority** in response generation. It overrides stage guidance and default behavior, including word limits. Use it to enforce specific rules, tone, or messaging strategy that the AI must always follow. Stage guidance (the conversation stages above) acts as secondary tactical guidance - it shapes the AI's approach but yields to your custom rules. ## Anti-Fabrication The AI never fabricates client stories, case studies, statistics, or results. It only references proof points that exist in your knowledge base. If the AI does not have a relevant example, it speaks in general terms rather than inventing specifics. ## Anti-Consulting Guardrails The AI is designed to avoid giving away too much value before qualifying a prospect. It redirects detailed technical questions toward a call or meeting rather than answering everything in a DM. This keeps conversations moving toward a booking rather than becoming free consulting sessions. ## Max AI Responses Per Conversation The "Max AI responses per conversation" setting on each sequence limits how many AI replies are sent in a single conversation: * Value 0 = unlimited responses * When the count is reached, AI is disabled on that conversation * Also checked before queuing follow-up messages ## Conversation Follow-Ups When a lead goes silent, the follow-up scheduler can automatically re-engage: | Setting | Default | Range | | ----------------- | ------- | ----- | | Follow-up enabled | false | - | | Wait days | 3 | 1-30 | | Max follow-ups | 2 | 1-10 | When a lead has not responded within the configured wait period, AutoReach first screens the conversation with the AI to decide whether re-engaging actually makes sense (it skips threads that already concluded naturally, for example a goodbye or an acknowledged booking link). If a follow-up is warranted, a message is generated with a fresh angle and sent automatically. Follow-ups respect: * The max follow-up count per conversation * The max AI responses per conversation limit * The activity window of the account that owns the conversation (per-account) ## Next Steps * [**Tone and Knowledge Base**](/ai-and-conversations/tone-and-knowledge): Customize the AI's voice and provide sales context * [**Conversation Analyzer**](/ai-and-conversations/conversation-analyzer): AI suggestions for improving your sequences # Tone & Knowledge Base Tone examples and the knowledge base are the two tools you use to control how the AI communicates on your behalf. Tone examples teach the AI *how* to sound. The knowledge base teaches the AI *what* to say. ## Which Should You Set Up First? Both matter eventually, but they solve different problems: * **Start with the Knowledge Base** if your offer has nuanced positioning, common objections, or details the AI needs to get right (pricing tiers, specific case studies, technical answers). Without it, the AI improvises, and improvisation on cold facts is risky. Even a single well-written 1-page positioning doc dramatically improves response quality. * **Start with Tone Examples** if your AI replies sound technically correct but feel robotic or off-brand. Tone examples are the fastest way to make the AI sound like you. Most users get the most value from a small Knowledge Base (2-3 documents) plus 5-10 tone examples. AutoReach also auto-captures tone examples from conversations that lead to booked meetings, so this library grows over time without manual work. If you're brand new and not sure where to start: skip both, let your sequence run for a week, then come back here. AutoReach will have auto-captured your first tone examples by then, and you'll know which knowledge gaps the AI keeps stumbling on. *** ## Tone Examples Tone examples are conversation samples that teach the AI how you want to sound. They are the most effective way to make AI responses feel authentic to your voice and sales style. ### How They Work Tone examples are stored **per-sequence**. Each example contains: * A **label** describing the scenario * A **stage** classification (Opener Reply, Discovery, Objection Handling, etc.) * A **conversation** with message pairs showing the exchange * A **source type** indicating how it was created When generating a response, the AI retrieves the most relevant tone examples based on the current conversation context. Examples cover all conversation stages including a dedicated "Not a Fit" category for gracefully handling leads that are not a match. ### Auto-Capture from Conversations Tone examples are automatically captured from real conversation outcomes: **Winning Conversations** - When a conversation reaches **Meeting Booked** status, the system extracts the best exchanges and saves them as examples, automatically classified by conversation stage. Duplicate examples are detected and skipped. ### Tone Summary A **tone summary** is a style guide automatically extracted from your tone examples by AI. It captures your vocabulary, communication style, and common patterns. You can regenerate it from the tone examples section to keep it current with your evolving voice. ### Managing Tone Examples Open your sequence's **Settings** tab and scroll to the Tone Examples section to: * **View** examples organized by stage * **Edit** text to better match your voice * **Delete** examples that don't fit your style * **Add** custom examples from conversations that went well Use contractions (I'm, you're, they've) in your examples. Non-contracted language sounds robotic in DMs. *** ## Knowledge Base The Knowledge Base stores your sales documents so the AI can reference them when generating responses. Upload playbooks, case studies, FAQs, and objection handling docs to make the AI smarter and more context-aware. ### How It Works Documents are uploaded per-offer. After uploading, AutoReach automatically processes and indexes your documents so the AI can search them for relevant content when generating responses. Knowledge base is scoped per-offer, so different offers can have different knowledge bases. This ensures AI responses are tailored to each target audience. ### When Knowledge Base Is Used * **Cold DM generation**: References your value proposition, metrics, and case studies * **Response generation**: Finds relevant objection handling scripts, product details, or case studies based on the conversation topic * **Offer-specific context**: Ensures Enterprise case studies do not appear in SMB conversations ### Uploading Documents **Supported file formats:** * **PDF**: Sales decks, case studies, whitepapers, one-pagers * **DOCX**: Playbooks, sales guides, process documents * **TXT**: FAQs, objection handling scripts, notes, quick-reference sheets You can upload up to **10 documents per offer**. Quality matters more than quantity. A few well-written documents outperform a large collection of vague or overlapping content. **How to upload:** 1. Navigate to your Offer and open the **Knowledge Base** section 2. Click the upload area or drag a file onto it 3. Select a file from your computer (PDF, DOCX, or TXT) Processing takes a few seconds per document. Once complete, your knowledge base is ready to use. ### What Makes a Good Knowledge Base Document **Strong candidates:** * **Case studies** with concrete results: "We helped Company X reduce churn by 40% in 3 months" gives the AI a specific proof point to reference * **Objection handling scripts**: Common pushbacks with your proven responses * **Pricing and packaging guides**: So the AI can speak accurately about what you offer * **Product or service overviews**: Feature descriptions, use cases, and differentiators * **Implementation and onboarding docs**: Timeline, process steps, and what the customer can expect * **FAQ documents**: Answers to the questions your sales team hears most often **Documents to avoid:** * Generic marketing copy with no specifics * Internal engineering or technical documentation not relevant to sales conversations * Excessively long documents where key points are buried (break these into focused sections instead) ### Tips for Better AI Responses 1. **Be specific with numbers and outcomes**: "Reduced onboarding time from 6 weeks to 10 days" is far more useful to the AI than "We speed up onboarding" 2. **Use clear section headings**: Processing works best when your documents have logical sections with descriptive headers 3. **Write in the voice you want the AI to use**: If your knowledge base documents are formal, the AI will lean formal. If they are conversational, responses will reflect that 4. **Cover common objections explicitly**: Write out the objection and your ideal response. The AI will adapt this to each conversation 5. **Update after wins**: When you close a deal, add the story as a case study. Fresh, specific examples improve response quality over time 6. **Remove outdated content**: Old pricing, discontinued features, or stale case studies can cause the AI to reference information that is no longer accurate ## Next Steps * [**AI Response Engine**](/ai-and-conversations/ai-response-engine): How the AI uses tone examples and knowledge base during generation * [**Conversation Analyzer**](/ai-and-conversations/conversation-analyzer): AI suggestions for improving your conversations # Conversation Analyzer The Conversation Analyzer reviews your conversations, identifies patterns, and generates actionable suggestions to improve your AI's outreach voice. ## How It Works When you run an analysis: 1. Fetches all replied conversations for the sequence 2. Classifies outcomes (positive, negative, neutral) based on conversation status 3. Analyzes patterns across your conversations using AI 4. Generates actionable suggestions for improvement Analysis is triggered manually- it is not automatic. ## Suggestion Types The Analyzer generates four types of suggestions: ### Modify Tone Example An existing tone example isn't matching your actual voice. The Analyzer identifies the misalignment and suggests specific edits. ### Add Tone Example A conversation stage is missing examples or has too few. The Analyzer identifies a gap and suggests a new example based on your successful conversations. ### Remove Tone Example A tone example contains anti-patterns or is consistently associated with stalled conversations. The Analyzer recommends removing it. ### Modify Prompt Your AI prompt needs adjustment. Each prompt modification suggestion targets a specific prompt field: * **AI Prompt**- general conversation AI instructions * **How AI Engages With Posts**- the prompt for engaging with leads' posts (X, LinkedIn, and Instagram) * **DM Generation Prompt**- cold DM generation prompt ## Applying Suggestions For each suggestion, you can: * **Accept**: Select the suggestion for application * **Discard**: Remove the suggestion After reviewing, click **Apply** to implement all accepted suggestions at once, or **Dismiss** to clear all results. ## When to Run Analysis Run the Conversation Analyzer after accumulating enough conversations for meaningful patterns- typically after 20+ replied conversations. The more data available, the more reliable the suggestions. ## Next Steps * [**Tone and Knowledge Base**](/ai-and-conversations/tone-and-knowledge): Manage the tone examples that suggestions modify * [**AI Response Engine**](/ai-and-conversations/ai-response-engine): How tone and prompt changes affect responses # Overview Autopilot automatically configures and runs your entire outreach pipeline based on your Offer. Instead of manually creating searches, sequences, and enrollment rules, Autopilot sets everything up and keeps it running continuously. ## What Autopilot Sets Up When you enable Autopilot, it configures: * **Lead discovery**: Creates recurring searches (X, LinkedIn, and/or Instagram) based on your Offer criteria * **Lookalike discovery**: Finds influencer accounts in your space and extracts their followers as prospects * **Sequence creation**: Builds outreach sequences with AI-generated prompts and tone examples * **Auto-enrollment**: Enrolls qualified leads into sequences automatically * **Buyer expansion**: Discovers new prospects daily from your existing seed sources * **Monitor resurfacing**: Re-checks lower-scored leads for new buying signals ## Prerequisites * At least one **Offer** created with your ICP and value proposition * At least one **social account** connected (X, LinkedIn, and/or Instagram) On Instagram, Autopilot seeds discovery from a rotating hashtag pool and builds a default follow, like, watch story, then DM sequence. ## How It Runs Once set up, Autopilot runs five continuous operations in the background: 1. **Auto-enrollment** moves qualified leads into sequences 2. **Buyer expansion** discovers new prospects daily from your seed sources 3. **Monitor resurfacing** re-checks lower-scored leads for new buying signals 4. **Lookalike rotation** finds fresh seed accounts when existing sources are exhausted 5. **Signal search** runs intent-based searches to surface prospects posting about your pain points These run alongside the standard enrichment, scoring, and sequence execution that every lead goes through. Your only job is reviewing conversations in your Inbox and closing deals. See [Continuous Operations](/autopilot/continuous-operations) for details. ## Autopilot or Manual Setup? Everything Autopilot configures can also be set up manually. The choice is about how much control you want. **Use Autopilot when:** * You're new to AutoReach and want a working pipeline today, not next week. * Your offer is broad enough that "find people who match my ICP and reach out to them" is the goal. * You'd rather spend your time