Built an AI Agent That Gets 10x Cold Email Responses

Cold email is broken. Everyone knows it. Response rates are in the gutter because most outreach looks exactly the same: generic templates with a {first_name} merge field and some vague value proposition.

The problem isn't cold email itself. It's that personalization at scale is genuinely hard. Researching each prospect, understanding their context, and writing something relevant takes time—time most salespeople don't have.

So we built an AI agent to do it.

What the Agent Actually Does

The workflow takes a list of prospects and, for each one, the agent:

1. Researches the company - Visits their website, reads their about page, checks recent news or blog posts
2. Researches the person - Looks at their LinkedIn activity, recent posts, job changes
3. Identifies relevant angles - Finds specific things happening at their company that relate to what you're selling
4. Writes a personalized email - Not a template with variables, but an actually personalized message based on real research

The output isn't "Hi John, I noticed you work at Acme..." It's "Hey John, saw your post about struggling with document search across your 50-person engineering team. We just helped a similar team at [Company] solve exactly that..."

Why This Gets 10x Better Results

The difference between a 1% and a 10% response rate usually comes down to one thing: does the recipient believe you actually care about their specific situation?

Generic emails say: "I want something from you."

Personalized emails say: "I understand your situation and might be able to help."

The AI agent does the research that makes the second type of email possible—at scale.

The Technical Setup

The workflow uses several tools:

• Web browsing - To research companies and people
• Collection search - To pull relevant case studies or content from your own knowledge base
• Email integration - To queue up the drafted emails for your review

The agent iterates through your prospect list, does deep research on each one, and outputs ready-to-send (or ready-to-review) emails.

A Note on Human Review

We built this with a human-in-the-loop. The agent drafts, you review. For a few reasons:

• Sometimes the research misses context only you would know
• Your voice and judgment still matter
• Fully automated outreach at scale can cross into spam territory

The goal is to save you 90% of the research and writing time, not to remove you from the process entirely.

Try It Yourself

The workflow template is available in Needle. Connect your email, add your prospects, and let the agent do the research.

Jan Heimes is Co-founder at Needle. He still writes some cold emails manually, but only to prove he can.