How I Built an AI Agent That Generates Leads from a LinkedIn URL?

— And Why It Might Be the End of Manual Prospecting

We used to spend hours writing cold emails.

→ Looking up LinkedIn profiles.
→ Copy-pasting job titles.
→ Crafting intros.
→ Rewriting each line so it “felt personal.”
→ Testing tools. Paying for plugins.
→ Hiring VAs. Managing inboxes.

And for what?

2% reply rates — if you're lucky.

So we built an AI agent to do the whole thing for us.

Not just the email writing.
The entire pipeline:
→ Scraping
→ Enriching
→ Writing
→ Delivering

All you do is drop in a LinkedIn URL.
Everything else happens automatically.

Here’s how it works — and how you can build it too.

---

🔧 The Stack

• n8n for automation (free, open-source, no-code)

• Apify or a local scraper for LinkedIn data

• Claude / GPT for email generation

• Slack, Google Sheets, Email, Notion for delivery

No browser extensions.
No form fills.
No fluff.

---

🧠 The Full Workflow Breakdown

Let’s walk through it step-by-step:

1. Trigger

It starts with a Webhook (or a scheduler).
You input a LinkedIn profile URL — manually or through another automation.

2. Input Parser

The workflow checks the URL.
It makes sure it's valid, extracts the right pieces, and ensures it’s clean for scraping.

3. LinkedIn Scraper

This is where the magic begins.
Using Apify or a headless browser, it scrapes key details:

• Name

• Title

• Company

• Industry

• Bio/About section

• Location

• Skills

• Education

No login. No API. Fully local and safe.

4. Data Cleaning & Transformation

Now the messy part.
The scraped data is cleaned using Set, If, and Function nodes in n8n.
It removes empty fields, trims whitespace, and normalizes the output.

Think of it as making your AI’s job easier.

5. Prompt Engineering Layer

This is the core.
We build a dynamic prompt that feeds into Claude or GPT.

It reads like this:

“Based on this person’s LinkedIn profile, write a cold email introducing our product. Focus on their job title, industry, and likely pain points. Keep it casual, concise, and human.”

Prompt includes:

• Profile data

• Email goal (e.g., intro, demo request, referral ask)

• Writing style and tone

• Optional personalization logic

This is where most agents fall flat — ours doesn’t.

6. Claude or GPT API Call

The prompt is sent to Claude or GPT via API.
We’ve tested both — Claude tends to be more structured; GPT a bit more creative.

In under 2 seconds, we get a personalized email back.

7. Response Handling

Once the AI responds, we parse the output:

→ Strip out unnecessary line breaks
→ Remove hallucinated data
→ Format it for delivery

8. Delivery Options

You can choose how and where to send the final message:

• Slack DM

• Email (via SMTP or Gmail)

• Append to a Google Sheet

• Push to Notion or CRM

• Return as a webhook response

You can even combine outputs — send the email to yourself, your SDR, and store it in Sheets at the same time.

---

🎯 What It Actually Does for Us

This one workflow books 20–30 qualified calls/month on autopilot.

But more than that — it changed how we think about lead gen.

Instead of:

• Spending time finding leads

• Manually researching them

• Carefully crafting the message

We’ve turned it into a system.
An always-on, no-fatigue, AI-powered prospector.

And the emails?
They don’t feel like AI.
They feel like you read the person’s mind.

Because technically… you did.

---

💡 Why This Matters (More Than Just Email)

This isn’t just about cold outreach.
This is about autonomous workflows.

Systems that:

→ Reason
→ Write
→ Act
→ Repeat

This same architecture can be used for:

• Warm intro requests

• Candidate screening

• Customer onboarding

• Referral automation

• Event follow-ups

• Partner outreach

• Internal summaries

Anywhere a human used to “read + write + send,”
AI agents can now do it — better and faster.

---

🧪 Bonus: What We’re Experimenting With

• Auto-tagging ICP fit based on role/title

• Looping through 50+ profiles at once

• A/B testing Claude vs GPT outputs

• Dynamic tone-shifting based on industry

• Delivering outputs as voice notes via WhatsApp

---

Final Thought: Build Once, Scale Forever

Most people are still building lead gen like it's 2015.
Tools duct-taped together.
Hustle over system design.
Manual steps everywhere.

But if you're in B2B and you're not automating intelligence — you're behind.

This AI agent isn’t just a cold email writer.

It’s a lens into what’s next:

A world where software doesn’t just follow instructions…
It makes decisions.

And we’re only getting started.

If you enjoyed this breakdown, follow along — I’ll be sharing more AI agent systems, prompt breakdowns, and automation builds over the next few weeks.

Welcome to the age of intelligent workflows. 🧠

Want to build this for your own business?

We go deep on these exact systems inside AcquisitionX — our private community where founders and marketers build AI-powered growth engines together.

✅ Access all 7+ ready-to-use AI agents
✅ Group live strategy calls
✅ Step-by-step workflow clones (Claude, GPT, LinkedIn, Cold Email, Reddit, and more)
✅ Battle-tested playbooks that generate 20–30 qualified leads/month — without paid ads or sales reps

Join us here 👇

https://www.skool.com/acquisitionx/aboutref=38f6b839e9cd400ea1272f6c5329c948