How to Automate LinkedIn Prospecting With AI? (2026)

Most LinkedIn prospecting automation fails the same way: the same generic note goes to hundreds of people, acceptance drops, and the account gets flagged.

AI can fix the part that matters, a real reason to connect for every lead, without putting your account at risk.

This guide shows how to automate LinkedIn prospecting with AI the safe way, with a workflow we tested on 25 real sales leaders.

There are two ways to do it: an AI Agent in Claude Code, Codex or Cursor, or an n8n workflow that runs from a form.

We ran both on a real Sales Navigator seat, with real leads and no messages sent. Copy the workflow below, or follow the same method with your own tools.

We will walk you through the process step by step.

Everything is in one public GitHub repo: the skill, the writing rules, a small helper script and the n8n workflow. Paste the prompt into your AI Agent, or pick the n8n tab to run it from a form.

Agent kit Free · Public

GitHub n8n

Let your AI agent set it up

Paste this prompt into Claude Code, Codex or Cursor. Your agent clones the public repo, follows its instructions and asks before every change.

Prompt · GitHub Copy

Clone https://github.com/evaboot-labs/ai-linkedin-connection-notes, read AGENTS.md
and follow the skill. Write my LinkedIn connection notes for today. Ask me who I
target first, and show me the notes before anything else happens.

In the repo

• The ai-connection-notes skill: search, preview, export, write, check, queue
• notes.py: the search filters, the facts, the checks and the sheet
• NOTES.md: who you write to, and the rules every note follows
• The n8n workflow, with its setup guide

Open the repo on GitHub

Set it up in n8n

Paste this prompt into your AI agent and it walks you through the import, one step at a time. Or open the workflow JSON on GitHub and import it yourself.

Prompt · n8n Copy

Set up the AI LinkedIn connection-note workflow in my n8n.
Clone https://github.com/evaboot-labs/ai-linkedin-connection-notes and follow docs/n8n.md.
Walk me through importing n8n/ai-linkedin-connection-notes.json, adding the Evaboot,
Anthropic and Google Sheets credentials, and filling in Your settings, one step at a time.

In the repo

• The workflow JSON: a form in, a sheet of checked notes out
• docs/n8n.md: the three credentials and the settings, step by step
• The same rules and checks as the script

Open the n8n JSON on GitHub Open the repo on GitHub

We used Sales Navigator, Evaboot and Claude in our example, but with a few tweaks, the same method works with another model or outreach tool.

What Can You Safely Automate in LinkedIn Prospecting?

The work before the click. Finding the right people, reading their profiles and drafting a message are slow by hand and safe to automate.

The click itself, sending from your account, is what LinkedIn watches. That is where automation needs the most care.

The workflow in this guide automates the first three rows completely and leaves the sending to you.

Which Tools Can Automate LinkedIn Prospecting?

Three families of tools do the job, and they carry very different risks for your account:

1. Cloud Automation Tools

Tools such as lemlist, HeyReach, Expandi or Waalaxy run your LinkedIn steps from their own servers, on a schedule, with daily limits you set.

They are the right choice when you send every day at volume. Pick one with limits, working-hours sending and reply detection.

2. Browser Extensions

Extensions click inside your own LinkedIn tab. They are cheap and quick to start, but they act from your browser session.

That makes them easier for LinkedIn to spot, and they stop when your laptop sleeps:

3. AI Agents With Sales Navigator Data

The newest option is an AI Agent connected to your data, such as Claude with the Evaboot MCP server. It finds the leads and writes the notes, and it sends nothing.

It pairs with either of the other two: the agent does the research and the writing, your cloud tool or your own hands do the sending. For a full comparison of senders, see our list of LinkedIn automation tools.

How to Stay Within LinkedIn's Limits

LinkedIn does not publish a safe daily number for connection requests. It does cap invitations per week and restricts accounts that look automated.

Four habits keep your account safe:

1. Start Low and Raise Slowly

Our workflow defaults to 20 leads a day. Raise it only if your acceptance rate holds for a couple of weeks.

A falling acceptance rate is the first warning sign, long before a restriction.

2. Warm Up a Quiet Account

An account that sent nothing for months and suddenly sends daily requests looks automated. Build up over a few weeks instead.

Visit profiles, react to posts in your feed and accept your own pending invitations in the meantime.

3. Withdraw Old Invitations

A long list of unanswered requests looks like spam. Withdraw old ones regularly from My Network > Manage > Sent.

If you keep hitting the cap, read how the LinkedIn weekly invitation limit works.

