The Designer's Advantage When Building with AI

The first version of Quartermaster showed every piece of task information inline. Status, context, links, descriptions — all of it, right there in the task list. It worked. You could read everything. And it was completely wrong.

I caught it because I think about interfaces the way a pilot thinks about altitude. At 30,000 feet, you want to see where you are relative to the whole area. As you descend, things sharpen — you start to understand where exactly you’re landing. At ground level, you can finally read the billboards. The initial Quartermaster UI was like reading billboards from cruising altitude. Functional, technically complete, but it ignored how people actually process information in layers.

That’s not a coding problem. That’s a design problem. And it’s the kind of thing a non-designer consistently ships without noticing.

The skill that transfers most

There’s a narrative going around that AI makes design skills obsolete. That anyone can build now, so designers have lost their edge. I’ve been having these conversations constantly — with designers who are worried, with developers who are dismissive, with founders who aren’t sure what to think.

Here’s what I’ve landed on: the single most transferable design skill in AI-native building is communication.

This sounds soft. It’s not. AI only understands what you effectively tell it. If the output is bad, it’s almost always because the input was vague. Designers have been doing this for years — explaining interface decisions to product managers, advocating in stakeholder meetings, articulating why a specific interaction pattern serves the user better than another. That entire muscle is exactly what good prompting requires.

The difference is specificity. When I’m prompting for UI, I’m calling out exact Tailwind classes — “give the div pt-24” — not saying “add more spacing.” I know the vocabulary because I’ve spent years working in these systems. A developer might get the spacing right eventually, but the path there is longer and the first pass is less precise.

AI only understands what you effectively communicate. Designers have been training for this their entire careers.

The user-centric lens developers skip

When we built the initial version of Ovii, you could create videos. That was the feature, and it worked. We were pumped. But it wasn’t until we got people into it and started collecting feedback that we realized we’d missed something: part of the problem wasn’t just making videos — it was figuring out what types of videos to make.

That’s a user-centric insight. A developer building to spec would have shipped the video creation flow, checked the box, and moved on. The jobs-to-be-done perspective — understanding not just what the user asked for but what they’re actually trying to accomplish — is something designers bring instinctively. It doesn’t show up in a test suite. It shows up in whether people come back.

Omotenashi, or: anticipate the need

One of my first companies was Button, where our CEO Michael Jocconi had previously worked at Rakuten. He brought this concept of omotenashi — the Japanese hospitality philosophy of anticipating what someone needs before they ask for it. A hotel that pre-sizes robes for each guest before they arrive. Nobody requested it. But the experience is better because somebody thought about it.

I’ve carried that into everything I’ve built since. When I’m working on an interface now, the question isn’t just “does this work.” It’s “have we anticipated what the new reality of work looks like for this person?” Is it a human using this, or an agent? What context does each of them need, and when? Those are design questions that shape architecture, not just aesthetics.

If you find a problem, congrats — it’s now your problem.

That used to be frustrating. A designer would spot an implementation gap and have to file a ticket, write up the issue, wait for an engineer to prioritize it. Now you can just fix it. The ownership loop closed. And the designers who lean into that — who treat the shipped product as their responsibility, not just the mockup — are the ones making outsized impact.

The taste conversation

I need to say something about taste, because too many designers are hiding behind it.

Too many designers are using taste as a moat that will somehow protect them from AI. That’s nonsense.

Taste matters. It’s probably the most important quality when building with AI. But the next generation of foundation models will have significantly better taste, and using yours as a defensive position is a losing strategy. What taste actually gives you right now is the ability to build opinionated products — to express a point of view through an interface. AI makes that accessible to more people, not fewer. That’s a good thing.

Cap Watkins wrote a piece years ago called “The Boring Designer.” The core idea was that the best designers choose the obvious solution over the inventive one. I’ve tried to apply that throughout my career, and it’s only become more relevant. When you can ship in hours instead of weeks, the temptation to over-design is stronger than ever. Taste isn’t about making things precious. It’s about knowing what to leave alone.

What I’d tell a worried designer

I don’t have to imagine this conversation. I’ve been having it constantly.

Design skills matter more now than they did a year ago. The emphasis shifts — less time in Figma arranging rectangles, more time thinking about information hierarchy, user goals, and whether you’ve actually solved the problem or just built the feature. Most days I still open Figma, but the work I do there is different. It’s graphic and visual work, not product and UX flows. The product design happens in the terminal now.

The designers who will make the most impact going forward are the ones who think deeply about real problems, understand what the user is actually trying to accomplish, ensure those goals get met efficiently, and — in the best case — anticipate what the user needed before they knew to ask for it.

That’s not a new skill set. It’s the skill set designers have always claimed to have. The difference is that now there’s nowhere to hide if you don’t actually have it.