For most of my career, the last mile of a design idea was translation.
I could define the flow, design every state, prototype the motion, and explain why the idea mattered. Then an engineer and I would translate all of that into working software. Inevitably, some of the feeling changed or got lost along the way.
AI is shrinking that last mile.
At first, that looks like a code-generation story. I think the more interesting part is what designers already bring to the work.
A product idea does not become software until someone makes hundreds of specific decisions. What happens first? What changes after a tap? What does the empty state say? What happens when the data is late, the input is wrong, or someone changes their mind halfway through? How should the interface respond, recover, and help them continue?
Designers already think this way. We turn an outcome into sequences, states, rules, content, and feedback. Those are design decisions, but they are also building instructions.
I have also spent most of my career sitting close to product management. At adidas, I work inside a product management team rather than a centralized design organization. After enough time in that seat, the product side starts to feel natural.
You develop a feel for what the business needs, why a feature matters to sales, what marketing will need to tell the story, and what content has to exist before anything can launch. You learn to separate an interesting interaction from a product worth funding. None of that appears automatically in a design file, but it shapes every good product decision.
That puts designers in an unusual position. We can understand the product intent, make it useful and expressive for people, and describe its behavior with enough precision to build it.
The part I was missing was engineering fluency. I have always loved that side of the work, but I did not have the language to move through it myself. I could describe the experience I wanted to an engineer. I could not inspect the system, make a change, and immediately see what the change taught me.
AI changed that learning loop. Now I can explain an interaction, see it running, find what is wrong, ask better questions, and keep dialing it in. I am learning the vocabulary through the act of building.
I felt this while making Taste Buds, a mobile web app for organizing potlucks with my family. I owned the product decisions, interface, motion, and implementation. I decided location autocomplete belonged in the first version because entering a place should not become the hardest part of creating an event. I did not know how to build it, so I worked through the integration with Claude until it worked.
At adidas, I reimagined our configurator as a working WebGL prototype using real garment geometry. Instead of describing cloth motion or a transition from customization to a saved design, I could build both and let people experience them. The conversation could begin with the actual feeling of the product.
It also makes my collaboration with engineers more specific. A working version exposes questions about feasibility, performance, architecture, and edge cases earlier. We can react to the same behavior instead of translating intent back and forth through a static artifact.
AI did not give me product judgment, visual craft, or care for the details. It shortened the distance between that judgment and working proof.
That is why I think designers are some of the best-placed people to build with AI. We have spent our careers turning ambiguity into concrete choices. Now those choices have somewhere to run.