reviewing replies than building searches. **Configure manually when:** * You're targeting a narrow, named list of companies or accounts and need precise control over who gets contacted. * You want to A/B test different sequences against different lead segments. * You already have a working pipeline and just want to add a specific search or sequence without touching the rest. You can also start with Autopilot to get going, then switch to manual control once you understand what's working. Autopilot doesn't lock you in. Pausing Autopilot leaves the searches and sequences it created in place; you take them over from there. ## Next Steps * [**Enabling and Disabling Autopilot**](/autopilot/enabling): Walk through setup, pausing, and disabling * [**Continuous Operations**](/autopilot/continuous-operations): Auto-enrollment, buyer expansion, and monitor resurfacing in detail # Enabling & Disabling ## Enabling Autopilot Enabling Autopilot is a one-click process. The setup runs in the background and completes in 1-2 minutes. > **What Autopilot replaces vs. what's still on you:** Autopilot creates the searches, sequences, tone examples, and enrollment rules so you don't have to. It does **not** replace daily limit configuration, conversation review in your Inbox, Engagement Engine content approval (unless you turn on auto-post), or meeting tracking setup. Think of it as automating the *setup and discovery* loop: the human-judgment parts (tone tuning, replying to hot leads, deciding when to scale up) still belong to you. ### Prerequisites * At least one **Offer** created with your ICP and value proposition * At least one **social account** connected (X, LinkedIn, and/or Instagram) * No existing active Autopilot config for the same Offer ### The Enable Flow When you click "Enable Autopilot," the following steps execute: **Step 1: Create Configuration** Autopilot is activated immediately. Any prior disabled configuration for the same Offer is removed first. The UI returns immediately while setup continues in the background. **Step 2: Generate AI Content** Autopilot generates a shared campaign prompt and tone examples for your Offer. This content is reused across all sequences created in the next step. **Step 3: Create Sequences (Per Platform)** For each connected platform (X, then LinkedIn, then Instagram), Autopilot: 1. Creates a sequence with AI-generated steps, tone examples, and auto-enrollment enabled 2. Generates keywords (or hashtags, for Instagram) from your Offer 3. Creates a recurring search for that platform 4. Finds a lookalike influencer account via AI search 5. Creates a seed extraction source (follower extraction for X and Instagram, or a people search with buyer expansion for LinkedIn) **Step 4: Background Loops Begin** Once setup completes, Autopilot's continuous operations (lookalike rotation, auto-enrollment, signal search) begin running automatically. ### What to Expect After Enabling * **Searches begin immediately**: Your first recurring searches start running * **Follower extraction starts**: Seed accounts begin extracting followers * **First leads appear**: Typically within the first few hours * **Auto-enrollment activates**: Qualified leads are enrolled into sequences automatically *** ## Pausing Autopilot Pause when you want to temporarily stop all Autopilot operations and resume later exactly where you left off. **When to pause:** Vacations, short breaks, account maintenance, or adjusting settings before resuming. **To pause:** Click the pause icon on the autopilot card. **To resume:** Click the play icon on the autopilot card. Autopilot restores searches, reactivates sequences, and resumes operations. When paused: * Searches are paused (preserved) * Sequences are paused * Seed sources and lookalike accounts are preserved * All leads and conversations are preserved *** ## Disabling Autopilot Disable when you are stopping outreach for the long term or want a clean slate. **When to disable:** Fully stopping outreach for an Offer, switching to manual-only mode, or pivoting to a different ICP. **To disable:** Click the trash icon on the autopilot card, then confirm by clicking **Remove**. When disabled: * Recurring searches are stopped * Sequences are paused * Lookalike expansion is disabled * All leads, conversations, and historical data are preserved ### Re-enabling After Disable When you re-enable Autopilot after disabling: * New sequences, searches, and lookalike accounts are created fresh * AI generates a new campaign prompt and tone examples * New seed accounts are discovered for follower extraction * Existing leads in your database are preserved and can be re-scored *** ## Pausing Individual Sequences You can pause a single sequence without affecting the rest of Autopilot: 1. Go to **Sequences** 2. Select the sequence 3. Click **Pause** This stops outreach for that sequence only. Other sequences and Autopilot operations continue normally. ## Next Steps * [**Autopilot Overview**](/autopilot/overview): All Autopilot operations at a glance * [**Continuous Operations**](/autopilot/continuous-operations): Auto-enrollment, buyer expansion, and monitor resurfacing # Continuous Operations Once Autopilot is enabled, five continuous operations run in the background: auto-enrollment, buyer expansion, monitor resurfacing, lookalike rotation, and signal search. Together, they keep your pipeline growing and ensure qualified leads reach your sequences without manual intervention. *** ## Auto-Enrollment Auto-enrollment is the bridge between lead scoring and outreach. It automatically moves qualified leads into sequences. ### How It Works Autopilot regularly checks for enrollable leads. Each cycle: 1. Fetches leads in Active buyer state 2. Filters out leads already enrolled in the target sequence 3. Validates each lead has the required platform data 4. Enrolls qualified leads into the matching sequence There is also a real-time enrollment hook: when scoring marks a lead as Active, enrollment can trigger immediately. ### Enrollment Criteria A lead is enrolled when all conditions are met: | Condition | Details | | -------------------- | ----------------------------------------------------------------------------------------------------- | | Buyer state | Must be Active | | Sequence status | Must be active with auto-enrollment enabled | | Offer match | Lead's offer must match the sequence's offer | | Platform data | X requires an X profile; LinkedIn requires a LinkedIn profile; Instagram requires an Instagram handle | | Not already enrolled | Lead must not already be in the sequence | ### Platform Routing Leads are automatically routed to the correct platform sequence: * Leads with X profile data go to the X sequence * Leads with LinkedIn profile data go to the LinkedIn sequence * Leads with an Instagram handle go to the Instagram sequence * Leads with data for more than one platform are enrolled in a sequence that supports one of those platforms ### Controlling Auto-Enrollment When Autopilot is enabled, auto-enrollment is turned on for all newly created sequences by default. To review leads before enrollment, disable auto-enrollment on the sequence. Leads will then appear on the Buyers page for manual review. *** ## Buyer Expansion Buyer Expansion discovers new prospects daily by mining your existing pipeline sources. ### X Expansion On X, expansion works by increasing follower extraction targets on your seed accounts: 1. Calculates a daily batch of followers to extract 2. Increases the extraction target on the seed account 3. Triggers the extraction worker to pick up new followers 4. New followers enter the scoring pipeline automatically Before expanding, Autopilot also triggers a **lead pool match** to find instant results from leads already in your database that match your ICP. ### LinkedIn Expansion On LinkedIn, expansion takes a role-based approach: 1. Rotates through a priority list of decision-maker roles, one role per day 2. Searches for people matching the current role using your Offer's targeting criteria 3. If a role returns no results, immediately rotates to the next role Roles are prioritized by seniority, searching for senior decision-makers first before moving to lower levels. This builds a well-rounded prospect list across different seniority levels at your target companies. ### Instagram Expansion On Instagram, expansion mines your existing seed searches, rotating through follower, following, and hashtag sources to surface fresh prospects. As with the other platforms, it respects the wait-for-completion gate: a source does not re-fire while the previous batch's leads are still draining through enrichment and scoring. ### Exhaustion Detection All platforms include automatic exhaustion detection: * **X accounts**: Once all available followers are extracted, the source is marked exhausted. Lookalike rotation then finds a new seed account. * **LinkedIn roles**: Once a role's results are exhausted, expansion moves to the next priority role. * **Instagram sources**: Once a source is fully mined, expansion is disabled for that source and rotation can find a new seed account. *** ## Monitor Resurfacing Monitor Resurfacing re-checks lower-scored leads for new buying signals. Instead of abandoning leads that are not ready yet, Autopilot periodically evaluates them and upgrades them to active outreach when their signals improve. ### Check Frequency Leads are checked on a regular schedule. Warmer leads are checked more frequently than colder leads, focusing attention on leads most likely to become ready for outreach soon. ### What Gets Checked During each resurfacing window, Autopilot checks three categories of signals: **Recent Activity** - Fetches new posts and social activity for the lead. New posts and engagement patterns are fed back into the scoring system. **LinkedIn Headline Changes** - Compares the lead's current LinkedIn headline against the stored headline. If a change is detected, the lead receives a score boost and tenure is reset. A job change often indicates new priorities and buying intent. **Company Jobs Refresh** - If the lead has an associated company and the company's hiring data is stale, Autopilot refreshes the company's job listings. Active hiring can indicate growth and budget availability. ### Score Upgrades When resurfacing detects new signals and the lead's score improves: * **Reaches Active threshold** - Lead is upgraded to Active buyer state and eligible for auto-enrollment into sequences * **Score improves moderately** - Lead moves to a warmer tier with more frequent re-checks * **No improvement** - Lead remains in its current tier and is re-checked at the next interval Leads upgraded to Active are picked up by auto-enrollment and enrolled into the appropriate sequence. *** ## Lookalike Rotation When an X or Instagram seed account's followers are fully extracted, Autopilot automatically discovers a new lookalike account and begins extracting its followers. This keeps your pipeline growing without manual seed management. *** ## Signal Search Autopilot periodically runs intent signal searches to find prospects actively posting about problems your product solves. These searches use your offer's pain points and target audience to surface high-intent leads from X, LinkedIn, and Instagram. ## Next Steps * [**Autopilot Overview**](/autopilot/overview): All Autopilot operations at a glance * [**Enabling and Disabling Autopilot**](/autopilot/enabling): Setup, pausing, and disabling # Overview The Engagement Engine builds a credible social presence on your accounts before you start outreach. It automatically posts content, engages with relevant posts, and grows your visibility so that when leads receive your messages, they see an active account with genuine engagement history. ## When to Run It The Engagement Engine matters most in two situations: * **Before you launch sequences on a new or inactive account.** Cold DMs from an empty account get ignored or flagged. Run the Engagement Engine for **at least 1-2 weeks** on a fresh account to build a posting history before you start sending outreach. Skipping this step is one of the most common reasons reply rates stay disappointingly low. * **Continuously, alongside your outreach.