Back to designing
I’ve been a design manager for a large part of my career. Earlier this year, I co-founded Steady and found myself designing a product and brand from scratch, and building most of the frontend with Claude Code. Being hands-on again felt great. When Steady didn’t work out, I started looking for projects where I could do more of that.
There was some uncertainty about whether I could operate as a high-level IC after so long in management. I trusted my product thinking, but the detailed execution that comes with being an IC designer hadn’t been part of my work for a long time. AI played a pretty big role in giving me the confidence to try.
Interestingly, I’ve since heard similar things from a couple of other design leaders who are moving back towards more hands-on work. Part of what seems to make that possible is the amount of making one person can now do with AI. It’s a small sample, but I wonder if we’ll see more experienced designers moving back into the work as AI makes it easier to be hands-on across more of the process.
Expanding my craft with AI
One of the most interesting things AI gives designers is access to areas where they understand what good looks like, but don’t necessarily have the craft skills to get there themselves.
Motion is a good example. My latest project was for a new workout format where the product is used during the class, with very little attention to spare. Working prototypes and motion were important to understanding whether the experience actually worked. I can imagine how an interaction should behave, but I’ve never been able to create motion myself. With Claude Code, those ideas could become working prototypes.
At one point, we had an idea that looked great in static screens and had quite a lot of support internally. Once it was prototyped in the context of a class, it was obvious that it didn’t work. We had previously considered building it with engineering to find out. Having the ability to create that prototype ourselves meant we could learn much earlier.
The same has been true of brand design at Steady, and of creating imagery.
AI doesn’t suddenly give a product designer the expertise of a motion or brand designer. It does give them access to parts of those crafts when they have enough of a foundation to direct the work and react to it, until something that could previously only live inside their head is actually produced.
When everything starts to look the same
The speed at which AI can get you to something that looks finished is pretty remarkable. I found Figma AI surprisingly good at UI creation, for example. It understands hierarchy, layouts and common interface patterns. Within a very short time, you can have something that looks like a properly designed product.
The problem is that “properly designed” is also starting to have a very recognisable look. Some of the early designs on my latest project got the same reaction from the PM and me: this isn’t it. Nothing was obviously wrong. They almost felt like wireframes rendered as finished UI — everything was where it should be, but there was very little that made the experience feel like ours.
I recently heard something similar from an investment fund that works with a lot of founders and early-stage companies. Founders are starting to see so many products with the same kind of AI-generated polish that they’re actually looking for design with some of the “clunkiness”.
We’ve spent years getting better tools that make it easier to produce polished design, and now getting to polished is becoming almost trivial. What happens after that is much more interesting.
Being able to make almost anything creates another problem: what should you make?
Having so many possibilities available also changes how you think about creative direction.
For the final workout screen on my latest project, we wanted to create a feeling of celebration. The first results from Figma AI made sense in a very literal way. It was recognisably celebratory, but it had very little to do with the feeling we wanted the product to have.
Things became much more interesting once the tools had directions to explore. What if it felt like being on a podium? What if medals, metals and gold became part of the visual language? What if we went somewhere darker, drawing from Berlin’s underground and techno scene? From there, AI was incredibly useful. The final prototypes I was able to create were much better than anything I could have made by myself.
Having access to almost endless possibilities makes human judgement and direction much more important. Sometimes it’s not even logic but simply being able to feel when something belongs to the experience or brand you’re trying to create.
The work still gets better with other people
One thing I’d forgotten after years of management is how much time designers spend working alone. AI changes that dynamic in an interesting way. There’s always something to bounce an idea off, ask for another direction, or simply react to. It can feel surprisingly collaborative, and I’ve found that genuinely rewarding.
At the same time, I think there’s a risk in how easy that makes it to stay in your own bubble. AI is very good at building on what you give it, but it also tends towards the familiar and is often too willing to agree with you. Another person comes with their own perspective, their own ideas, and a much greater willingness to tell you when they think you’re wrong.
Some of the best work on my latest project came from exactly that. One session with the PM and an engineer stands out: we were sketching and building on each other’s ideas, challenging some along the way, and ended up with the direction we actually used.
AI can make an individual designer much more capable and make solo work feel less solo. But if it becomes increasingly easy — and rewarding — to work this way, I wonder what that does to our instinct to seek out other people, and ultimately to the work we produce.
What I think my job is now
I started this period wondering whether I could really go back to being a hands-on designer. I don’t worry about that much anymore. AI expands what a designer can make, how quickly they can explore, and how far they can reach beyond the craft skills they already have. But those new capabilities come with tensions that I think are worth paying attention to.
We can get to polished design very quickly, while more and more of it starts to look the same. We can explore almost endless possibilities, which begs the question “which direction do we choose?”. And we can have something that feels like a collaborator, while potentially becoming less inclined to seek out people who bring genuinely different perspectives.
These aren’t arguments against using AI. They’re questions that become more important as it becomes part of how work gets done. How do we keep the work distinctive? How do we keep exploring the problem and different product ideas when it’s so easy to jump to UI? And who do we involve in the process?