Last week I asked an AI to create a dashboard for Phora's analytics view. Thirty seconds later I had a beautiful interface. Cards, charts, clean typography. Looked professional at first glance.
Then I looked at what it showed. Seven metrics at once, three of them irrelevant. The most important number — active user trends — was buried at the end of the third row. Nobody would scroll that far.
The AI did exactly what I asked: it generated a dashboard. It didn't design one.
Generating ≠ designing
Generation is pattern matching. The AI saw what dashboards look like and produced something that looks like a dashboard. Cards, because dashboards have cards. Charts, because dashboards have charts.
Design is different. Designing means knowing that this user opens the page once a week, wants one number fast, and wants to leave immediately. They don't need seven cards. They need one number, large, centered.
This isn't a prompt problem. No matter how precisely I describe what I want — the AI doesn't understand why I want it. It doesn't know my users. It doesn't know the context in which they'll open the page. It hasn't seen how many people abandoned the previous version.
When generation is enough
Not every interface needs deep design thinking. An admin panel used by three people can be AI-generated just fine. A settings page, an internal tool — these work from templates.
The problem starts where the user makes a decision on the interface. Purchasing, signing up, configuring something that matters to them. Every pixel becomes a statement: what you show first, what you hide, where you direct attention. These are business decisions, not aesthetic choices.
In my own product, the AI writes the code, but the design decisions are mine. Which feature gets space on the main page. What's one click away. What goes deeper. These aren't layout tasks — they're strategic choices.
The difference is invisible — until it matters
An AI-generated interface and a designed one can look identical at first glance. The AI version might even be prettier. But look at conversion rates, usage patterns, drop-off points — the difference is clear.
Design isn't what you see. Design is what you decided.
There's always a next level.
If you like what you see — whether you're building a product or a team — I'd love to hear about it.