** Even on established accounts, ongoing engagement keeps you visible in your prospects' feeds. When a lead sees your account commenting on a post they care about, then receives a DM from you, the message lands as a follow-up to existing presence rather than as a cold outreach. You can run sequences without the Engagement Engine (it's not a hard requirement), but reply rates are typically meaningfully higher when you do. ## What the Engagement Engine Does The Engagement Engine automatically: * **Posts content** aligned with your offer and content pillars (X tweets and LinkedIn posts) * **Engages with relevant posts** in your niche (likes, comments, follows/connections) * **Maintains human-like patterns** with randomized timing, skip days, and natural variation ## How It Works The system runs continuously throughout your activity window, distributing actions into natural sessions with randomized timing. ## Available Platforms | Platform | Action Types | | ------------- | ------------------------------------------- | | **X** | Tweets, likes, follows, replies | | **LinkedIn** | Posts, likes, connection requests, comments | | **Instagram** | Likes, comments, follows, story views | You can run the Engagement Engine on multiple accounts simultaneously. Instagram engagement targets are discovered from niche-relevant hashtags generated from your offer and refreshed on a rolling basis. The same human-like timing, skip days, and natural variation apply across every platform. ## Account Persona When setting up the Engagement Engine, you can optionally define a persona to make generated content sound like a real person rather than a brand account: * **Full name and role** - e.g. "Sarah Chen, Growth Lead" * **Location** - e.g. "Los Angeles, CA" * **Voice and personality** - 1-2 sentences describing tone, e.g. "Direct and data-driven, shares real numbers and builder insights." All three fields are optional. If provided, the AI uses them to shape the voice and perspective of generated posts and engagement comments. ## Daily Activity The Engagement Engine generates a randomized mix of actions each day. Action counts vary daily to mimic natural human behavior, including occasional skip days with minimal activity. See [**Engagement Automation and Daily Actions**](/engagement-engine/engagement) for more details on daily activity patterns. ## Engagement Pod (LinkedIn, Instagram, and X) Engagement Pod is an opt-in feature where AutoReach users automatically engage with each other's posts to boost algorithmic reach. When you publish a post, AutoReach detects it and schedules a small group of pod members to like and comment on it shortly after. The same happens in reverse for posts from other pod members. How it works: * Up to 5 pod members are selected per post, with a short, randomized gap between their engagements to keep the pattern natural * AI-generated comments are tailored using each pod member's offer context and persona, so they don't look templated * A 3-day cooldown prevents the same author/engager pair from showing up repeatedly, which would look reciprocal The Engagement Pod toggle lives on each account's Engagement Engine card and is **on by default**. Turn it off if you don't want your account participating - you'll lose the inbound engagement boost on your own posts as well. ## Content Approval All generated content enters an approval queue by default. You can: * **Approve**: post as-is at the scheduled time * **Discard**: permanently remove the content * **Regenerate**: replace with a new AI-generated version * **Edit**: modify the text before approving Enable **auto-post** to skip manual review and post content automatically. A separate toggle controls whether engagement comments (and X replies) are auto-approved; it is labeled **Auto-comment** on LinkedIn and Instagram, and **Auto-reply** on X. Pending approvals expire if not reviewed. (Instagram has no auto-post toggle, since the Engagement Engine does not publish standalone Instagram posts.) ## Next Steps * [**Content Strategy & Pillars**](/engagement-engine/content-strategy): How content themes are organized * [**Engagement Automation**](/engagement-engine/engagement): Daily activity patterns and engagement targeting * [**Content Generation & Approvals**](/engagement-engine/content-generation): How content is created and reviewed # Content Strategy & Pillars The Engagement Engine organizes content around a **content strategy** with rotating weekly themes. This structure ensures your posts stay relevant to your niche and build a consistent professional presence. ## Content Pillars Each strategy has **content pillars** - core themes representing different angles of your expertise and value proposition. Pillars are AI-generated from your offer, ICP, and pain points. Example pillars for a sales automation company: 1. Industry Insights - market trends and analysis 2. Practical Tips - actionable advice and how-tos 3. Case Studies - results and examples 4. Thought Leadership - original perspectives on challenges 5. Community Engagement - audience interaction and discussions ### How to Define Effective Pillars Pillars are generated automatically, but understanding what makes a good pillar helps you evaluate and refine the AI's output. Strong pillars share a few characteristics: * **Each pillar covers a distinct angle** of your expertise. Avoid overlap - if two pillars both address "tips for sales teams," merge them or sharpen the distinction. * **Pillars should appeal to your ICP's interests**, not just your product features. A cybersecurity company might have a "Compliance and Regulation" pillar because that is what their buyers care about, even though the product is a technical tool. * **Mix educational and engagement-oriented pillars**. Purely educational content builds authority, but pillars focused on community discussion or audience interaction drive replies and visibility. **Example pillars by industry:** * **Marketing agency**: Brand Strategy, Campaign Performance, Client Stories, Platform Updates, Creative Process * **HR tech startup**: Hiring Trends, Employee Retention, DEI Insights, Workplace Culture, Product Use Cases * **Financial advisor**: Market Commentary, Retirement Planning, Tax Strategy, Client Wins, Economic Outlook ## Plan Structure The content strategy generates: * **Content pillars** based on your offer * **Weekly themes**- one per week, cycling through each pillar in rotation * **Post ideas per pillar**- ready-to-use content seeds Each week, all generated content (posts, engagement comments) aligns with that week's pillar. This creates a natural rotation that keeps content fresh and varied. ### How Weekly Themes Work in Practice Each week is assigned a focus pillar. If you have 5 pillars, the rotation cycles through all 5 over five weeks, then repeats. Week 1 might focus on "Industry Insights," Week 2 on "Practical Tips," and so on. By Week 6, the cycle restarts with "Industry Insights" but with a different weekly theme within that pillar. This means your audience sees variety week to week, and over time every pillar gets roughly equal attention. The weekly theme gives the AI a specific angle within the pillar - for example, under "Industry Insights," one week's theme might be "emerging trends in AI-powered sales" while another week focuses on "how buyer behavior is shifting." ### How Post Ideas Are Generated Each pillar comes with specific post ideas- ready-to-use content seeds that the AI draws from when generating posts. These ideas are created based on your offer description, target audience, and pain points. They are tailored to your niche and designed to resonate with the type of people you are trying to reach. When the Engagement Engine creates a post, it selects from the current week's pillar ideas, adapts the idea to the platform (shorter for X, longer for LinkedIn), and generates the full content. ### Reviewing and Customizing the Plan After your strategy is generated, you can review the pillars, themes, and post ideas in the Engagement Engine settings. While the AI handles the initial generation, you can regenerate the strategy at any time if your offer changes or if you want a fresh set of content ideas. Updating your offer description, pain points, or target audience will produce a more relevant strategy on regeneration. ## How the Strategy Is Generated The content strategy is AI-generated using: * Your offer context and ICP * Pain points and target audience details * Current date and year (prevents stale references) * Professional voice and tone For LinkedIn, the AI also generates topic-specific approaches for engaging with other people's posts, aligned with your content pillars. ### Tips for Getting Better AI-Generated Content * **Write a detailed offer description**: The more context the AI has about what you do, who you serve, and what results you deliver, the more specific and relevant the generated content will be * **Include real pain points**: Generic pain points produce generic content. "Sales teams waste 10 hours a week on manual data entry" is far more useful than "improve productivity" * **Keep your offer updated**: If your positioning, pricing, or target audience changes, update your offer and regenerate the strategy so content stays aligned * **Review the first week's output**: Check the posts generated for the first week. If the tone or topics feel off, refine your offer description and regenerate ## Strategy Creation Your content strategy is created when you: 1. **Manually configure** the Engagement Engine with an associated offer 2. **Enable Autopilot** - which may create an Engagement Engine as part of its setup No manual pillar selection is needed. The AI determines the optimal pillar structure from your offer. ## Keyword-Based Engagement Targeting When the Engagement Engine engages with other posts (likes across platforms, comments on LinkedIn and Instagram, replies on X), it uses keywords derived from your content pillars and offer to find relevant content. Only posts matching your niche keywords are targeted for engagement. ## Next Steps * [**Content Generation**](/engagement-engine/content-generation): How content is created from pillars and themes * [**Engagement Automation**](/engagement-engine/engagement): How content is distributed and how pillars guide engagement targeting # Content Generation & Approvals All Engagement Engine content is AI-generated using your offer context, content pillars, and the current week's theme. You control what gets posted through the approval queue. ## Generation Flow 1. **Input**: Your offer, ICP, pain points, weekly pillar, and current date 2. **AI generation**: Content is generated using your configured AI models 3. **Approval routing**: Content enters the approval queue as **Pending Approval**, or goes directly to the post queue if auto-approval is enabled ## X Tweets Tweets are short-form content generated daily based on the current week's pillar: * Concise and engaging, optimized for the X platform * References current year and recent trends * Matches your professional voice and tone ## LinkedIn Posts LinkedIn posts are longer, narrative-driven content: * Professional tone suited to the LinkedIn platform * Designed for engagement, comments, and shares ## Engagement Comments and Replies When the Engagement Engine comments on LinkedIn or Instagram posts, or replies to tweets on X, it generates contextual responses: * References the original post specifically * Adds value to the conversation (not generic "great post" responses) * Matches your professional voice * Concise and conversational ## Year Awareness All content generation prompts include current date and year context. This ensures posts reference recent trends and do not contain outdated year references. *** ## Approval Queue All generated content enters the approval queue as "Pending Approval" unless auto-approval is enabled. Each item shows the content text and its scheduled post time. ### Approval Statuses | Status | Description | | ---------------- | ---------------------------------------------------- | | Pending Approval | Awaiting your review | | Approved | Approved and waiting to post at scheduled time | | Queued | Picked up for posting | | Posted | Successfully published (tweets/posts) | | Completed | Successfully published (engagement comments) | | Discarded | Permanently removed by you | | Skipped | Expired without review or auto-approval was disabled | | Failed | Post attempt failed | ### Actions **Approve** - Post the content as-is at its scheduled time. The status moves to "Approved" and the item is queued for posting. **Regenerate** - Replace the content with a new AI-generated version. The new version enters the queue as a new "Pending Approval" item. **Discard** - Permanently remove the content. No replacement is generated. **Edit** - Modify the text before approving: 1. Open the approval queue 2. Edit the text as needed 3. Approve the edited version ### Auto-Approval Two separate toggles in your Engagement Engine settings control auto-approval: * **Auto-post** - Generated posts (tweets, LinkedIn posts) skip the approval queue and post automatically * **Auto-comment** - Generated engagement comments (on other posts) skip the approval queue When enabled, content posts automatically at scheduled times without manual review. You can still view posted content in activity logs. When disabled, all content requires manual review before posting. ### Approval Expiry Pending approvals expire after **48 hours**. Stale approvals are automatically marked as "Skipped". This prevents a backlog of outdated content from accumulating if you do not review the queue regularly. ## Next Steps * [**Content Strategy and Pillars**](/engagement-engine/content-strategy): How content themes are organized * [**Engagement Automation**](/engagement-engine/engagement): How engagement targeting works * [**Engagement Engine Overview**](/engagement-engine/overview): Return to the overview # Engagement Automation While your account posts content, the Engagement Engine simultaneously interacts with relevant posts in your niche - liking, commenting, and following to build visibility. ## Engagement Types ### Likes Target posts matching your niche keywords on X and LinkedIn. ### Comments and Replies AI-generated contextual responses that reference the original post and add value to the conversation - not generic responses like "great post." These run as comments on LinkedIn and Instagram, and as replies to tweets on X. Comments and replies can be auto-approved (the Auto-comment / Auto-reply toggle) or held for your review in the approval queue. ### Follows / Connection Requests Follow relevant accounts in your niche on X and Instagram, or send connection requests on LinkedIn. ## Keyword-Based Targeting Engagement uses keywords derived from your offer and content pillars to find relevant content. Only posts matching your niche keywords are