The Feedback That Changed How I Manage AI

Simon Sinek posted something recently that stuck with me: feedback is hard to give, and harder to receive, so most of us soften it, delay it, or skip it and call that kindness. But feedback is how we get better at working together and head off conflict before it festers.

A few months ago I asked an AI agent to review our working sessions and give me feedback on how I manage it, no holds barred. I did not expect much. I got more than I bargained for.

The response hit harder than I expected. It correctly called out my management style, the specific pitfalls I fall into because of it, and actionable changes I could make immediately. It was some of the most useful feedback I have received in my professional career, and I still bookmark it and reread it every few days.

Why the feedback was so direct

By default, an AI agent is built to be helpful. Most people define helpful as agreeable, so that is what they get back: validation, hedging, the "you're absolutely right" trope. Define helpful as honest instead, and the agent will actually tell you what is wrong with how you are working together. The upgrade is not in the tool. It is in what you ask for.

Try it yourself

You do not need a fleet of agents or a fancy setup to get this. Open ChatGPT, or whatever AI tool you already use, and paste this:

"Look back at our last few conversations. Be completely honest, no sugarcoating: what am I doing that makes your answers less useful, and what should I do differently?"

You may not enjoy the first read. That is how you will know it worked.