GTM Engineering for LinkedIn: Build a Signal-to-CRM Outreach Workflow

TL;DR: You can build a LinkedIn outreach system the way you build the rest of your GTM stack: signals in, scoring, execution, outcomes out. The one rule is that your automations move data, and the LinkedIn actions stay in an app on your own computer. This guide wires it up with Google Sheets, a Slack channel, webhooks, n8n and the Claude or ChatGPT connector.

The architecture

Five stages, each with one owner:

Stage	Owner	How it connects
1. Capture signals	RB2B, Clay, PostHog, LinkedIn itself	Slack channel, Google Sheet, PostHog sync, LinkedIn URLs
2. Score	Reachy AI	ICP drafted from your website, every contact scored
3. Execute	Reachy AI	Connection note and follow-ups, paced from your machine
4. Reply	Reachy AI and you	AI drafts, you approve from Slack, Telegram or Claude
5. Record	Your CRM	Webhook to n8n, Make or Zapier

Why not send LinkedIn actions straight from n8n? Because every “LinkedIn API” node that does it either runs a headless browser in a datacenter or uses a session cookie from a cloud IP. Both are the patterns LinkedIn restricts first (why). Keep the automation on data, and keep the browser on your machine.

Step 1: Capture signals into a source Reachy AI can read

Pick the inputs that match your motion. You can combine several in one campaign.

LinkedIn-native signals (no plumbing). Paste a post, group, event, company page or Sales Navigator search URL. Reachy AI collects the people at its own pace. This is the highest-intent source for most teams (the playbook).

Website visitors. In RB2B, Vector or Warmly, send identified visitors to a dedicated Slack channel. In Reachy AI, paste that channel URL as a source. Each message with a LinkedIn profile becomes a contact.

Clay tables and other lists. Route rows to a Google Sheet, with Clay’s Google Sheets export or an n8n, Make or Zapier step. The sheet needs one column with the LinkedIn profile URL. Other columns (company, trigger, custom line) can be used in messages. Share it as “Anyone with the link can view”, paste the link and turn on Keep in sync. Reachy AI reads it daily and only adds new rows (sheet format and pitfalls).

Product signals. Sync PostHog users or import a cohort, for trials that stalled or accounts that went quiet (example).

GTM engineering tip: write the reason for the row into a column (“visited pricing twice”, “commented on competitor launch”). It becomes the first line of the message, which is what makes a signal-based message work.

Step 2: Score against your ICP

Paste your website URL and Reachy AI drafts your ICP: roles, seniority, company size, industries. Edit it, then every contact from every source is scored against it, and the best fits are contacted first.

This is the step most DIY builds skip, and it matters more on LinkedIn than on email. Your weekly budget per account is around 150 to 200 connection requests (2026 limits). Scoring decides whether that budget goes to buyers or to whoever happened to be at the top of the sheet.

Step 3: Execute at a safe pace

Write the connection note and two follow-ups, or let the included AI draft them from your ICP and the signal. Use variables like {{FIRSTNAME}} and {{COMPANY}}. Keep the note short with no pitch, and stop after two follow-ups. Signal-sourced prospects answer early or not at all.

Reachy AI then sends from a dedicated browser on your computer, during your working hours, with warm-up for new accounts and per-account daily caps. To scale beyond one account, add more and let Reachy AI rotate campaigns across them.

Step 4: Handle replies without living in LinkedIn

When someone answers, Reachy AI sorts the reply (interested, objection, closed, or over to you) and drafts an answer with your booking link or offer. Nothing is sent until you approve it. You can approve from the app, from Telegram, Slack or Discord on your phone, or from Claude and ChatGPT (how AI replies work).

Step 5: Push qualified contacts to the CRM

In Reachy AI, add a webhook (a URL from n8n, Make, Zapier or your own endpoint) and turn it on for the campaign. Every new contact that passes your ICP filter is POSTed to the URL as JSON, with the profile data Reachy AI collected (name, LinkedIn URL, job title, company and more).

A minimal n8n flow:

1. Webhook node: receives the contact.
2. HubSpot (or Salesforce, Pipedrive) node: search contact by LinkedIn URL, create if missing.
3. Set node: tag the source and campaign name.
4. Optional Slack node: post high-score contacts to a sales channel.

You can also send any list to a webhook on demand from the export menu, which is handy for backfilling a CRM the first time.

Step 6: Put an AI agent on top

Connect the Reachy AI connector to Claude or ChatGPT and your assistant becomes the operator of this system. It can:

• Build LinkedIn searches from a plain description of your buyer
• Score a list you paste against your ICP
• Check every message draft against LinkedIn writing rules before it ships
• Create a campaign, or add new sources to a running one
• Show the hottest prospects and the replies waiting for approval

The assistant only queues requests. Reachy AI applies them when the app is open, and does every LinkedIn action within its limits. A prompt cannot raise a cap. Details in LinkedIn MCP for Claude and ChatGPT.

Build checklist

• At least one signal source feeding a campaign (LinkedIn URL, Slack channel or synced sheet)
• A “reason” column or signal referenced in the first message line
• ICP reviewed and edited
• Short connection note with no pitch, two follow-ups, then stop
• Reply approvals routed to the channel you actually watch
• Webhook live on the campaign, tested with one contact in the CRM
• Weekly review with Claude or ChatGPT: which source and which message performed

Next step

Download Reachy AI and wire up one source and one webhook during the 14-day trial, which is the full product with no card. For the bigger picture of where this layer sits, read where LinkedIn fits in your GTM stack.