candidates for engagement. This keeps interactions focused and relevant rather than indiscriminate. ## Smart Targeting The Engagement Engine prioritizes accounts where your interaction is most likely to be noticed and reciprocated, optimizing for visibility and response rate. ## Engagement and Outreach Are Independent Engagement Engine activity and outreach sequences run simultaneously and independently. The Engagement Engine builds your account presence while sequences handle direct lead outreach. Neither blocks the other. *** ## Daily Action Allocation The Engagement Engine generates a randomized mix of actions each day. Counts vary daily to mimic natural human behavior. ### Daily Activity Each day, the Engagement Engine performs a mix of actions across your connected platforms: **X (Twitter):** Likes, follows, replies, and tweets - with daily variation in counts. **LinkedIn:** Likes, connection requests, comments, and posts - with daily variation in counts. **Instagram:** Likes, comments, follows, and story views - with daily variation in counts. (No standalone posts.) Some days have reduced or zero activity for certain action types, and occasional skip days with minimal activity simulate natural human behavior. Outreach sequences are **not affected** by skip days - only Engagement Engine activity pauses. ### Action Scheduling Actions are distributed into natural sessions within your activity window, with randomized timing between actions to avoid patterns. ### Account Safety The Engagement Engine automatically reduces activity if account errors occur, protecting your accounts from platform detection. If errors persist, activity scales down progressively until the issue resolves. ## Next Steps * [**Content Strategy and Pillars**](/engagement-engine/content-strategy): How pillars guide engagement targeting * [**Content Generation and Approvals**](/engagement-engine/content-generation): How content is created and reviewed # Overview AutoReach generates short-form vertical videos (reels) from a topic and your offer. You get a finished MP4 plus a ready-to-paste caption, so you can post inbound content to Instagram, TikTok, LinkedIn, X, or YouTube Shorts without filming or editing anything yourself. Reels are a content tool, not an outreach action. They build the kind of active, credible presence that makes your outreach land, in the same spirit as the [Engagement Engine](/engagement-engine/overview). ## Where It Lives Reels are generated per account, from the **Videos** tab on an account's detail page (**Accounts → \[account] → Videos**). Each account keeps its own character photo, voice, and reel history. ## What You Provide * An **offer** to ground the video in your product, audience, and value proposition * A **topic** (or let AutoReach suggest topics for you) * A **format** that sets the structure and length (see [Creating a Reel](/ai-video-and-reels/creating-a-reel)) * Optionally: a character reference photo and a voice, so the on-camera presenter looks and sounds like you (see [Setup, Requirements, and Costs](/ai-video-and-reels/setup-and-costs)) ## What You Get When a reel finishes, AutoReach delivers: * A finished **vertical MP4** you can download * A **caption** with hashtags, ready to paste into the platform's caption field * The full **script** the video was built from * The final **duration** Generated videos are stored for **3 days**, so download anything you want to keep. ## Posting Reels are **not** auto-posted to any platform. You download the MP4, add trending audio if you want, and post it yourself. This keeps you in control of timing, audio, and platform-specific captions, and lets you reuse a strong reel across multiple platforms. ## How It Works at a High Level Each reel is produced through an automated pipeline: AutoReach writes a script from your topic and offer, generates a voiceover, builds synchronized on-screen captions, plans the visual cuts, generates the imagery and motion, then renders everything into a finished video with cinematic polish. The whole process runs in the background and typically takes several minutes. You can watch progress in real time and retry any reel that fails, paying only to redo the step that broke rather than the whole video. See [Creating a Reel](/ai-video-and-reels/creating-a-reel) for the full lifecycle. ## Next Steps * [**Creating a Reel**](/ai-video-and-reels/creating-a-reel): Formats, inputs, and the generation lifecycle * [**Setup, Requirements, and Costs**](/ai-video-and-reels/setup-and-costs): API keys, character photos, voices, and pricing # Creating a Reel Open the **Videos** tab on an account, click **Generate Reel**, and fill in the options below. The reel queues immediately and processes in the background. ## Reel Options ### Offer Select the offer the reel should speak to. The offer's audience, value proposition, and knowledge base shape the script, so a well-defined offer produces a sharper video. See [Offers and Knowledge Base](/core-concepts/offers-and-knowledge-base). ### Topic Enter the subject of the reel, for example *"Why cold DMs fail before the second follow-up."* If you are not sure what to make, click **Suggest Topics** and AutoReach proposes ideas tailored to your offer and the format you picked. ### Format The format sets the structure, pacing, and length band of the video: | Format | What it is | | ---------------------- | ------------------------------------------------------------------------------- | | **Talking Head** | A presenter speaks to camera. Uses your character photo and voice. The default. | | **No Person** | Narrator-only. Voiceover over visuals, with no on-camera presenter. | | **Short Motivational** | A punchy, high-energy clip (roughly 8 to 15 seconds). | | **Listicle** | A "Top N" style rundown. | | **UGC** | A casual, selfie-style first-person take, the look of organic creator content. | | **UGC with Inserts** | The selfie-style take plus cutaway shots that illustrate the point. | ### Energy Vibe (optional) Set the delivery tone, or leave it on automatic and AutoReach picks one that fits the format. Options: * **Calm and authoritative** * **Urgent and punchy** * **Vulnerable and personal** * **Sarcastic and dry** * **Inspirational and warm** ### Creative Brief (optional) A free-text box for any extra direction on framing, visuals, or tone that the energy vibe selector does not cover. Optional. ### Include CTA Toggle whether the reel ends with a call to action that points back to your offer. ### Quality Choose **Fast** for quicker, cheaper generation or **Full** for higher-fidelity video. Fast is a good default while you are dialing in topics and formats; switch to Full for reels you intend to publish widely. ### Duration Override (optional) Each format has a natural length range. You can clamp the final video to a specific length within that range if you need it to hit a particular runtime. ### Cost Estimate Before you generate, AutoReach shows an estimated cost for the reel based on your format, quality, and duration choices. See [Setup, Requirements, and Costs](/ai-video-and-reels/setup-and-costs). ## Creative Variation So your reels do not all feel the same, AutoReach automatically varies the creative direction (the hook style, visual treatment, and opener) from one reel to the next, avoiding choices it used on your recent videos. You do not need to configure this; it happens on every generation. ## The Generation Lifecycle A reel moves through these statuses: | Status | Meaning | | -------------- | ------------------------------------------------ | | **Queued** | Waiting for a worker to pick it up | | **Generating** | Being produced (the current step is shown) | | **Ready** | Finished. MP4, caption, and script are available | | **Failed** | A step could not complete (the reason is shown) | While **Generating**, AutoReach shows the current step so you can see where it is: 1. **Script** - writing the script from your topic and offer 2. **Voice** - generating the voiceover 3. **Captions** - building synchronized on-screen captions 4. **Cuts** - planning the visual cuts and effects 5. **Visuals** - generating the imagery and motion for each cut 6. **Render** - assembling the final video Generation runs in the background and typically takes several minutes. You can leave the page and come back. ## Retrying a Failed Reel If a reel fails, the row shows a **Retry** button and a short reason. Retrying re-runs the pipeline but reuses every step that already succeeded, so it resumes from the step that broke instead of starting over. You only pay again for the failed step and the steps after it. If a worker restarts while a reel is mid-generation, AutoReach automatically recovers it on the next boot and resumes from where it left off; you do not need to do anything. ## Viewing and Managing Reels The Videos tab lists every reel for the account with its status, duration, and actions: * **Download** the finished MP4 * **View Script** to read the full script and caption * **Retry** a failed reel * **Delete** a reel you no longer need Remember that finished videos expire after 3 days, so download anything worth keeping. ## Next Steps * [**Setup, Requirements, and Costs**](/ai-video-and-reels/setup-and-costs): API keys, character photos, voices, failures, and pricing * [**AI Video and Reels Overview**](/ai-video-and-reels/overview): How reels fit into your content strategy # Setup, Requirements & Costs Reel generation uses several AI providers under the hood. Because AutoReach runs them with your own API keys, you connect those providers once, then generate as many reels as you like. ## Required API Keys Add these in **Settings → AI & Models**. Reel generation needs all of them: | Provider | Used for | | -------------- | --------------------------------------------------------- | | **Anthropic** | Writing the script and planning the visual cuts | | **OpenAI** | Transcribing the voiceover to build synchronized captions | | **ElevenLabs** | Generating the voiceover | | **fal.ai** | Generating the imagery, motion, and on-camera lipsync | An optional **Pexels** key lets AutoReach pull in stock footage for some shots. If a required key is missing, the reel fails with a configuration error that tells you which provider to connect. ## Character Photo and Voice For formats with an on-camera presenter (Talking Head, UGC, UGC with Inserts), set up a character so the presenter looks and sounds like you. Both live on the account's **Videos** tab. ### Character Reference Photo Upload a reference photo of the person who should appear on camera, and label the angle it shows (front, left profile, right profile, looking up, looking down, or other). A clear, well-lit, front-facing photo gives the most consistent results across reels. AutoReach reuses this reference so the presenter stays recognizable from one video to the next. If you change the photo, the cached description refreshes automatically on the next reel. ### Voice Pick an **ElevenLabs voice** for the account. You can use one of ElevenLabs' prebuilt voices or a voice you have cloned in ElevenLabs. The selected voice is used for every reel on that account until you change it. ## Costs Each reel's cost depends on its format, quality preset, and duration, and is split across the providers above (script, voiceover, transcription, image generation, and video generation). Higher quality and longer durations cost more; the **Fast** preset and shorter formats are the cheapest. AutoReach shows a **cost estimate** in the generation modal before you commit, and records the actual cost once the reel finishes. Because generation runs on your own provider keys, these charges come from those providers directly. To keep costs down while experimenting, use the **Fast** quality preset and shorter formats (Short Motivational, UGC) until you have a topic and look you are happy with, then switch to **Full** for the reels you plan to publish. ## Storage and Expiry Finished videos are stored for **3 days**, then removed. Download any reel you want to keep. The script and caption remain visible in the reel's history entry, but the MP4 itself is gone after expiry, so save the file locally or repost it before then. ## When a Reel Fails Failures are labeled with a short category so you know what to do: | Category | What it usually means | | --------------- | ------------------------------------------------------------------------------------ | | **Config** | A required API key or character photo is missing | | **Script** | The script could not be generated; retry usually clears it | | **Voice (TTS)** | Voiceover generation failed (check your ElevenLabs key and voice) | | **Visuals** | Image generation failed; retry resumes from this step | | **Lipsync** | The on-camera lipsync step failed | | **Render** | The final assembly failed | | **Refusal** | A provider declined the request on content-policy grounds; adjust the topic or brief | In almost every case, click **Retry**: it resumes from the failed step and reuses everything that already succeeded, so you do not pay to redo the whole video. For a refusal, soften or reword the topic and creative brief before retrying. ## Next Steps * [**Creating a Reel**](/ai-video-and-reels/creating-a-reel): Formats, inputs, and the generation lifecycle * [**AI Model Configuration**](/settings-and-configuration/ai-models): Where API keys are managed # Meetings, Call Briefs & CRM AutoReach integrates with Calendly and Cal.com to detect when leads book meetings, and generates AI-powered call briefs to prepare you for every conversation. ## When to Set This Up Set up meeting tracking **early, ideally during onboarding**, even if you don't think you need it yet. Two reasons: * **Without a webhook, AutoReach can't know when a lead books.