Automate outbound prospecting with an AI SDR that researches leads, writes outreach, follows up and updates your CRM on its own.
A fully autonomous AI SDR that researches prospects, generates personalised outreach, follows up intelligently, interprets replies and updates your CRM without manual effort. Users gain consistent outbound volume, higher quality messaging, faster response handling and a predictable flow of qualified conversations, all powered by an AI driven system that operates continuously in the background.
Set up an Attio workspace with a pipeline called Outbound SDR and create fields for lead source, ICP fit score, last outreach step, last contact date, objection notes, next action and owner.
Import a CSV of leads or connect your existing lead source so that new prospects are added into Attio with basic fields like name, company, role, email, website and LinkedIn URL.
In n8n, create a workflow that triggers whenever a new lead is created in Attio with status set to New or when a lead is moved into a Ready for outreach stage.
Add an n8n node that sends the lead’s company domain, role and any public description or notes into Relevance AI so it can enrich the lead with website insights, industry, size, tech stack clues and potential pain signals (optional enhancement: swap or augment this step with Clay for deeper enrichment and social signals).
Configure Relevance AI to return a structured JSON object that includes ICP fit indicators, key pain themes, relevance tags and a recommended priority score for that lead.
Pass the enriched lead data and ICP signals from Relevance AI into Claude via n8n and ask it to decide the best outreach angle, write a personalised first line, draft a full cold email and propose a follow up sequence tailored to that prospect.
Map Claude’s output back into Attio fields such as first_line, email_step_1, email_step_2, email_step_3, objection_script and next_action so the system has a stored playbook for that contact.
Use n8n to send the first email from your outbound inbox (for example via SMTP or your email provider) using the personalised first line and main email body that Claude generated and then update the lead status in Attio to Step 1 sent (optional enhancement: replace raw SMTP here with Instantly or Lemlist for stronger deliverability, warmup and inbox rotation).
Add time delay and conditional nodes in n8n so that if there is no reply after a set number of days, the workflow automatically sends the next follow up email written by Claude and increments the outreach step in Attio (optional enhancement: run the follow up sequence from Instantly or Lemlist while still syncing states back to Attio).
Configure n8n to watch your inbox for replies and when a response arrives, send the thread into Claude so it can classify the reply as positive, neutral, objection or not interested and suggest an appropriate response or next action.
Update the Attio record with Claude’s reply classification, suggested response and new next_action, and optionally let n8n send the reply directly if the classification is a simple positive or basic objection that does not need human review.
Set up a scheduled n8n workflow that once a day queries Attio for all active leads, sends a batch summary into Claude to identify stuck deals or promising prospects and then writes back recommended changes to messaging, new angles and priority leads to contact manually or with a higher touch sequence.
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 turns your outbound process into an agentic AI SDR that can research leads, choose angles, write outreach, follow up and update your CRM with minimal human input. The system runs on four core layers. Attio holds your leads and their current state, Relevance AI enriches and scores them, Claude decides what to say and what to do next, and n8n orchestrates the entire loop.
Everything starts in Attio. You set up an Outbound SDR pipeline with stages like New, Ready for outreach, Step 1 sent, In sequence, Replied, Qualified and Closed. Each contact record has fields such as ICP fit score, pain themes, last outreach step, last contact date, next action and objection notes. Attio becomes the memory for your agent so the system can always see who you have contacted, what was said and what is supposed to happen next.
Leads enter the system either from a CSV import, manual entry or a connected lead source. As soon as a new lead lands in Attio with a status like New or Ready for outreach, n8n triggers the agent workflow. The first job is to give the AI enough context to decide whether this person is worth pursuing and how to speak to them. To do that, n8n sends key data to Relevance AI, such as the company domain, the role, and any existing notes or public description.
Relevance AI pulls extra information from the company website or other accessible fields and turns it into structured signals. It can identify the industry, approximate team size, product type, positioning, possible tech stack hints and language that points to pain points or priorities. From that, Relevance AI returns a JSON object that includes tags, ICP fit signals and a recommended priority score. At this point you have moved from a cold CSV line to an enriched prospect with actual context.
n8n then sends this enriched data to Claude. This is where the agent becomes intelligent. Claude reads the present state of the lead, the ICP rules you define, and the extracted signals from Relevance AI, then decides the best outreach angle. It designs a narrative that connects your offer to the prospect’s likely pain points, writes a personalised first line that references something relevant about the company or role, and generates a sequence of outreach emails with different angles across multiple steps. It can also output objection handling scripts and suggested next actions when certain reply types come in.
All of that output is mapped back into fields in Attio. The contact record now contains the first line, the main opening email, several follow up variations, and notes on how to handle likely objections. At this point you have a ready to run playbook for that prospect, generated on the fly without an SDR touching it.
The next part is sending and sequencing. In the base version of this blueprint, n8n sends emails directly via SMTP or your email provider. It pulls the correct email body and subject line from Attio, sends the message, then updates the lead status and last contact fields. A delay node in n8n waits a set number of days, checks if there has been a reply, and if not, sends the next follow up email that Claude generated. This continues until the sequence completes or the prospect replies. For users who want best in class deliverability, this step can be enhanced by swapping the sending layer to tools like Instantly or Lemlist, while keeping Relevance AI, Claude and Attio as the intelligence and memory.
Reply handling is where the agent feels alive. n8n monitors your inbox for replies, then forwards the conversation thread to Claude. Claude classifies the reply as positive, neutral, objection, referral, not interested or spam. It then suggests an appropriate response and a next action, for example booking a call, sending a case study, handling a pricing objection, or closing the record. Those outputs are written back into Attio so you see a clear history of the decision process. For simple positive replies or predictable objections, you can let n8n send Claude’s reply automatically and move the lead to a Qualified or In conversation stage. For complex or high value deals, you can require human review before a response is sent.
The system also runs daily reflection. A scheduled workflow in n8n can pull a list of active and recently contacted leads from Attio and send a compressed summary into Claude. Claude can look across multiple conversations at once and recommend changes to messaging, new angles for specific segments, and a list of stuck leads that should be escalated or closed. This lets the agent learn from patterns instead of treating every lead in isolation.
If you want more aggressive enrichment and prospecting, you can layer Clay on top of Relevance AI at the enrichment stage. Clay can source and refine lead lists at scale, while your core blueprint still uses Relevance AI and Claude for agentic decision making and Attio for memory. If you want enterprise grade outreach volume and deliverability, you can plug Instantly or Lemlist into the sending stage and let n8n coordinate between them and Attio.
The end result is an agentic AI SDR system that behaves like a junior SDR team member. It researches leads, prioritises them, writes and sends sequences, interprets replies and suggests or sends the next move, while your CRM always remains up to date. Your human team then spends their time on live calls and high leverage deals, rather than typing cold emails and chasing non responders.
Share it with your network