AI isn't replacing creators. It's changing what makes them valuable.

Sergio Slansky, Principal Design Manager @Microsoft (formerly Nike) — a designer's view on what AI changes about building, and what it doesn't.

I've always believed the best designers are builders. Not because they make the most polished screens or know the latest tools, but because they take responsibility for turning ambiguity into clarity. They understand people, problems, technology, and business, and they connect those threads into something worth using. The craft was never really about screens. It was about understanding what people need and building experiences that get them somewhere they couldn't get to as easily before.

AI doesn't change that. It raises the stakes.ew

We're entering a moment where the distance between an idea and a working product is collapsing. We can explore, prototype, write, design, and ship faster than ever. That's thrilling. It's also where the trouble starts, because when the cost of creation goes down, the value of judgment goes up. That single trade is the whole story of what's about to happen to our work.

The unicorn was always a signal

For years we celebrated the unicorn: the designer who could code, the engineer who understood experience, the founder who could design, build, and sell. They were rare because real capability across disciplines took years. A designer going technical wasn't just learning syntax — they were learning architecture, constraints, debugging, deployment, and the invisible decisions that make software hold together. An engineer moving toward design had to absorb human behavior, interaction patterns, accessibility, and the emotional side of a product.

Those skills still matter, and AI doesn't erase them. Let me be blunt about one thing: the idea that AI turns everyone into an expert misses the point entirely. Experts are valuable precisely because they carry depth, scars, and judgment that no prompt hands you.

What AI actually does is quieter and more useful. It lets people move past the edges of their discipline. A designer can explore a technical direction without first becoming a senior engineer. An engineer can prototype an experience without waiting on a spec. A founder can test an idea before spending a quarter on it. For most of software's history, good ideas died in the gap between thinking and making — you needed the team, the budget, the runway, the time. AI shrinks that gap. The first version of an idea is becoming cheap.

That's an incredible gift. It's also a trap.

The problem with infinite possibility

The danger was never that we couldn't build. It's that we'll build everything. I've watched this go both ways, long before AI.

The first product I worked on at Nike was Adapt — the app that controlled a self-tightening shoe. We had about ten months. The thing everything else depended on was trust: we were asking an athlete to put their foot into a shoe that tightened on its own and believe it wouldn't fail them mid-game. For someone whose body is their livelihood, that is not a small ask. Earning it was the product. But we didn't stop there. We added preset tightness Modes for a run or a stretch on the bench. We added adaptive lighting for personality. We built a "locker room" for pairing multiple shoes — a real need, but for a team equipment manager, not the athlete we'd spent ten months trying to win. We were building for a consumer and a professional in the same MVP, and every feature quietly asked the user to figure out which product they were holding. None of them were bad ideas. That was the trap. Each was defensible on its own; only stepping back did you see we'd answered "how do we add this?" four times and never really sat with "should this exist — in the first version, for this person?"

And it shipped, with fanfare. The Back to the Future lineage gave it a launch most products never get; Jayson Tatum debuted them on an NBA court against the Raptors, and other stars followed (Dončić, Kuzma, Morant, Fox, Stewart, Plum). But the thing people actually bought it for wasn't the thing we'd built it for. Adapt became a hypebeast collectible, not a shoe you reached for on an ordinary Tuesday. The accessibility future we'd imagined — the real reason a shoe that ties itself matters, for people who can't easily tie their own — never got its shot. The team spun up two more iterations on the platform. A few years later, Nike shut it down. It didn't fail for lack of ambition. It failed because it was trying to be too many things for too many people to ever be the one thing someone truly needed.

A different product taught me the other half. Nike's time-off system ran through a virtual machine into an antiquated tool with a brutal UX. Most people didn't bother — they took their time off and never logged it. The ones who did burned seven-plus minutes learning the thing to file a single request, and managers mostly never went in to approve. The company was flying blind on its own PTO. There were ten problems worth solving. We solved three: an employee requests time off, a manager approves or denies, and a manager sees the team's calendar. That was the MVP, and we left the rest on the floor on purpose. Within a month, a third of all requests came through it — on word of mouth, nobody forced to switch. Fourteen months later it was two-thirds, managers included.

Same lesson, from both directions. Adapt was strongest where it was most focused and weakest everywhere we let it sprawl. Time Off won because we had the discipline to solve less. Products rarely fail because a team couldn't ship one more feature; they fail because they forgot the problem they were there to solve. Just because we can solve a hundred problems doesn't mean we should pour a hundred solutions into one experience. That's not innovation. Most of the time it's just the failure to make a decision.

Judgment is the new craft

When anyone can generate ideas, prototypes, and features on demand, execution stops being the thing that sets work apart. Judgment does. Knowing which problems matter. Understanding what people actually need versus what they'll click on once. Recognizing the line between a capability that creates value and one that just creates complexity. Having the taste to feel when something is right, and the discipline to cut what isn't. Judgment is discernment with a reason behind it — not just seeing what matters, but knowing why it matters to this person. And being willing to be accountable for the answer.

This is the opportunity, and it's a bigger role than the one we've been handed. Builders aren't just making artifacts anymore — they're editors of possibility. The one who stands out won't be defined by a single discipline, but by how well they connect them: customer empathy, product thinking, technical curiosity, business sense, and storytelling, pulled into something coherent. AI is a force multiplier for that work. But the hardest parts stay human: choosing what matters, creating focus, making the tradeoff, seeing the real problem underneath the technology.

Because customers don't care about capabilities. They care about outcomes. Nobody wakes up grateful that a product uses AI. They care that it helped them do something better, faster, or in a way that felt more human.

AI gives us more power. Judgment gives it direction.

I'm optimistic about where this goes. AI lets builders move faster, explore more, and create things that used to take a room full of people. That's real, and it's worth being excited about.

But the future won't belong to whoever creates the most — AI makes creating easy. It'll belong to the people who create with intention. Who ship, watch what actually happens, and change course when the work tells them to. Because a great product isn't the sum of everything it can do; it's the result of every deliberate choice about what it should do. And the honesty to keep making those choices after it's in someone's hands.

The hardest part of building was never making something. It's knowing what deserves to exist.