The Cost of Letting AI Think First

We've all gotten faster at using AI. We upload documents, ask questions, get analyses and summaries, move on. It feels efficient (it often is). But as I use AI in my professional and personal life (and based on some of the articles that appear in my feeds) I've started wondering what we might be losing in the process.

Look at the words cited as similar to discernment. Would you want to lose the ability for any of them?

We might be giving away our discernment.

Discernment isn't a flashy word. It doesn't show up in capability frameworks or LinkedIn skills sections. But it's one of the most valuable things a professional carries. It's the ability to look at information, raw, conflicting, incomplete, and know what matters.

Discernment is built slowly. Through experience, through mistakes, through paying attention over a long time. And here's the thing about discernment: it needs to be generative before it can be reactive.

What does that mean in practice?

Forget about AI for a moment, when you do the first pass of thinking yourself, when you sit with a problem, read a document, form an initial view, your brain is doing something important.

It's connecting new information to everything you already know. It's noticing dissonance. It's generating questions. That process is where discernment lives.

When an AI does the first pass instead, the process is different. Your thinking becomes reactive. You're no longer generating a perspective because you're evaluating someone else's.

And evaluating is easier than generating, which is exactly why it feels so efficient.

The risk isn't that the AI is wrong. It might be perfectly accurate. The risk is that you never formed your own view to test it against.

That's not a trivial loss.

In high-stakes professional situations like a board report, client recommendation or decision about people, your discernment is often the most valuable thing in the room. Not the data. Not the summary. Your read of what the data means, informed by context and experience that no AI has access to.

What concerns me (and a plethora of writers) is that if you've outsourced the first pass consistently enough, that read gets harder to access. Slower and less reliable. Basically you're still capable of discernment. But you're less practiced at it.

So what to do?

Be intentional about when you let AI think first.

There's a difference between using AI to extend your thinking and using AI to replace the thinking you haven't done yet. The first is leverage. The second is shortcut. And shortcuts, used habitually, have a cost.

The most effective professionals I see working with AI aren't the ones who use it most. They're the ones who know when to read the document themselves first. When to sit with the problem before asking for a summary. When to form a view before inviting one in. They use AI to pressure-test their thinking. Not to generate it. That's discernment in an AI-assisted world.

Practical Application

If this resonates and you want to go deeper, two books worth your time:

• Thinking, Fast and Slow: Daniel Kahneman. The definitive account of how we form judgements, where we go wrong, and why fast thinking isn't always the asset we think it is

• Range: David Epstein. On why broad, slow, effortful learning builds the kind of flexible thinking that discernment depends on