The best AI skills may turn out to be management skills. One I’ve been thinking about a lot is what I call “collaborative interrogation.”

It has the intensity of an interrogation, but without the adversarial or confrontational part. The goal is to get to a better answer together.

Ask a question, then really read the answer. Look for holes. Notice assumptions. Find the part that doesn’t quite make sense and push on it. Ask the hard follow-up. Follow an interesting thread somewhere you didn’t expect. And when it matters, trust but verify.

This has probably been my biggest unlock in doing high-quality work with AI. The high quality value is often not in the first answer. It’s in the fifth question.

If AI gives you a recommendation, don’t just accept it. Read the reasoning and figure out what it might be missing. If it gives you a number, understand where it came from. If it makes an assumption that doesn’t feel right, challenge it. If it proposes a strategy, push on the weakest part. If an answer opens up an interesting new question, follow it.

This isn’t really “prompt engineering.” It’s a much more human skill: listening carefully, exercising judgment, and knowing where to probe.

And that’s remarkably similar to managing people.

A good manager doesn’t give someone an assignment and blindly accept whatever comes back. They engage with the work, ask hard questions, challenge assumptions, spot gaps, and help someone get from a good first answer to a much better one.

And the best people to work with are good at the other side of this. They engage constructively with hard questions, explain their thinking clearly, and push back when appropriate without becoming defensive.

After all, more and more of what we’re doing with AI looks a lot like managing. And many of the skills that make someone a great manager are going to make them great at working with AI too.