The Smart Way to Use AI

I don’t think the future is about humans versus machines. It’s about learning when to lean on each of them.

About a year ago, I took my first ride in a Waymo in Los Angeles. I’ll admit I was a little nervous. Sitting in the back seat with nobody behind the wheel just felt wrong. Every stop sign, every turn, and every pedestrian had me wondering if this thing was about to make a mistake. By the end of the ride, though, I had to admit it was pretty impressive.

Later that day I needed another ride, but it was outside Waymo’s service area, so I called an Uber. Within a few minutes I became convinced the driver was high as a kite. Maybe he wasn’t, but that’s certainly how it felt. I remember sitting there thinking, I was probably safer in the Waymo. Then I caught myself. Was I really safer? Or was I buying into the idea that because a machine doesn’t get tired, distracted, or emotional, it must automatically be the better choice?

I’ve thought about that moment a lot over the past year because I think it captures something much bigger than self-driving cars. As AI has become part of my daily routine, I’ve realized that the Waymo ride wasn’t really about transportation. It was about trust.

Not long ago I was reviewing a legal contract. I thought I’d be clever and save my lawyer a little time—and maybe save myself a few dollars. I asked ChatGPT to analyze the agreement, cleaned up the response so it sounded like I had written it myself, and sent my lawyer what I believed were my eight biggest concerns.

His reply came back pretty quickly. In so many words, he told me I was worried about all the wrong things. The problem wasn’t that ChatGPT was bad. The problem was that I had started trusting the output before I had really challenged it myself. I had let the machine do too much of my thinking.

A few weeks later, I asked both ChatGPT and Claude the same medical question. Their initial answers were surprisingly different. After several follow-up questions, they eventually moved much closer together, but getting there took time. That said, the back and forth left me with a much better list of questions to ask my doctor. It didn’t replace the doctor. It made the conversation with the doctor better.

When you’re dealing with an important medical issue, getting a second, or even third, opinion is generally a smart move. The same goes for legal advice. Two experienced lawyers can read the same contract and see different risks or opportunities. That’s not a flaw in the system. That’s usually where the value comes from. The more I use AI, the more I think of it as another opinion. It’s often an incredibly helpful opinion. It’s sometimes a surprisingly insightful one. But it’s still just one opinion.

Sometimes ChatGPT will catch something Claude misses. Sometimes Claude will make me think about an issue differently. Sometimes both will miss something that a human expert spots immediately. And now and then, the AI notices something that the human overlooked.

That’s exactly why I find myself seeking multiple perspectives more often than ever. Not because I trust AI less, but because I trust the process of comparing thoughtful opinions more. Whether those opinions come from two AI models, two lawyers, two doctors—or some combination of all of them—the exercise almost always sharpens my thinking.

That’s why I don’t think the future is about humans versus machines. It’s about learning when to lean on each of them. The people who get into trouble will be the ones who assume the first answer is the right answer, whether it comes from a lawyer, a doctor, ChatGPT, Claude, or a self-driving car. The people who will make the best decisions are the ones who keep asking questions, compare different viewpoints, and then apply their own judgment.

Maybe that’s the real lesson from my Waymo ride. It wasn’t that machines are better than people. It wasn’t that people are better than machines. It was that neither deserves blind faith. Whether I’m riding in a driverless car, reviewing a legal contract, or facing a medical question, I’ve come to the same conclusion. The smartest answer rarely comes from the first opinion.