Design debt is the new tech debt
Or: the faster we can ship, the more design matters
AI-powered dev teams can now build fully working interfaces at a speed that would’ve been unthinkable two years ago. The results often look good at first glance. But show them to real users and cracks begin to show. No amount of visual fidelity will save an interface nobody thought through.
The dynamic has flipped: Design used to be the “fast” team, running ahead of development, getting iterations in before the final code was created. Now AI is enabling interfaces to be built automatically so that working software can arrive before the thinking that should’ve shaped it. These interfaces impersonate a finished product, and it’s often hard to parse which bits are good and which bits aren’t.
This isn’t a crisis for design. It’s actually the clearest proof of its value in years.
When you can generate a decent-looking UI in minutes, the differentiator is no longer whether something works. It’s whether it works for the person using it. Three teams can each build a checkout flow in an afternoon. All three will function. Chances are none of them will share patterns with the rest of the product, match a user’s mental model, or survive first contact with real people.
That’s design debt. And just like tech debt, you pay for it later… with interest. Naming that cost is how you make the case for doing the thinking upfront.
For product teams building with AI, the bottleneck has moved. Speed of production is no longer the constraint. Understanding users is. Knowing how they see the world, what they’re actually trying to accomplish, where the product fits into their life. That’s the work that’s hard for AI to replace.
The right response isn’t to slow down. It’s to spend the extra time better.
Now that AI can generate a working checkout flow in an afternoon, let’s build three and test them with users before picking one. If a prototype takes two hours instead of two days, explore the alternative you always had to leave behind. The teams who get this right aren’t using AI to ship faster. They’re using it to ship better. More alternatives explored. More assumptions tested. More thinking baked in before anything goes live.
The teams that use the time they’ve gained to understand before they ship are the ones that will build things people actually want to use.
The rest will ship fast and create work for the cleanup crew.