I tested an AI persona of myself

Last week, I pointed Claude at several months of my Slack messages, emails, and Google Docs and asked it to build a behavioral profile of me.

Not a personality quiz. Not a strengths assessment. A profile built from how I actually communicate, make decisions, and operate day-to-day.

Then I tested it: 10 scenario-based questions across five rounds. “In this situation, what would Brett do?” The AI and I mostly aligned. But the misses were the whole point.

The Misses

The scenario: A teammate shares a half-baked product idea. It has potential, but clear gaps.

The AI predicted I’d highlight what I liked, then raise the concerns—the standard “feedback sandwich.” But it was wrong.

My truer instinct is to absorb the idea and add my own angle so the flaws get fixed as a byproduct, without anyone feeling “corrected.” It’s collaborative redirection. The idea gets stronger, and the person walks away feeling like we built something together, not like they just got feedback.

I’d been doing this for years without naming it. The AI’s miss forced me to articulate it.

A few times during the test, I caught myself giving the “measured” answer—the one that sounded like good judgment. Then I’d sit with it and realize… no, that’s not what I actually do.

In one question, the AI predicted I’d lean into the conversation when a teammate pings me mid-flow — turn a quick answer into a real discussion. I said no — I stay focused and get back to it. But the more I sat with it, the more I recognized the pattern. When someone brings me something interesting, I engage. The AI saw it in my data. I just hadn’t named it yet.

So What

That’s the thing about testing a persona of yourself—you’re not just evaluating the AI’s accuracy. You’re evaluating your own honesty. Sometimes you stop arguing and realize the mirror was more honest than you were.

This took less than an hour. No coach. No 360 review. No waiting for a performance cycle.

If you want to try this yourself, you don’t need a special setup. If you use Slack and Google Workspace, the data already exists. The only question is whether you’re willing to argue with what it shows you.

1. Ask AI to build a persona of you from your real work data — Slack, email, docs, all of it.
2. Run scenario-based questions against it.
3. Pay attention to the misses—especially the ones where you’re not sure if you or the AI got it wrong.

The version of you that exists in your head is not the version that shows up in the data. The gap between those two is the most useful thing you can learn about yourself.