The Problem Was Never Generation
The promise is seductive: describe what you want and AI builds it. The catch? Nobody tells you what happens when you still don't know what you want.
Before You Create, You Must Decide
The entire ecosystem of generative AI is being constructed atop an assumption nobody bothers to question: that the hardest part of software is generating the artifact.
The narrative suggests that once you possess a clear vision of what you want, the battle is won. And in some respects, that's true: generating code and interfaces with AI is astonishingly rapid and accessible. The speed of execution can be genuinely jaw-dropping.
Yet anyone who has actually designed digital products knows that generation is never the bottleneck.
The genuinely demanding part of the craft is everything that precedes it: deciding which problem to tackle, for whom, why this one rather than another, and why right now.
That conversation cannot fit inside a prompt because it has no defined shape yet. It is ambiguous, messy, politically fraught, and full of unanswered questions.
Refining vs. Directing
There is a world of difference between improving something that already exists and questioning whether it ought to exist at all.
The former requires analysis; the latter demands holistic business context, deep intuition about human behavior, and the hard-won experience of having made costly mistakes before, paired with the awareness to recognize when you're on the verge of making them again.
Modern AI tools can inspect an interface and say: "This checkout flow has too many friction steps; let's collapse them."
What an algorithm cannot say is: "You shouldn't be designing this feature in the first place because the real problem our customers are struggling with is entirely different."
That distinction is precisely what separates a visual executor from a true Product Designer.
"A designer's value was never in the screen. It was in knowing what deserves to exist, and defending that conviction before anyone opens a tool."
That conviction isn't developed through prompt engineering. It is forged through live projects, uncomfortable stakeholder debates, and having lived through the aftermath of an engineering team building something nobody asked for.
Generating outputs is becoming easier by the week. But deciding what actually deserves to be generated remains the most demanding, and most valuable, aspect of product design.