There's a skill no model has ever learned. We can argue about whether AI has taste, or preferences, or can be prompted toward inspiration. But there's one skill no designer, leader, or strategist will lose to it. Knowing when to stop gathering evidence and commit.

AI is generative by nature. It will keep producing options, framings, considerations, variations, for as long as you let it. That's useful early on. It expands the aperture, surfaces gaps, stress-tests assumptions. But at some point a decision has to be made. And that decision cannot be made by generating more.

It has to be made from a felt sense of what will work. Informed by all the reasoning. But not bound to it. This is the cut-off point, and it's where design judgement actually lives.

Generating more is not deciding

The risk AI poses to creative work isn't that it replaces judgment. It's that it makes judgment easy to postpone forever. To keep reasoning, keep generating, keep optimizing, in a way that looks like precision but might be more of avoidance. Committing, really committing, requires trusting your own read on a situation.

This idea, that you have to cut yourself off, is probably for most, still a fairly new one.

Before AI, the creation process set its own limits. You committed to one or a few direction and stuck to it. Not because you were certain it was correct, but because you had no choice. You followed research in combination with instinct, taste, intuition in many cases, because testing 148 versions before deciding was never actually on the table.

Today, that's changed. We can create and test all 148 versions if we want. So the decision of which one to ship becomes the real work. How is that decision made? Are we sitting with the 148 versions, checking in with our gut, weighing taste before we ship? Less and less.

So what's doing more and more of the deciding? Data.

When you can test 148 versions, choosing is the work

When the ability to create and measure outweighs the care it takes to listen, understand, and shape taste. What we're doing is moving further away from design, and closer to design by algorithm.

There are places where that's exactly right. Intelligent interfaces, for instance, showing a user exactly what they need at the right moment. Algorithms excel there.

And there are places where it works against you. When what you actually want is novelty. A unique product. Something of your own. Something that's never existed before.

Algorithms can always tell you how something works. Is someone dropping off here? Is this feature used more than that one? But the why becomes speculation. And being able to explain why something works, not just that it does, is what makes expertise transferable. It's how knowledge moves from one person or project to the next.

The why travels further than the how

That 'why-knowledge' later shows up as intuition, as taste. As a felt sense of what will resonate. It coexists with rational analysis, the best design decisions involve both, in sequence, with human judgment deciding when each has done its job.

This tension isn't solved by rejecting AI. It isn't solved by insisting intuition always wins, either. It's solved by getting better at naming what human judgment is actually doing. Not to justify it to AI. But to protect it from being mistaken for something lesser than it is.

Bias, in the sense that matters here, isn't a flaw in our thinking. It's the accumulated weight of everything we've learned about what works. The goal isn't to remove it, but knowing when to trust it.