Insight
Designing with AI
AI is your tool — it makes producing the work nearly free, five layouts or fifty. But it can't tell you which one is right, and without the fundamentals of the craft you're just choosing at random.
The old cost structure
Most of a designer's hours went into producing the artifact — drawing the screens, building the prototype. Deciding which direction to take was a sliver on top of a mountain of execution.
Execution collapses
Generating five dashboard layouts — or fifty — now costs almost nothing. The mountain of production flattens; what took a week arrives in minutes.
Fundamentals are the filter
Five cheap options go in; one right decision comes out. The narrow part in the middle isn't the AI — it's the designer's fundamentals doing the filtering. Take the fundamentals away and the filter is just a coin flip.
A few years ago, when a producer needed a renewals dashboard, the work was mostly making. You'd spend two or three days in Figma — drawing the screens, laying out the tables, wiring a prototype — before anyone could even react to the idea. If the direction was wrong, you'd spent the days to find that out. The cost sat in production, so that's where the time went.
Now I type something like "generate five dashboard layouts for insurance brokers based on these requirements" and five of them come back in minutes. Concepts, user flows, a first pass at the copy, a React component to click through — what used to take days takes minutes. The making stopped being the expensive part.
What's left is the part that was always there, buried under the production: looking at those five and knowing which one a broker will actually live in at 8am while chasing a loss run. That judgment didn't get cheaper. It got exposed.
And here's the part that's easy to miss in the excitement: AI hands you options, not judgment. If you don't know why one layout works and another doesn't — hierarchy, flow, what the broker actually needs — you're not choosing, you're guessing. The tool just makes a weak designer faster at picking the wrong thing. Fundamentals aren't optional now; they're the whole edge. Design is still a discipline.
So the skills shift. Prompting, orchestrating the tools, evaluating fast, dropping four of the five without regret. But the decision is still mine, and it rests on everything the craft taught me long before any of this. The tool generates five dashboards; it doesn't know which one earns its place in a broker's morning. I do.
Making got cheap; the discipline to choose didn't.