Last week, I was working on a piece of writing and considering a post I read on LinkedIn that said something along the lines of this: if it takes the reader longer to read the piece than it took the writer to write the piece, there's a problem.

The point was being raised in consideration of AI-generated writing, and I think it’s correct. It's not hard to come up with lots of examples where this is true—especially when it comes to books or long-form essays. Writing demands clear thinking and an ability to articulate your thoughts in a coherent manner, and that often takes a fair amount of time.

But as I began to converse with AI about that very point, it pushed back. It didn't agree with my position, and no matter how I argued it, it continued to try to show me that it wasn't true, giving me examples where it “proved” its case. It was suggesting that I simply not believe the thing that I believed—and I would argue know— to be true.

In many ways, my experience gets at the fundamental question of the moment: should you take the suggestion that AI gives you? In my case, AI was making an argument it could defend logically. I'm sure there's an example out there where a writer took less time to write something than it took someone to read it. However, my argument was rooted in principle—and in the lived experience of being a writer.

Oftentimes, the suggestions that AI makes are good ones. Perhaps AI surfaces a point you haven't thought of or gives you a framework that feels perfect for the task. It's genuinely useful across a wide range of activities. But as someone who uses AI all the time, in almost everything, across multiple businesses, I can tell you with certainty that AI is good at some things and not so good at others. And there's still an incredible amount of discernment, or, to use another word, taste, that you have to bring to every interaction with AI.

AI's best feature is also its biggest bug. What I mean is that it scratches an itch that we all have, which is to be done with the thing that we're trying to do, whether that's writing or thinking or something else. AI offers closure, and it offers closure quickly. But closure might not be what we need, even if it's often what we want.

In the end, I think it's perfectly fine to listen to what AI tells you, so long as you listen to yourself first. Remember that you have agency and experience that the model doesn't have, and ultimately, you understand the world in a way that the model simply doesn't.

So when it gives you a suggestion or trots out a piece of advice, don’t just accept it wholesale. Do the critical thinking required to make the most informed choice.