Hand this case to your agent and build your own version

A key section of this article is locked, and an agent opening the link cannot see it either. Unlock it first, then copy the full text for your agent in one click.

Let Your AI Agent Build a B2B Lead List That's Actually Worth Cold Emailing

Lead enrichment is now something an AI agent can call directly. Describe your ideal customer in one sentence, and the agent finds companies first, then people. It fetches work emails only for the shortlisted people and returns a list where every row has a source. This breaks down why it works, gives a five-step SOP you can follow right away, and includes a prompt to give your AI agent.

Machine translation. Read the original

The problem

You want to cold email a B2B product but have no accurate customer list. Spreadsheet tools take time to learn and cost a subscription. Searching by hand is slow, and the lists are often inaccurate.

Who it's for

Indie developers building B2B products, sales and growth leads at small teams, and founders testing cold email for lead generation.

What you get

A written customer profile and a prospect list. Companies are filtered first, then people. Every email has a source. You spot-check the list, then send to a small batch.

How the workflow runs

  1. 01 Write the customer profile
  2. 02 Find companies first, then people
  3. 03 Enrich only the shortlist
  4. 04 Deduplicate and spot-check
  5. 05 Send a small batch and watch replies

Tools and accounts

  • Firecrawl's people and company data service (your AI agent calls it)
  • ChatGPT, Claude, Grok Bot, or any other AI agent that can connect to Firecrawl
  • Your own email sending tool

Before you start

  • You can describe what kind of companies and job titles fit your product
  • A Firecrawl account. Start with a small batch on the free credits
  • You're willing to spot-check a few rows by hand before sending

Firecrawl, the web data company, launched a people and company data service for AI agents on November 2, 2026. The company calls it "a headless Clay for agents." Clay is a lead enrichment tool many sales teams use. You wire up data steps one cell at a time in a spreadsheet.

According to Firecrawl, the service searches over 250 million people and 26 million companies. It finds work emails, job titles, and company details. Connect Firecrawl to ChatGPT, Claude, or Grok Bot, and you make your request in one sentence. You get 1,000 free credits each month. The tools are simpler now. Whether the list is any good still depends on how narrow your customer profile is before you run it, and whether you check a sample by hand before you send.

Start With the Workflow

Here is the full workflow for having your AI agent pull a B2B customer list. Your agent handles the first four stages. The last stage is yours.

Stage What it involves Output
Write the customer profile Spell out the industry, company size, region, technologies in use, and the job titles and seniority you want A checklist you can verify line by line
Find companies Filter for companies that match A target company table
Find people Look for people only at these companies, by job title and seniority A shortlist
Enrich Get work emails and background only for shortlisted people. Leave blanks where nothing is found A list where every row has a source
Spot-check Open a few rows at random and confirm the person still works there and the email is real A list you can send to

Firecrawl's companion skill sets rules for the agent too. If an email can't be found, the field stays empty. The agent doesn't guess from the first.last@company pattern, and it doesn't look up personal emails or private phone numbers. It builds the list. It doesn't send emails for you.

The coverage and pricing figures above come from Firecrawl's own announcement. We haven't verified them or tried the service ourselves. Firecrawl says it is "88% cheaper than Clay at scale." That comparison uses the list price of Firecrawl's annual Standard plan against the list price of one fully enriched record on Clay's Launch plan. Not every use case will save that much.

Why It Works: Three Reasons

  • Lead enrichment is now something an agent can call directly: Before, you had to learn how to chain nodes in a spreadsheet tool and pay for a subscription. Now you describe who you want in one sentence. The agent calls the data source itself. You don't need to learn any interface.
  • A narrower profile means a more accurate list and lower cost: Data is billed per record, so a wide scope spends credits on people who will never buy from you. Finding companies first and people second narrows the search at every step.
  • Manual spot-checks guard the last line: People in the data source may have already left their jobs, and emails may bounce. Checking a few rows before you send is much cheaper than sending a pile of bounces that hurts your sending domain's reputation.

Build It Step by Step: A Five-Step SOP

Five-step flow for having an AI agent pull a customer list, from writing the customer profile to sending a small test batch

You and your agent can both follow these five steps. The setup and calls for the agent are on the skill page firecrawl-lead-gen. Give that page to your AI agent. Then have it check each step against the "Done when" column as it goes.

Sign in to unlock this key section

New accounts get 10 free credits, plus 1 more for each daily check-in. Spend 1 credit to unlock and it stays open for you.

Safety Limits

  • Work identities only: Get only work emails and job titles. Don't look up personal emails or private phone numbers, and don't use this to dig into a specific person's private life.
  • Follow the laws where your recipients live: Marketing emails need a real sender identity and an unsubscribe option. Personal data in the EU and similar regions has stricter protection. If you aren't sure, ask someone who knows the rules.
  • You approve the money and the permissions: You confirm every credit spend and every third-party data source's terms. The agent can't accept them for you.
  • Keep keys out of the chat: When you need an API key, enter it yourself in your terminal or the app settings.

Common Mistakes

  • A profile that's too broad: If you only write "tech company executives," the agent will pull in a pile of irrelevant people, and your credits go to waste. Write the must-haves in a form you can judge one by one.
  • Treating the free credits as enough for 1,000 people: Data sources charge per record, from a few to more than ten credits each. How many people the free tier covers depends on which data source you use. Ask the agent for a price quote before you run it.
  • Treating a preview as the full record: In search results, a last name may be hidden, and an "email available" flag doesn't mean a real address exists. Enrich the record before you count it as complete.
  • Trusting a guessed email: Addresses built from the first.last@company pattern are often wrong, and they bounce. Leave the field blank if nothing is found.
  • Blasting without spot-checking: People may have left their jobs long ago. A flood of bounces can send your sending domain to the spam folder. Spot-check first, then send in small batches.

Want to see how your rivals find customers first? Read the case Your Rival's Growth Playbook Is Hiding in Plain Sight. Running a local business? See one-line local business lead lists.

Where it usually breaks

  • 01A customer profile that's too broad gives an inaccurate list and wastes credits.
  • 02Treating the free credits as enough to look up 1,000 people.
  • 03Treating a search preview as complete information.
  • 04Trusting emails guessed from a pattern.
  • 05Blasting without spot-checking, so bounces damage your sending domain.
Done with this one? Mark it to track your path.Sign in
Save it for laterSign in
Questions about this piece?Ask AI about this

Comments (0)

No comments yet. Start the conversation.

Sign in to join the discussion.Sign in