4. Write a Note Worth Accepting

The safest request is one people accept. A short note with a real reason does more for your account than any limit setting.

That is the part this guide automates, with one rule: every note rests on a fact from the lead's own data.

How to Automate LinkedIn Prospecting With AI: Step by Step

Our workflow turns an intent signal into a queue of approved connection notes, one per lead, in six steps.

The agent decides only where judgement is needed, the notes; the search, the facts and the checks are the same on every run.

1. Pick One Intent Signal

An intent signal is a recent event that gives you a reason to reach out now. Sales Navigator has two that work well for connection notes:

Pick one per list. Our test used changed jobs, for UK sales leaders at software companies with 51 to 500 employees.

2. Build the Search and Preview It for Free

Ask the agent for the people in plain words: "Heads of Sales and VPs of Sales in the United Kingdom who changed jobs recently". The Evaboot MCP turns that into a Sales Navigator search.

Our first search returned 6,950 people across all sectors, including coaches and consultants. Too broad to write good notes for.

Adding industry and company size cut it to 106, all in software. The agent adds those filters to the URL with one command in the repo, then previews the list for free.

Here is what those filters look like in Sales Navigator, on another search:

Reading the preview costs nothing. Check the first titles and companies before you export anything.

3. Export the Leads With Evaboot

The agent exports your daily number, 20 in our test, with Evaboot. It costs 1 credit per lead.

The export carries the facts a good note needs: the current title and how long they have held it, their headline and profile summary, posting frequency, and the company's description and growth.

One field changed our plan: all 20 leads had a "Low" posting frequency. So "I saw your recent post" was not an option, and the new role became the main hook.

4. Write One Note per Lead With AI

The agent writes one note per lead, or a skip with a reason. The rules live in a one-page file, NOTES.md:

A good hook is specific. "Congrats on the new role" is weak; "After building a 45+ seller digital team at SailPoint, I'm curious what you'll build for Maia" comes straight from the lead's own summary and their new company's product.

5. Check the Notes Automatically

A script in the repo checks every note before a human reads it: the 300-character limit, a named hook, a reason for every skip, and repeated endings.

The last check earned its place. Our first draft ended 18 of 19 notes with "Happy to connect", which reads like a template. The check now fails that, and the agent rewrote the endings.

6. Approve and Send

The notes land in a sheet, one row per lead, with the hook next to the note so you can judge it at a glance:

Edit any note you want, mark the ones to send, then send them your way:

To refill the queue every week, put the search on a recurring Evaboot export: each run returns only new people. That is what we call IBAO (Intent Based Automated Outreach): fresh, signal-led leads flowing in without a manual pull.

How to Run It in n8n

n8n runs the same agent from a form. You paste a Sales Navigator search URL and the number of leads:

The workflow exports the leads with an Evaboot MCP Client node, builds the facts in a Code node, has Claude write the notes, runs the same checks, and appends the rows to your sheet:

We tested it with the posted-on-LinkedIn signal on 5 leads. It wrote 5 notes on the first run, each built on a line of the lead's own summary.

That test taught us one thing: the posted signal pulls in marketing leaders, who post more. Exclude marketing titles in the search when you use it.

Example: 20 New Sales Leaders, 19 Notes

We ran the agent on UK sales leaders who changed jobs, at software companies with 51 to 500 employees. Here is what each step returned:

The hooks split three ways: the new role itself (all 19), company growth where it was striking (up 117%, 88% and 49% in a year), and a line from the lead's own profile in most notes.

The skip is the honest part. A lead with no recent change and an empty profile gets no note, rather than a generic one that hurts your acceptance rate.

LinkedIn Connection Note Templates

These four structures come from the notes our agent wrote. Fill the brackets from the lead's data, never from guesswork:

Keep the ending varied and the note under 220 characters. For longer messages after the accept, see our guide to hyper-personalized LinkedIn messages.

Common Mistakes

Four mistakes cause most failed LinkedIn automations:

They all lower the acceptance rate, which is what protects your account.

1. One Template for Everyone

"I'd love to connect" sent to 500 people reads like what it is. Start from a signal and give every note its own hook.

2. Letting AI Invent the Hook

Asked to personalise, a model will happily mention a post or a funding round that does not exist. Give it only the export fields, and let it skip.

3. Automating the Sending Too Fast

Going from zero to a hundred requests a day is the fastest way to a restriction. Warm up, start at 20, and watch the acceptance rate.

4. Pitching in the Note

A product pitch in a connection request gets ignored or reported. The note opens a door; the conversation comes after the accept.

Frequently asked questions