> For the complete documentation index, see [llms.txt](https://doc.autoreach.tech/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.autoreach.tech/ai-and-conversations/ai-response-engine.md). # 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.md): Customize the AI's voice and provide sales context * [**Conversation Analyzer**](/ai-and-conversations/conversation-analyzer.md): AI suggestions for improving your sequences --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://doc.autoreach.tech/ai-and-conversations/ai-response-engine.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.