I’ve been wrong before. Not occasionally, consequently. Wrong about markets, wrong about timing, wrong about which technologies would matter and when. Wrong with enough confidence that I didn’t bother checking. That history is why I’ve made a habit of asking a question most people avoid: What if I’m wrong about this?

It sounds simple. It isn’t. Because asking it seriously means entertaining the possibility that the thing you’ve built your position around, the narrative you’ve been publicly backing, the bet you’ve already made, might be off. That’s not a comfortable place to sit. So most people don’t.

They do something easier instead. They find people who agree with them, content that confirms them, and communities that reward their conviction. It feels like rigor. It’s actually insulation.

Take a look at what’s happening with AI right now.

There are two camps, and both have largely stopped thinking. The first is all-in; AI changes everything, the transformation is inevitable, the only question is how fast. The second thinks it’s overbuilt and overhyped; a bubble dressed in technical language, dangerous in ways its evangelists refuse to acknowledge. Both camps have real arguments. Both camps also have something more dangerous than a wrong opinion: they have an identity wrapped around being right.

Once that happens, curiosity dies. You’re no longer processing evidence; you’re defending a position.

The bullish camp looks away from the structural problems. The hallucination problem isn’t just an engineering nuisance; for some use cases, it’s a disqualifying flaw. Most enterprise implementations are still underperforming expectations. The gap between impressive demos and reliable execution at scale is real and wide. These aren’t reasons to dismiss AI. They’re reasons to think more carefully about where the actual value accrues versus where the story is getting ahead of reality.

The skeptical camp has its own blind spots. The capability curve of the last 18 months isn’t something you can explain away with dismissal. Millions of people have already changed how they work, not because they were told to, but because the leverage is real. When critics like Gary Marcus and Ed Zitron are right, they’re worth reading closely. When they keep moving the goalposts after being proven wrong, that’s worth noticing too.

I’m bullish on AI. I use it daily. I see the leverage, the new categories being built, the ways it’s compressing what used to take weeks into what takes minutes. But I read the skeptics seriously, not to balance my portfolio of opinions, but because the holes in my thinking are most visible to the people who disagree with me. That’s not a sign of weakness. It’s the only way I know how to think clearly about something this consequential.

Here’s the part that most pieces like this one won’t say: intellectual flexibility isn’t just a virtue. It’s a competitive advantage, and most organizations don’t have it. The room where strategy gets made is usually populated by people who’ve already agreed. The data gets selected to support the direction already chosen. Dissent gets managed rather than engaged. And when the assumption at the center of the strategy turns out to be wrong, nobody saw it coming because nobody was seriously trying to.

If you’re leading through this period of AI adoption, the most dangerous thing you can have right now isn’t a wrong opinion. It’s an opinion you’ve stopped questioning.

The future doesn’t belong to the most optimistic people in the room or the most cynical. It belongs to the ones still asking hard questions about their own positions, when everyone else has already decided