** It will still send your booking link in messages, and leads will still book meetings, but those bookings don't get marked in your pipeline automatically. You end up tracking meetings in your head or in a spreadsheet. * **Meeting status feeds back into the system.** When a conversation reaches Meeting Booked, AutoReach extracts the winning exchanges as tone examples for future replies. Without webhook tracking, the AI doesn't learn from your wins. If you don't have a paid Calendly plan, use **Cal.com**, which is free and has one-step webhook setup. If you only have a custom booking URL, you can still inject it into messages via `{{booking_link}}`, but you'll need to track bookings manually. *** ## Booking Integration ### Supported Platforms * **Calendly** - requires a paid subscription for webhook support * **Cal.com** - free, no paid subscription required * **Custom booking URLs** - any scheduling page (no automatic tracking) Calendar configuration is **per-account**. Open an account from the **Accounts** page and use the Calendar section. The form has two parts: (1) the booking page (provider + URL), and (2) meeting tracking (provider-specific webhook setup, surfaced once the booking page is saved). ### Calendly Setup (Easier, No Form Field Required) Calendly webhooks require a paid Calendly plan (Standard, Teams, or Enterprise). AutoReach now uses an **invisible attribution** approach for Calendly. You no longer need to add a custom form field to your event, bookers see Calendly's default name and email fields only. 1. In the account's Calendar section, choose **Calendly** and paste your booking URL 2. Generate a Personal Access Token in [Calendly API settings](https://calendly.com/integrations/api_webhooks) and paste it into AutoReach 3. Click **Auto-fetch** to populate your Organization URI (or paste it manually) 4. Click **Register webhook** - AutoReach creates the webhook for you via the Calendly API Attribution happens invisibly: AutoReach appends `utm_content=:` to the booking URL injected via `{{booking_link}}`, and the webhook handler reads it from the booking's tracking metadata. Bookers see no extra fields and there is no friction at the booking step. > **Legacy events:** If you previously added a "Username" form question, you can leave it or remove it. AutoReach falls back to the first form answer when `utm_content` is not present. ### Cal.com Setup Cal.com is free and does not require a paid plan. Setup is a two-step process: **Step 1: Register the webhook** 1. In the account's Calendar section, choose **Cal.com** and paste your booking URL 2. Copy the webhook URL AutoReach generates 3. In Cal.com, go to **Settings → Developer → Webhooks → New Webhook** 4. Paste the webhook URL, subscribe to the `BOOKING_CREATED` event, and save **Step 2: Add a hidden username field to each event type** Bookings need a `username` identifier so AutoReach can match them to a lead. To keep the field invisible: 1. In Cal.com, open the event type → **Advanced → Booking Questions** 2. Click **Add a question** with type **Short Text** and identifier `username` 3. Check **Disable input if the URL identifier is prefilled**, which hides the field from bookers 4. Save AutoReach prefills `username` via the `{{booking_link}}` URL, so the field is filled invisibly and bookers never see it. Repeat step 2 for every event type you want to track. ### Custom Booking URLs You can use any booking URL without webhook integration. The `{{booking_link}}` variable still works for injecting your URL into messages, but AutoReach will not automatically detect bookings. You will need to update meeting status manually. ### How Lead Matching Works When a booking webhook fires, AutoReach matches the attendee to a lead using the tracking parameter embedded in your booking URL via the `{{booking_link}}` template variable, or by matching the attendee's email address. ### What Happens When a Meeting Is Booked When a lead is matched: 1. The lead's status is set to **Meeting Booked** 2. Any pending follow-up actions are cancelled 3. The meeting is recorded in the conversation thread 4. Meeting count is updated in sequence statistics 5. A winning tone example is auto-captured from the conversation ### Using `{{booking_link}}` in Templates The `{{booking_link}}` template variable injects your calendar URL with lead-specific tracking: ``` Book a time that works: {{booking_link}} ``` This becomes a personalized booking link with the lead's identifier appended as a tracking parameter. *** ## Call Brief Generation Call briefs are AI-generated pre-call preparation documents that combine lead data, conversation history, and buyer intelligence into actionable talking points. ### How to Generate a Call Brief 1. Open any conversation in your **Inbox** 2. Click the **Call Brief** button 3. The brief generates in a few seconds and appears inline The brief is saved to the conversation's metadata for future reference. ### What's Included Call briefs include sections tailored to the available data: | Section | Content | | ------------------------- | ----------------------------------------------------------------------- | | Lead Snapshot | Current role, company, industry, size, location, buyer score and status | | Company Intelligence | LinkedIn company data, funding, tech stack | | Conversation Recap | Summary of all prior messages, key topics, commitments | | Pain Points Identified | Specific challenges mentioned in the conversation | | Talking Points and Agenda | Suggested discussion topics tailored to the lead | | Recent Activity | Recent posts from the lead | | Offer Alignment | How your offer maps to their situation | | Objection Preparation | Anticipated pushback and counter-points | | Competitive Intel | Competitive context from your offer | | Recommended Next Steps | Stage-aware suggestions for what to propose | Each section is only included if relevant data is available. ### Data Sources The brief draws from: * **Lead profile**: Bio, headline, location, email, website * **Company data**: Industry, size, funding, tech stack * **Career history**: Work experience, education, skills * **Web enrichment**: Company technologies, enrichment summary * **Conversation transcript**: Full message history * **Buyer intelligence**: Fit, intent, timing, and composite scores * **Recent posts**: Latest social activity * **Offer context**: Your value proposition and pain points ### Stage-Aware Recommendations The brief adapts its recommendations based on conversation status: | Status | Brief Focus | | -------------- | ------------------------------------------------------ | | Meeting Booked | Focuses on meeting preparation, never suggests booking | | Replied | Focuses on advancing toward a booking | | Pending | Focuses on follow-up strategies | *** ## Chrome Extension CRM Pipeline The AutoReach Chrome Extension brings your CRM pipeline into your browser. It works on LinkedIn and X, providing lead management, pipeline tracking, and account connectivity. ### CRM Pipeline Stages Leads move through a visual pipeline. The typical progression is **New → Requested → Accepted → Contacted → Replied → Meeting → Won**, but **On Hold**, **Won**, and **Lost** can be reached from any stage at any time. | Stage | Description | | --------- | ----------------------------------- | | New | Freshly added to AutoReach | | On Hold | Temporarily paused | | Requested | Connection request sent | | Accepted | Connection request accepted | | Contacted | Ongoing conversation started | | Replied | Lead has engaged with your outreach | | Meeting | Meeting booked | | Won | Deal closed | | Lost | Disqualified or no longer pursuing | Leads auto-progress through stages as sequence activity occurs. You can manually drag leads between stages to override. ### Adding Leads On both LinkedIn and X profile pages, the extension injects an **"Add to Leads"** button into the profile action row. Lead addition is manual: you identify prospects and add them via the extension panel. The extension is a **CRM tool**, not an automated lead generator. It does not scan feeds or match ICPs automatically. ### Account Connection The extension handles connecting your LinkedIn and X accounts to AutoReach. It connects your active browser sessions and detects the platform automatically based on the current page. ### LinkedIn Connection Tracking When you send a connection request through AutoReach, the extension tracks the connection status and supports auto-withdrawal. The lead moves to the Requested stage automatically. ## Next Steps * [**AI Response Engine**](/ai-and-conversations/ai-response-engine): How AutoReach handles conversation follow-ups * [**Chrome Extension Setup**](/getting-started/chrome-extension): Install and configure the extension # AI Model Configuration AutoReach runs AI across many tasks (finding and scoring leads, writing and replying to messages, enrichment, and more). How that AI is powered, and whether you can choose models, depends on your billing mode. ## How AI is billed You have two options: * **Included credits (default).** AI runs on AutoReach's models, billed in credits. You start with **free credits** and can top up anytime from the **Credits** page. There are no API keys to configure. On credits, each task runs on a tuned default model, so model selection is managed for you. * **Bring your own key (optional).** Add your own **OpenAI, Anthropic, or DeepSeek** key in **Settings > AI & Models** for unlimited usage billed directly by your provider with no markup. With your own key, you can also choose the primary and fallback model for each category below. > Email finding uses your own **Findymail** API key, which is separate from the above and only needed for email discovery. ## Model Categories AutoReach uses AI across multiple categories. Each has independent primary and fallback model settings: | Category | Purpose | | ------------------ | ----------------------------------------------------------------- | | Content Writing | DMs, replies, first messages, templates, warmup, copilot | | Buyer Scoring | Fit/intent/timing analysis and lead qualification | | Pain Inference | Infers each lead's pain from their full profile before scoring | | Keyword Generation | X + LinkedIn search keyword generation | | Classification | Objection detection, negative content, deal signals, ICP matching | | Profile Finding | Email, LinkedIn, X profile, and website discovery | | Web Enrichment | Company research, news, social profiles, insights | | Lookalike Search | Finding influencers similar to target profiles | | Research & Copilot | Follow-up research, copilot web search, topic research | | Lead Relevance | Post and tweet relevance analysis using reasoning | | Call Briefs | Pre-call meeting preparation documents | ## Per-User Overrides > Choosing models per category requires your own API key (the bring-your-own-key option above). On included credits, every category runs on a tuned default and the model picker is not used. The cost guidance in the rest of this page applies to bring-your-own-key usage, where you pay your provider directly; on credits you are billed in credits and see your usage on the **Credits** page. Each category can be customized with a primary and fallback model selection. The system validates your choices against available models and checks that the required API key is present. Invalid or missing configurations fall back to defaults. ## How Fallbacks Work If the primary model fails (timeout, rate limit, API error), AutoReach automatically switches to the fallback model for that operation. This provides resilience without manual intervention. You do not need to do anything when a fallback occurs- the system handles it automatically and the output is still delivered. The only visible effect is that your cost reflects the blended rate of both models rather than just the primary. If both the primary and fallback models fail (rare, but possible during provider outages), the operation is retried later. ## When to Change Model Settings **Good reasons to change models:** * You want to reduce costs by using a cheaper model for high-volume categories like scoring or classification * You want higher quality DMs or posts and are willing to pay more for a premium content writing model * You have API keys for only one provider and need all categories to use that provider's models **When to leave defaults alone:** * If you are just getting started, the defaults are a solid starting point * If your current output quality and costs are acceptable, there is no need to experiment * If you are unsure which models to use, the defaults are well-tested ## Cost Implications Model pricing varies significantly. As a rough guide: * **Nano-class models** are the cheapest option - ideal for scoring, classification, and keyword generation where volume is high * **Mid-tier models** (Sonnet-class) offer a good balance and work well as fallback models * **Premium models** (Opus-class) deliver the highest quality but cost considerably more per token - reserve these for content writing if budget is a concern Since scoring and classification run on every lead while content writing only runs when sending messages, moving scoring to a cheaper model produces larger savings than optimizing content writing. ## Recommended Configurations **Budget-conscious setup:** Use nano models for scoring, classification, and