An autonomous AI system that reads inbound messages, handles objections and books sales calls for you.
A fully autonomous deal-closing system that responds to inbound leads instantly, generates personalised replies, handles objections, qualifies prospects, books calls automatically and keeps your CRM updated without human input. Users get faster response times, higher conversion rates and a sales pipeline that advances itself with minimal manual work.
Set up an Attio workspace with a pipeline such as New lead, Responded, Qualified, Call booked, Won and Lost and add fields for source, channel, last_message, intent, lead_score, objection_type, next_action, owner and ai_reply_overridden.
Connect your main inbound channels to n8n, such as a sales inbox via IMAP or API and any lead forms you use, and create a workflow that triggers whenever a new inbound message or form submission is received.
In that n8n workflow, normalise the incoming data so each new inquiry has a clean payload that includes name, email, channel, raw_message, product_of_interest and any UTM or source data.
Add an n8n node that sends the raw_message and basic context such as channel, product and source into Relevance AI so it can classify intent, detect sentiment, flag spam and extract key entities like budget hints, timeline and use case.
Configure Relevance AI to return a structured JSON object with fields such as intent_label, sentiment, spam_flag, problem_statement, urgency_level and a suggested lead_score.
Add a simple filter in n8n so that only leads above a minimum lead_score or from priority sources go through the full AI reasoning flow while low value or obvious spam records are marked as Ignored or Closed in Attio.
Pass the original message, the Relevance AI labels and any existing Attio contact data into Claude via n8n and ask it to decide the current buying stage, identify the main objection or question, propose the best next action and generate a full reply in your brand tone.
In the same Claude call, request that it output a structured block with fields such as buying_stage, objection_type, qualification_notes, suggested_reply, include_booking_link flag and an internal_reasoning_note.
Map Claude’s structured output back into fields and use an Attio node in n8n to create or update the contact, logging the latest message, buying_stage, lead_score, objection_type, suggested_reply, next_action and internal_reasoning_note.
Add a decision step in n8n so that if the message is spam or very low intent, the workflow simply updates Attio with a Closed or Ignored status and does not send a reply.
For valid leads, use n8n to send Claude’s suggested_reply back through the original channel, such as via your email provider API, and automatically update the Attio stage to Responded or Qualified depending on Claude’s buying_stage output.
For cases where Claude recommends booking a call, configure n8n to insert your Calendly or scheduling link into the reply and when a call is booked, capture the event via webhook, attach the booking details to the Attio contact and move the stage to Call booked.
Add basic error handling branches in n8n around Relevance AI and Claude nodes so that if a call fails or times out, the workflow logs the failure to an Attio activity field and optionally posts an alert into Slack for manual follow up.
Update your Attio layout to include fields for first_response_time, total_ai_replies, ai_reply_overridden, and reason_for_loss so you can track how often the agent is accurate and where humans step in.
Create a separate n8n workflow that runs daily, pulls all open deals from Attio that are stuck in mid funnel stages, sends a batch summary into Claude to get prioritisation and messaging tweaks and writes the updated next_action and notes back into Attio so you always know which deals to push forward next.
Before going fully live, duplicate the workflow into a test pipeline in Attio, run a set of sample leads through it including multilingual messages and complex objections, and confirm that every branch from spam to high intent behaves correctly and sends safe, on brand replies.
Build and deploy custom AI agents to automate business workflows.
Meet your AI thinking partner for any complex task.
CRM re-engineered for high-velocity revenue teams.
Build custom workflows with code-level power and no-code speed.
This blueprint builds an autonomous deal-closing system that reads every inbound message from your sales channels, analyses intent, generates personalised replies, handles objections and books calls automatically. Instead of manually checking emails or DMs, the agent takes over the entire front end of your sales pipeline and pushes only high-intent or complex conversations to you.
The system begins with Attio, which acts as the central CRM and memory layer. You set up a pipeline that mirrors a real sales process and add fields for intent, objection, next action, lead score and AI reply data. Every time a lead interacts with you, the AI writes back to Attio so your CRM becomes a real-time reflection of your entire funnel.
n8n is the automation backbone. It listens to your inbound channels such as your sales inbox or lead forms, captures every new inquiry and standardises the message payload. n8n ensures every message, no matter the source, enters the workflow cleanly with consistent fields so AI reasoning stays accurate. This prevents broken logic from messy or inconsistent inputs.
Relevance AI is the classification engine. Each raw message is passed into Relevance AI so it can extract intent, sentiment, timeline clues, budget hints and emotional tone. It also flags spam or low quality messages. This step transforms an unstructured message into structured data your agent can rely on. You also apply a threshold so only leads with a minimum lead score or from priority channels go through the full reasoning flow, which keeps costs under control for high-volume inbound.
Claude is the reasoning and reply engine. It receives the original message, the Relevance AI signals and any past interaction history from Attio. Claude decides which buying stage the lead is in, what the core objection or question is and what the ideal next action should be. It then writes a personalised reply that matches your voice and aligns with the buyer’s stage, context and emotional tone. This transforms Claude from a text generator into a real agent performing multi step decision making.
Claude also outputs structured data, including the buying stage, objection type, internal qualification notes and whether the reply should include a scheduling link. n8n maps these outputs into Attio fields so every change to lead status is recorded. This builds a complete audit trail of how the agent thinks and what decisions it makes.
Before sending a reply, the workflow checks for spam or low intent. If the lead is clearly not a fit, n8n updates Attio with a Closed or Ignored status and avoids sending any response. This prevents the system from wasting cycles on noise or engaging with low quality traffic.
If the lead is valid, n8n sends the reply through the original channel and updates the lead’s stage accordingly. When Claude suggests a call booking link, n8n inserts your scheduling link automatically. When a prospect books a call, the webhook triggers n8n to update Attio with the booking details and move the lead to Call booked.
Error handling is built in. n8n catches failures from Claude or Relevance AI and logs them in Attio or optionally in Slack so you can see when a message did not process correctly. This gives you visibility into failures without breaking the entire pipeline.
To improve performance over time, Attio includes fields for first response time, AI replies sent, replies overridden by a human and reasons for lost deals. This lets you track accuracy and identify where the agent struggles. The blueprint also includes a daily workflow that pulls all active deals into Claude, which analyses stuck leads and suggests new angles, follow ups or prioritisation so you always know which opportunities to push forward next.
Before going live, you run a full simulation with a test Attio pipeline. You test multilingual messages, long emails, random spam, complex objections, emotional messages and high intent leads. This ensures every branch behaves correctly and replies stay safe and on brand.
The result is a true agentic closer that does more than send templated replies. It understands messages, thinks through them, decides what to say, sends the reply, updates your CRM, monitors stuck deals and escalates only when necessary. You eliminate the need for manual triage, your inbound gets answered instantly and your pipeline moves faster with less human input.
Share it with your network