Recently I was reflecting on my AI usage, and I realized that I never give direct commands to an AI agent. For example, when the AI agent generates bad code, I never say: "This code is bad, fix it". Instead I say something like: "The code uses this pattern, do you think it is a good approach?", and let it figure it out on its own. In 90% of cases it finds the flaw and fixes it. Sometimes it proposes an even better fix than the one I had in mind. Sometimes the AI's approach was right, and I was wrong.