keyword generation. Use a mid-tier model for content writing. This minimizes per-lead costs while keeping message quality reasonable. **Quality-focused setup:** Use a premium model for content writing (DMs, posts, replies). Keep scoring and classification on nano or mid-tier models - these tasks do not benefit as much from premium models. This gives you the best message quality without overspending on scoring. ## Settings UI The AI Configuration section in Settings shows all categories with primary and fallback dropdowns. Changes take effect immediately and are validated on save. ## Dynamic Cost Impact Your model selections directly affect pipeline costs. The cost estimation system reads your actual model configuration and calculates blended pricing based on observed fallback rates. See [**Cost Estimation**](/settings-and-configuration/cost-estimation) for details. ## Next Steps * [**Pipeline Cost Estimation**](/settings-and-configuration/cost-estimation): How model choices affect per-lead costs # Multi-Account Management Manage your X, LinkedIn, and Instagram accounts from a single AutoReach dashboard. Each account has its own limits, activity window, and health status, and is assigned a proxy from your proxy pool (proxies can be shared across accounts or dedicated, see [Proxies](#proxy-configuration) below). ## Account Slots One subscription includes **one account slot per platform**: 1 LinkedIn, 1 X, and 1 Instagram account. Offers and sequences are unlimited. To connect more than one account on the same platform, contact support at . There is no self-serve add-on for extra accounts. Each connected social account can also attach up to **2 email mailboxes** (one Gmail and one Outlook) for email outreach. See [Email Channel](/outreach-and-sequences/email-channel). ## Account Status Each account shows one of these statuses: | Status | Description | | --------- | ----------------------------------------------- | | Active | Operating normally | | Paused | Paused - no actions scheduled | | Suspended | Suspended due to safety issue - review required | | Expired | Authentication expired - re-connect required | ## X Accounts ### Authentication X accounts connect through the AutoReach Chrome Extension, which links your active browser session to AutoReach. ### Configuration * **Daily action limits**- configurable per account * **Activity window**- time frame when actions execute (default is two blocks, 09:00-13:00 and 14:00-18:00, Mon-Fri, weekends off, max 8h/day total), per-account, so each X account can have its own start/end/timezone. Set on the account's detail page under the Configuration tab. * **Calendar**- per-account booking link and webhook setup (Calendly or Cal.com) * **Browser fingerprint**- per-account OS spoofing (Auto/macOS/Windows/Linux) * **Proxy**- the proxy assigned to the account from your proxy pool (see [Proxy Configuration](#proxy-configuration)) ## LinkedIn Accounts ### Authentication LinkedIn accounts connect through the Chrome Extension, same as X. ### Configuration Each LinkedIn account exposes the same per-account configuration as X (activity window, calendar, browser fingerprint, assigned proxy) from **Accounts → \[account] → Configuration**. ### Connection Limits LinkedIn connection requests have a daily limit tracked per-account. You can set the daily connection limit on each LinkedIn account to any whole number from 1 to 100 (default 15, which is the recommended starting point). See [Supported Actions](/outreach-and-sequences/supported-actions#connection-request) for deferral behavior. ## Instagram Accounts Instagram accounts connect through the Chrome Extension and expose the same per-account configuration as X and LinkedIn (activity window, browser fingerprint, assigned proxy). ## Proxy Configuration Proxies are **optional** and managed in one place: the [**Proxies**](https://github.com/tiwanaca/autoreach-docs/tree/main/proxies/README.md) page. From there you can view your proxy pool and which accounts are assigned to each proxy, add a proxy, edit or remove a proxy, and assign a proxy to a specific account. One proxy can be **shared across all three platforms** (LinkedIn, X, and Instagram) or you can assign a dedicated proxy per account, your choice. You have two proxy options, chosen during onboarding: * **Managed proxy**- an IPRoyal ISP residential static proxy provided by AutoReach. Activation is included in the $20 signup fee, then it costs **$15/mo** per managed proxy after the trial if you keep it. By default one managed proxy is shared across your accounts. * **Bring your own (BYOP)**- use your own proxy at no extra cost. Enter the host, port, username, password, and type (HTTP or SOCKS5). AutoReach verifies connectivity before saving so you do not assign a dead proxy. If an account has no proxy assigned, it cannot run actions and is paused until you assign one. ### Antidetect Browsers If you run AutoReach inside an antidetect browser (SunBrowser, GoLogin, Multilogin, Dolphin Anty, Incogniton, Kameleo, AdsPower, and similar tools), you must configure your proxy inside the antidetect browser itself, not through the AutoReach extension. These browsers manage proxies at the browser launch layer and block extensions from calling Chrome's proxy API. If the extension tries to apply proxy settings, the browser kills the popup process before any error can be caught, which looks like the extension crashing or closing instantly. How AutoReach handles this: * The extension tries to auto-detect antidetect browsers from the user agent and skips proxy API calls when it matches * Detection is best-effort. Most antidetect browsers spoof their user agent to look like vanilla Chrome, which is the whole point of using one, so auto-detection will often miss them on first launch * Your proxy still works normally because the antidetect browser routes traffic through it at the network layer What you need to do before opening the extension on an antidetect browser: 1. Configure the proxy inside your antidetect browser profile first (this is the default workflow for these tools) 2. Do not configure a separate proxy in the extension popup. Leave proxy settings empty or set to "browser-managed" 3. If the popup closes instantly the first time you open it, that is the proxy API being called before detection kicks in. Set `__skip_proxy_api` to `true` in the extension's `chrome.storage.local`, then reopen the popup If you hit this crash, contact support with your browser name and version so we can add it to the detection keyword list and prevent it for other users. ## Pausing and Resuming **Pause**: All pending actions are cancelled and no new actions are scheduled. Associated warmup strategies and sequences are automatically paused. **Resume**: Account reconnects and operations resume. Previously paused warmup strategies and sequences are reactivated. ## Account Health Each account card on the Accounts page displays: * **Status badge**- the account's current state: Active, Sleeping (outside its activity window), Paused, Setup (no proxy assigned yet), Inactive, Expired, or Suspended * **API calls today**- how much of the account's daily activity allowance has been used so far today, shown as `used / total`. Every action AutoReach performs through the account consumes API calls: sending DMs and connection requests, enriching profiles, engagement actions (likes, comments, follows), and searches. When the allowance is used up, AutoReach automatically pauses activity on that account and resumes after the counter resets the next day. This is a safety mechanism that keeps the account's daily activity at a healthy level. A dash means the counter has not loaded yet. * **Proxy**- Configured (verified and working), Not verified, or Required (no proxy assigned) * **Last active**- when the account last performed an action * **Added**- when the account was connected Check the Accounts page regularly. Error indicators may signal proxy issues or authentication expiration. ## Deleting Accounts From an account's detail page, click **Delete Account** to permanently remove it. Deletion is irreversible and removes: * The account connection * All associated conversations, sequence enrollments, and warmup state for that account Deleting an account does **not** cancel or remove its proxy. Because a proxy can be shared across multiple accounts, proxies are managed separately on the [Proxies](https://github.com/tiwanaca/autoreach-docs/tree/main/proxies/README.md) page. If you want to stop paying for (or remove) a proxy after deleting an account, remove it there. Removing a managed proxy cancels its IPRoyal order and lowers your $15/mo proxy add-on accordingly. You will see a confirmation modal listing the consequences before the delete proceeds. Active campaigns using the account are affected. **Base account protection**- the primary account on your workspace cannot be deleted. If you need to remove it, contact support. A separate modal blocks the action and explains this. ## Next Steps * [**Account Safety**](/settings-and-configuration/account-safety): Error handling, cooldowns, and cascade pause behavior * [**Blacklisting**](/settings-and-configuration/blacklisting): Exclude specific accounts from outreach # Proxies AutoReach routes each connected account's traffic through a proxy so your activity comes from a stable residential IP rather than a datacenter IP or your own home connection. Proxies are optional and are managed from the **Proxies** page in the dashboard. ## Do I need a proxy? A proxy is optional but recommended, especially for LinkedIn and Instagram, which are sensitive to IP changes and datacenter IPs. You can run AutoReach without one, but a stable residential IP reduces the chance of session disruptions and verification challenges. You pick your proxy setup during onboarding, and you can change it at any time from the Proxies page. ## Two ways to get a proxy ### Managed proxy (we provide it) * An ISP residential static proxy provisioned through our provider. * Activation is included in the $20 signup fee, then each managed proxy is $15/month after the trial if you keep it. * The $15/month is added to your subscription as a separate line item, starting after the trial. * One managed proxy can be shared across your LinkedIn, X, and Instagram accounts, or you can buy more than one and assign them separately. ### Bring your own proxy (BYOP) * Free. Use a proxy you already have from any provider. * Enter the host, port, and (if required) username and password, then choose HTTP or SOCKS5. * AutoReach verifies the proxy can connect before saving it. If it cannot connect, you see an error and nothing is saved. ## The Proxies page Open **Proxies** from the sidebar. The page shows your proxy pool and which accounts are assigned to each proxy. From here you can: * **Add a bring-your-own proxy**: enter the connection details, which are verified before saving. * **Buy a managed proxy**: pick a location. Activation is included in the signup fee; it adds a $15/month line item to your subscription after the trial. * **Edit a bring-your-own proxy**: update the connection details (re-verified on save). * **Remove a proxy**: see [Removing a proxy](#removing-a-proxy) below. * **Assign a proxy to an account**: point any LinkedIn, X, or Instagram account at a specific proxy in your pool. ## Sharing vs dedicated One proxy can serve all three of your accounts (LinkedIn, X, and Instagram), or you can give an account its own dedicated proxy by adding more proxies and assigning them. A shared residential proxy is fine for a single operator running one account per platform. If you run heavier volume, a dedicated proxy per account adds isolation. ## Removing a proxy You can remove any proxy, including the default one. * Accounts running on the removed proxy are detached and paused, since an account cannot run safely without a proxy. Assign them another proxy to resume. * If you remove a managed proxy, the underlying order is cancelled and your $15/month proxy charge drops by one unit. Remove all managed proxies and the monthly proxy charge is removed entirely. * If you remove the default proxy, another proxy in your pool is automatically promoted to default so newly connected accounts still have one to use. ## Troubleshooting **"Proxy Connection Failed"**: open the Proxies page. For a bring-your-own proxy, re-check the host, port, and credentials and save again (this re-verifies it). For a managed proxy, contact support at . **An account is paused right after I removed its proxy**: this is expected. Assign the account a proxy from the Proxies page, then resume it. ## Related * [Connecting Your Accounts](/getting-started/connecting-accounts) * [Multi-Account Management](/settings-and-configuration/multi-account) * [Account Safety](/settings-and-configuration/account-safety) # Pipeline Cost Estimation AutoReach provides real-time cost estimates for your lead discovery and enrichment pipeline. Costs are dynamically calculated based on your AI model configuration. ## Dynamic Pricing Cost estimates reflect your **actual model selections**. When you change models in Settings, the estimate updates immediately. ### Blended Pricing Each category's cost is blended between your primary and fallback models based on observed fallback rates. This means your cost estimate reflects real-world usage rather than assuming the primary model handles every request. ## Cost Components The main cost drivers per lead: ### Extraction (Lead Discovery) * Keyword generation * Sentiment analysis ### Enrichment * **LinkedIn profile