- The OG GTM Engineer
- Posts
- Your ICP is killing your AI GTM strategy
Your ICP is killing your AI GTM strategy
Stop trying to slice databases. Start discovering data patterns that actually drive buying behavior
Jordan Crawford
September 17, 2025
I had two customer conversations this week that made me realize something.
People are still clinging to the old ICP model, and it's destroying their ability to use AI effectively.
Here's what happened. Both prospects said the same thing: "Jordan, you'll analyze our CRM and tell us which plays to run for which prospects, right?"
I had to tell them no. I literally don't know how to do that.
Here's the problem with the traditional approach. People don't exist in neat database categories waiting to be sliced by revenue and industry. Even if you know someone works at a $50M manufacturing company, that tells you nothing about their actual situation.
Think about it.
Do you think someone wakes up in the morning after they hired their 1001 employee and their situation drastically changes?
No, that’s absurd.
So why are you targeting 50-1,000 employees?
And what would you say to them anyway?
Two people in the exact same role, same industry, same company size. One is dealing with a regulatory compliance crisis. The other just got budget approved for a major expansion. Completely different situations. Completely different messages needed.
You can't solve this by updating your ICP, you need better segmentation based on pain.
You need a fundamentally different approach.
Here's what works.
Get four people in a room.
The executive who fires your SDRs. Make them eat humble pie.
The founder who actually knows customers.
The SDR or AE who closed recent deals.
All of your customers (every closed/won call transcript).
Open ChatGPT.
Use the “Thinking” model or the “Pro” model.
Paste in a customer transcript.
Give ChatGPT context about you and your customer.
Then ask it to spend serious time online searching for data connected to try to connect to what the customer said that would indicate they were in a situation where they needed your product.
ChatGPT will find things you never knew existed. A news article from six months ago. A regulatory fine. A partnership announcement. A federal contracting database entry. Data connections you couldn't have discovered manually.
Do this for 20-30 won accounts.
You'll discover corners of the internet you didn't know existed. You'll see how different data sets connect to actual buying behavior. Real dollars transacted. Real problems solved.
This becomes the raw material for your AI strategy. Not made-up personas. Not demographic guesswork. Actual patterns from actual buyers connected to actual data.
The key insight here is understanding what AI can and can't do.
AI can research for you. Structure data for you. Build workflows for you. But it cannot think for you.
It doesn't know your buyer's emotional state. It doesn't understand the nuanced context of their situation. It can't communicate with the empathy that closes deals.
Those are uniquely human strengths. Double down on them.
Use AI as a thinking machine that helps you think. Not a replacement for your judgment.
The companies getting AI right aren't trying to automate away human intelligence. They're using AI to surface insights that make human intelligence more powerful.
Start with 20 transcripts and ChatGPT. Discover what data actually matters for your buyers. Then build your segmentation strategy from there.
This is the backbone for the PVP (Permissionless Value Prop) and PQS (Pain-Qualified Segment).
The old ICP model assumes buyers fit into neat categories. The new model assumes buyers exist in complex, discoverable situations.
Most GTM approaches are about scaling volume, my approach scales value.
If you have any questions, reply and I’ll get back to you,
Real Jordan