The popular framing is that AI is the expert and the designer is the prompt engineer. That's backwards.

The strongest argument against that framing is a structural blind spot: it lacks embodied spatial reasoning. It can't understand physical constraints the way a designer does from years of working with real products, real spaces, real people.

Handing AI the "teacher" role in a design collaboration is giving the class to the student. The designer's job isn't to react to what AI generates. It's to define the constraints before AI generates anything.

Try it on your next brief. Before you open a prompt, write down the physical, spatial, and contextual parameters of the problem. Give those to the AI as a constraint layer, not as a question. See how much the output changes.