finding**- web search to find LinkedIn profiles for X-sourced leads * **Website finding**- web search for company websites * **Web enrichment**- deep company analysis from websites (most expensive per-lead operation) * **Email finding**- third-party API cost (your own Findymail key, no AutoReach markup) ### Deep AI Analysis Scoring costs depend on the amount of profile data available for each lead. Leads with more enrichment data (posts, company info, web research) cost more to score than leads with minimal profile information. ## Example Cost Scenarios The following examples give a rough sense of costs for common workflows. Actual costs depend on your model selections, lead data completeness, and enrichment settings. **Finding 100 leads from a tweet search (default models):** * Extraction (keyword generation + sentiment analysis): \~$0.01-0.03 * Enrichment (LinkedIn profile finding, website finding, web enrichment): \~$0.80-1.50 * Buyer scoring: \~$0.02-0.05 * **Estimated total: \~$0.85-1.60 for 100 leads** Web enrichment (deep company analysis from websites) is by far the most expensive single operation. If you disable it, enrichment costs drop significantly. **Scoring 500 existing leads (rescore only, no enrichment):** * Using the default efficient scoring model: \~$0.10-0.40 * Using a premium model for scoring: \~$0.50-2.00 * Dropping scoring to a nano model: \~$0.05-0.15 **LinkedIn content search with 50 posts, 3 intent categories:** * Similar extraction costs to tweet search * No LinkedIn profile finding needed (leads already have LinkedIn URLs) * **Estimated total: \~$0.15-0.60 for the discovered leads** ## How Model Selection Affects Costs Your choice of AI models has a dramatic impact on per-lead costs. Nano-class models (designed for classification and scoring) cost roughly 10-30x less per token than premium models (designed for content writing). Actual cost per lead depends on how many tokens each operation consumes- operations on richer profiles naturally use more tokens regardless of model tier. * **Nano models** (best for scoring, classification, keywords): Lowest cost, suitable for high-volume operations where speed matters more than nuance * **Mid-tier models** (Sonnet-class): Good balance of quality and cost, often used as fallback models * **Premium models** (Opus-class and equivalent): Highest quality output, best for content writing where tone and personalization matter, but significantly more expensive per token For budget-conscious users, the biggest savings come from keeping scoring and classification on nano models and reserving premium models only for content writing (DMs, posts, replies). ## Checking Your AI Spending You can review your actual AI usage and spending in the **AI Usage panel on the Dashboard**. This shows token consumption and costs broken down by category, so you can see exactly where your budget is going and identify opportunities to optimize. ## Reducing Costs 1. **Disable web enrichment**- the most expensive per-lead operation; skip it for lead sources where you already have company context 2. **Use lower-cost models for scoring and classification**- these high-volume categories process every lead, so even small per-token savings add up quickly 3. **Filter early**- apply keyword filters and exclusions before enrichment to reduce the number of leads entering the pipeline 4. **Choose efficient fallback models**- an expensive fallback model increases your effective rate 5. **Run smaller batches first**- test with 20-50 leads before running large searches to validate your keywords and filters 6. **Review the cost estimate before running**- the estimate updates in real time as you adjust search parameters, so use it to experiment before committing ## Next Steps * [**AI Model Configuration**](/settings-and-configuration/ai-models): Customize model selections to optimize cost # Account Safety AutoReach protects your accounts using browser identity management, rate limiting, health monitoring, and human behavior simulation. ## How to Stay Safe (the short version) Most account issues come from a small set of avoidable mistakes. Three rules cover the majority of them: 1. **Start with low daily limits and ramp slowly.** Begin at 15-20 actions per day per account, then increase gradually over weeks, not days. Sudden jumps to 50+ on a fresh account are the #1 cause of bot detection. 2. **Run the** [**Engagement Engine**](/engagement-engine/overview) **for 1-2 weeks before launching sequences on new or inactive accounts.** Outreach from a silent account looks suspicious to the platform. 3. **Don't run AutoReach and manual mass-actions at the same time.** If you're manually liking 50 posts while AutoReach is also active, the combined activity can trigger rate limits. The rest of this page covers what AutoReach does automatically when something goes wrong, and what requires your intervention. **You don't need to read it preemptively**, come back when you see an error status on an account or get a notification email. ## Account Health Monitoring AutoReach continuously monitors account health and classifies errors into actionable categories. ### Error Classification AutoReach classifies errors into categories- some are auto-recoverable (like rate limits and timeouts), while others require manual review (like bot detection or captcha challenges). Each category has an appropriate cooldown period before retrying. Errors that are auto-recoverable resolve on their own after the cooldown (rate limits, timeouts, expired auth, proxy errors, IP blocks). Bot detection, captcha challenges, and AI-provider credit exhaustion are flagged as manual-fix-required - the account stays paused until you address the root cause and resume it from the Accounts page. ### Email Notifications You receive email alerts for significant account issues. Transient errors (like occasional rate limits or timeouts) do not generate emails- only persistent or serious problems trigger notifications. ### Cascade Pause When a serious error occurs, AutoReach pauses not just the account but all associated activity- warmup, sequences, and pending actions. This prevents further issues while you investigate. After a cascade pause, manual resume is required. ## Emergency Pause AutoReach automatically pauses an account when it detects severe issues (bot detection, captcha challenges, IP blocks). When emergency pause activates: 1. All pending actions are cancelled 2. Account shows paused status 3. You receive a notification (for bot detection, captcha, and AI-provider credit issues - which require manual review) 4. **Manual resume** is required for bot detection, captcha, and AI-provider issues. **Auto-resume** happens for IP blocks, expired auth, and proxy errors once their cooldown expires ## Browser Identity Management Each connected account is assigned a consistent browser identity that persists over time, making activity appear as normal browser use rather than automated traffic. ## Human Behavior Simulation All actions include realistic timing and pacing designed to mirror normal human browsing patterns rather than automated scripts. ## Negative Content Screening Before engaging with a lead's post, content is checked for sensitive topics: **Skipped topics**: Death, obituaries, serious illness, disasters, violence, self-harm **Allowed topics**: Work frustrations, business failures, layoffs, industry criticism, competitive pressure If the content check fails, the action proceeds rather than blocking. ## Best Practices 1. **Use a quality proxy**- proxies are optional and managed on the [Proxies](https://github.com/tiwanaca/autoreach-docs/blob/main/proxies/README.md) page. One proxy can be shared across your LinkedIn, X, and Instagram accounts, or you can assign a dedicated proxy per account. A dedicated proxy per account isolates risk best, but a shared residential proxy is fully supported 2. **Keep limits conservative**- especially on new accounts 3. **Monitor the Accounts page**- check error rates regularly 4. **Warm up new accounts**- use the Engagement Engine before launching outreach 5. **Pause proactively**- if you notice unusual behavior, pause before AutoReach triggers an emergency pause ## Next Steps * [**Multi-Account Management**](/settings-and-configuration/multi-account): Manage accounts, proxies, and limits * [**Engagement Engine**](/engagement-engine/overview): Warm up accounts safely # Account Signals & Interaction Orbit AutoReach tracks two layers of buying behavior beyond what leads explicitly post about: **Interaction Orbit** (individual engagement patterns) and **Account-Level Signal Aggregation** (company-wide purchase intent). *** ## Interaction Orbit (Dark Funnel) Interaction Orbit detects buying behavior revealed through who leads engage with on social media. ### How It Works AutoReach tracks **who each lead replies to** on social media. When a lead replies to posts from specific accounts, those accounts become "orbit targets." The system builds a map of engagement targets per platform, tracking interaction count and first/last seen dates. ### Target Classification Each orbit target is classified into categories like **Competitor**, **Adjacent Vendor**, **Thought Leader**, or **Peer** using your offer's competitor list and AI analysis. ### Cluster Detection Orbit targets are grouped into **clusters** - collections of related accounts a lead engages with together within a recent time window. Example: A prospect replies to 3 data warehouse companies and 2 BI tool accounts in a short period, forming a "data stack evaluation" cluster. When a new cluster forms, it generates an orbit signal that feeds into buyer scoring. Competitor clusters are the strongest signal, as they indicate active evaluation of solutions in your space. ### Lead Profile Integration When orbit data is available, the lead profile shows: * **Orbit targets** - accounts the lead engages with * **Clusters detected** - groups of related engagement targets * **Cluster velocity** - engagement frequency (interactions per week) ### Privacy Interaction Orbit only tracks **public engagement** - public replies, likes, and follows. Private DMs and conversations are never tracked. This is equivalent to what you would manually discover by viewing someone's social activity. *** ## Account-Level Signal Aggregation Account-Level Signal Aggregation combines buying signals from multiple leads at the same company into a single **heat score** - an indicator of company-wide purchase intent. ### How Heat Score Works The heat score reflects the overall buying intent at a company. It factors in: * **Signal strength**: Stronger signals (like asking for recommendations) contribute more than weaker ones (like a single post like) * **Recency**: Recent signals carry more weight than older ones * **Team breadth**: Signals from multiple people at the same company are a stronger indicator than signals from just one person ### Company Identification Companies are identified by LinkedIn company ID (preferred) or normalized company name. Generic values like self-employed, freelance, or consultant are filtered out. Only companies with **2 or more leads** are included in heat scoring. ### Signal Types The following signal types are aggregated across all leads at a company: | Signal | Description | | -------------------------- | -------------------------------------------- | | Competitor Engagement | Engaging with competitor accounts | | Engagement Pattern | Category research patterns | | Orbit Cluster | Dark funnel cluster detection | | Hiring | Company hiring signals | | Tool Mention | Mentions of tools or technologies | | Switching | Signals of switching or evaluation | | Alternative Search | Searching for alternatives | | Asked Recommendation | Asked network for recommendations | | Funding | Funding rounds | | Pain Match | Matches a pain point your solution addresses | | Product Launch | Product launches | | Mergers and Acquisitions | M\&A activity | | Geographic Expansion | Geographic expansion | | Cost Cutting | Budget pressure signals | | IPO Filing | IPO filings | | Complained | Complained about a problem you solve | | Custom Intent | Custom intent signal from your ICP | | Own Post Engagement | Engaged with your posts | | Own Post Reply | Replied to your posts | | Own Post Repeat Engagement | Engaged with your posts multiple times | ### Heat Categories | Category | Description | Action | | -------- | ------------------------------------------------- | ---------------------------- | | **Hot** | Strong, multi-threaded signals across the company | Urgent, coordinated outreach | | **Warm** | Consistent signals from diverse team members | Priority outreach | | **Cool** | Light or isolated signals | Standard cadence | ### Heat Trend | Trend | Meaning | | ------- | ----------------------------- | | Rising | Score increasing day-over-day | | Stable | Score flat | | Cooling | Score decreasing | ## Next Steps * [**Buyer Intelligence**](/core-concepts/buyer-intelligence): How individual lead scoring works * [**Analytics**](/settings-and-configuration/analytics): Track overall pipeline performance # Blacklisting Accounts Blacklist specific accounts to prevent AutoReach from discovering, enriching, or engaging with them. Useful for excluding competitors, existing customers, and contacts who have requested not to be contacted. ## How Blacklisting Works Blacklisted accounts are matched by username (case-insensitive) across platforms. When you blacklist an account: 1. The account is added to the blacklist 2. All matching leads are found (case-insensitive