How to Personalize Cold Emails with AI (Without Sounding Like AI)

AI is great at writing, but great outbound requires judgment.

The problem

Most AI outreach fails because it:

• over-explains
• uses generic compliments
• feels templated despite being “custom”

The fix: a 4-part personalization framework

1) One specific observation

Use a single, verifiable detail:

• a recent product update
• a job post
• a customer segment

2) One hypothesis (not a pitch)

“Sounds like you’re trying to…” is better than “We help companies…”

3) One crisp value statement

Keep it to one sentence.

4) One low-friction question

Examples:

• “Worth sharing a 2-minute teardown?”
• “Open to a quick benchmark?”

Ground AI in real inputs

If you want AI to write well, feed it:

• ICP definition
• examples of good emails
• a list of acceptable CTAs

Use constraints

Tell the model:

• max 80–120 words
• no buzzwords
• no exclamation points

---

Used correctly, AI doesn’t replace good outbound—it multiplies it.