I use AI constantly as a second opinion — to pressure-test an idea, think through a decision, or show me something I might be missing.
But there’s an easy trap.
Instead of asking:
“I have this problem. Help me understand what might be causing it and what I should consider.”
we often ask some version of:
“I have this problem. I think this is how we should solve it. Help me make the case.”
AI will happily do that. It can organize your thinking, strengthen your argument and give you a very convincing case for the solution you already prefer.
The problem is, you never really asked: Could I be wrong?
If you actually want a second opinion, ask for one
Before telling AI what you think, try:
- What do you see based only on these facts?
- What am I missing?
- Which of my assumptions might be wrong?
- What are the strongest alternative explanations?
- Make the best case against the obvious solution.
- What information would you want before making a recommendation?
- What evidence would change your recommendation?
Then tell it what you were thinking.
Get the second opinion before you influence the second opinion.
Now apply that to a real founder problem
One of the most common frustrations founders have as their company grows is: “Why won’t my people take more ownership?”
My managers keep bringing decisions back to me that they should be able to make themselves. I’ve tried giving them more responsibility, but they still don’t take ownership. Help me create a better accountability system so they stop escalating everything to me.
AI can give you a very good answer: decision matrices, role clarity, accountability frameworks, meeting structures and management coaching.
Except there’s a question hiding underneath the one you asked: What if your managers aren’t the problem?
Maybe you regularly override their decisions. Maybe you tell them they have authority but still expect to be consulted. Maybe mistakes get a much bigger reaction from you than good independent decisions get rewarded. Maybe you’ve hired capable people but continue acting as the senior person in their function.
You asked AI how to fix your managers. You never asked whether you might be creating the behavior.
A better starting prompt:
My managers routinely escalate decisions I believe they should make themselves. Before recommending a solution, give me the strongest explanations for why this might be happening — including things I may be doing as the founder to create or reinforce the behavior.
This problem existed long before AI
Founders have always had to work harder than everyone else to get unfiltered disagreement.
Walk into a meeting and say, “My managers need to take more ownership. How do we fix that?” and you’ve already framed the problem for the room. How many employees are going to respond, “Actually, I think you’re the reason they don’t”?
Some will. A lot won’t.
Research published in Science in 2026 found that across 11 leading AI models, AI affirmed users’ actions 49% more often than humans in the advice scenarios studied. (source) So the agreement problem is real. But the leadership problem is older than AI.
AI didn’t create it. It can accelerate it.
If you use AI to reinforce the way you already see a problem, it can give you an enormous amount of articulate confirmation very quickly. But you can also use that same speed to get something founders have always struggled to get enough of: dissent.
The tool is new. The leadership problem isn’t.
And sometimes the most valuable thing AI can tell you isn’t how to solve the problem. It’s that you may be wrong about what the problem is.