username match) 3. All sequence enrollments for those leads are identified 4. All pending, queued, and deferred actions for those leads are **cancelled** 5. The leads are **removed** from all sequences The response includes a count of removed sequence leads so you can see the impact. ## Adding to Blacklist ### Single Account On the **Leads page**, click the menu on any lead and select **Blacklist**. Enter the username to blacklist. ### Bulk Blacklist On the **Leads page**, select multiple leads in the table using the checkboxes. A floating selection bar appears at the bottom- click **Blacklist** to blacklist all selected leads at once. The same cascade removal logic applies to each. ## What Blacklisting Prevents * **Discovery**: Blacklisted accounts are excluded from search results * **Enrichment**: Skipped during profile enrichment * **Engagement**: Never receive messages, likes, follows, or comments * **All sequences**: Applied globally across all sequences and offers ## Removing from Blacklist Find the account in the blacklist and click **Remove**. The account becomes discoverable and engageable again in future runs. Removing from the blacklist does not restore previously deleted sequence leads or actions. ## Common Use Cases * **Competitors**- avoid wasting enrichment costs on accounts you'll never contact * **Existing customers**- prevent duplicate outreach * **Do-not-contact requests**- comply with prospects who asked to stop receiving messages * **Internal contacts**- exclude your own team, CEO, or board members ## Next Steps * [**Account Safety**](/settings-and-configuration/account-safety): How AutoReach protects your accounts * [**Multi-Account Management**](/settings-and-configuration/multi-account): Managing accounts and their settings # Webhooks Webhooks let you receive real-time notifications when important events happen in AutoReach. Instead of checking your dashboard manually, AutoReach sends an HTTP request to a URL you specify whenever a lead is scored, a reply comes in, a message is sent, and more. This is useful for connecting AutoReach to your CRM, Slack workspace, internal tools, or any custom workflow that can receive HTTP requests. *** ## Available Events Each webhook subscribes to **one event**. To receive notifications for multiple events, create a separate webhook for each. The following events are available: | Event | Description | | ------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | | **Lead Created** | A new lead has been added to your account (from any source) | | **Lead Scored** | A lead has been scored by the buyer intelligence engine | | **Reply Received** | A lead replied to your DM or message | | **Message Sent** | An outreach message (DM) was sent to a lead | | **AI Response Sent** | The AI auto-responder sent a follow-up message in a conversation | | **Meeting Booked** | A lead booked a meeting through your calendar link | | **Booking Link Detected** | A lead shared their own scheduling link in a reply (AI auto-reply is turned off for that thread so a human takes the booking) | | **Enrichment Completed** | A lead's profile enrichment has finished | *** ## Creating a Webhook 1. Go to **Settings > Webhooks** 2. Click **Add Webhook** 3. Enter a **name** for the webhook (e.g., "CRM Sync" or "Slack Alerts") 4. Enter the **URL** where you want to receive events (must use HTTPS) 5. Select which **event** you want to subscribe to 6. (Optional) Enter a **signing secret** for payload verification (minimum 16 characters) 7. Click **Create** You can create up to **10 webhooks** per account. > **Tip:** Start with just one or two events to verify your integration is working before subscribing to everything. *** ## Testing a Webhook Before relying on a webhook in production, test it to make sure your endpoint is reachable and responding correctly. 1. Go to **Settings > Webhooks** 2. Click the edit icon on the webhook you want to test 3. In the edit dialog, select an event type and click **Send Test Payload** 4. AutoReach sends a sample payload to your URL and shows whether it succeeded or failed *** ## Editing and Deleting Webhooks To **edit** a webhook, click the edit icon next to it and update any fields (URL, events, name, secret, or enabled/disabled status). Click **Save** to apply changes. To **delete** a webhook, click the delete button next to it. Deleted webhooks stop receiving events immediately. You can also **disable** a webhook without deleting it. Disabled webhooks are saved but do not receive any events until re-enabled. *** ## Payload Format Every webhook delivery is a JSON payload with three fields: * **Event name** - which event triggered this delivery * **Timestamp** - when the event occurred (ISO 8601 format) * **Data** - the event-specific information ### What data is included The data varies by event type, but for lead-related events you can expect: * **Lead identity**: name, username, bio, headline, location, email (if found), profile links for X and LinkedIn * **Scoring**: overall buyer score, fit/intent/timing breakdown, buyer state, suggested outreach strategy, recommended channel * **Source**: how the lead was discovered, the original content or post that surfaced them * **Enrichment**: professional summary, work experience, education, skills, certifications, languages, company details, and recent posts * **Web intelligence**: company website data and additional research findings For message and reply events, the payload includes the message content and conversation context. *** ## Verifying Webhook Signatures If you set a signing secret when creating your webhook, every delivery includes a signature header (`X-Webhook-Signature`) that you can use to verify the payload came from AutoReach and was not tampered with. The signature is an HMAC-SHA256 hash of the raw JSON payload, prefixed with `sha256=`. To verify: 1. Read the raw request body (before parsing JSON) 2. Compute HMAC-SHA256 using your signing secret as the key 3. Compare your computed signature with the value in the `X-Webhook-Signature` header > **Warning:** Always use a constant-time comparison function when verifying signatures to prevent timing attacks. *** ## Retry Behavior If your endpoint is unreachable or returns an error, AutoReach retries the delivery up to **3 times** with increasing delays between attempts. If all retries fail, the webhook shows an error status in the dashboard with the failure reason. You can check the last delivery status and any errors for each webhook in the webhooks list. *** ## Use Cases ### CRM Integration Subscribe to **Lead Scored** and **Meeting Booked** events to automatically create or update contacts in your CRM (Salesforce, HubSpot, Pipedrive, etc.) whenever AutoReach identifies a qualified buyer or books a meeting. ### Slack Notifications Subscribe to **Reply Received** and **Meeting Booked** events to get instant Slack alerts when a prospect responds or schedules a call. Use a Slack incoming webhook URL as your endpoint, or route through Zapier/Make. ### Custom Workflows Subscribe to **Enrichment Completed** to trigger your own analysis pipeline when a lead's profile data is ready. Or use **Message Sent** to log outreach activity in a custom reporting system. ### Lead Routing Subscribe to **Lead Created** and route new leads to different team members or tools based on the lead's location, industry, or score. *** ## Next Steps * [**AI Model Configuration**](/settings-and-configuration/ai-models): Customize which AI models power your scoring and messaging * [**Account Safety**](/settings-and-configuration/account-safety): Configure rate limits and safety settings for your connected accounts * [**Cost Estimation**](/settings-and-configuration/cost-estimation): Understand and estimate your pipeline costs # Analytics & Reporting AutoReach provides a comprehensive analytics suite to help you understand your outreach performance, track AI spending, and monitor account health. All analytics are accessible from the dashboard and the Analytics section of the app. *** ## Dashboard Overview The main dashboard shows your key pipeline metrics at a glance: | Metric | What It Measures | | ---------------------- | ----------------------------------------------------------- | | **Pipeline Created** | Total leads in your account (across all sources) | | **Buyers Identified** | Leads scored as qualified buyers by the intelligence engine | | **Outreach Delivered** | Leads who have been contacted | | **Meetings Generated** | Total meetings booked through your calendar link | A **potential pipeline** value is displayed in the hero section, showing estimated revenue from qualified leads based on your offer's deal size. These numbers update in real time as your sequences run and leads progress through the pipeline. *** ## Daily Summary The daily summary gives you a snapshot of today's activity across all your sequences and the Engagement Engine: * **DMs sent** - how many direct messages went out today * **Comments/replies posted** - engagement actions on lead posts * **Likes** - posts liked as part of sequences or warmup * **Follows** - new accounts followed (X and Instagram) * **Connection requests** - LinkedIn connection requests sent * **Total actions** - combined count of all actions today This summary appears as a **toast notification** when you open the app, giving you a quick snapshot of today's activity. *** ## Activity Feed The activity feed is a chronological log of every completed action across your account. It shows: * **What happened** - the action type (like, DM, comment, follow, connection request) * **Who it was for** - the lead's name and username * **Which sequence** - the sequence that triggered the action (or "Engagement Engine" for warmup actions) * **Platform** - whether it happened on X, LinkedIn, or Instagram * **When** - the exact timestamp * **Preview** - a snippet of the message sent or the post engaged with The feed combines both sequence actions and Engagement Engine activity into a single timeline, sorted newest first. Use it to review what AutoReach has been doing on your behalf and spot-check message quality. *** ## Sequence Performance Each sequence has its own performance view showing: * **Lead status breakdown** - how many leads are at each stage (pending, active, contacted, replied, meeting booked, completed) * **Step-level analytics** - completion counts for each step in your flow (e.g., how many likes sent, how many DMs delivered) * **Reply rate** - percentage of contacted leads who responded * **Meeting rate** - percentage of contacted leads who booked a meeting * **Conversion funnel** - visual breakdown from enrolled leads through to meetings Access sequence performance by clicking on any sequence in the Sequences list. *** ## AI Usage Tracking AutoReach tracks every AI call made on your behalf so you can monitor token consumption and costs. ### Usage Summary The AI Usage panel on the Dashboard shows aggregate stats for a configurable time period (up to 90 days): * **Total tokens used** - prompt tokens (input) and completion tokens (output) * **Total cost** - estimated cost in USD based on model pricing * **Breakdown by action** - which activities consumed the most tokens (scoring, DM generation, replies, classification, etc.) * **Breakdown by model** - cost split across different AI models (GPT, Claude, etc.) * **Daily trend** - a day-by-day chart of token usage and cost ### Recent Activity Log Below the summary, a scrollable log shows individual AI calls with: * The action that triggered it (e.g., "DM generation", "buyer scoring", "keyword generation") * Which model was used * Token counts (prompt and completion) * Cost for that specific call * Timestamp This helps you identify which activities are driving your AI costs and whether any particular action is unusually expensive. *** ## AI Health Monitoring AutoReach monitors the health of your AI providers in real time. If both your primary and fallback AI models fail (for example, due to exhausted API credits or a provider outage), a warning banner appears on your dashboard. The warning includes: * **What happened** - whether it is a quota issue or a provider error * **Which provider** - the affected AI service * **When** - when the issue was first detected You can dismiss the warning once you have resolved the issue (e.g., added credits to your API account). > **Tip:** Enable auto-recharge on both your OpenAI and Anthropic accounts to prevent quota exhaustion from interrupting your outreach. *** ## Account Health Monitoring A health banner on the Dashboard alerts you when a connected account has issues. It shows the account's status and the reason for any problems (rate limit, bot detection, auth error, etc.). A healthy account shows no banner. *** ## LinkedIn Pipeline Analytics If you use the Chrome Extension CRM pipeline for LinkedIn, additional analytics are available: * **Funnel stages** - leads broken down by stage (new, requested, accepted, contacted, replied, meeting, won, lost) * **Needs follow-up** - how many leads are overdue for a follow-up message * **Reply rate** - percentage of contacted leads who replied * **Meeting rate** - percentage of contacts that converted to meetings *** ## Next Steps * [**Cost Estimation**](/settings-and-configuration/cost-estimation): Estimate costs before running a pipeline * [**AI Model Configuration**](/settings-and-configuration/ai-models): Change which models are used to optimize cost vs. quality * [**Account Safety**](/settings-and-configuration/account-safety): Adjust rate limits and safety settings