How I think design is changing

Two ideas run through everything I write and every talk I give. The first is that execution, turning a design into a working product, used to be the slow part: the handover to developers and the build that followed. AI has made it fast, so the part that matters now is judgment, knowing what good looks like and being able to steer toward it. The second is that the systems we design have to be readable by machines and not only by people, or they fall behind. Most of what follows is one of those two ideas worked out in a specific place.

Direction over execution

For years, execution was the slow, effortful part of design work: turning a decision into something real took a team, time, and a handoff. A whole practice grew up around managing that gap, handoff tools, spec files, component libraries. AI has collapsed the effort, and a designer who knows what a good page looks like can now produce a running application, not a static mockup, without a developer in the loop. The industry spent twenty years treating execution as the main constraint, and for most work it no longer is.

The easy conclusion is that designers are being replaced, and the demos make it tempting. Install a few skills, describe a website, and a polished one appears in minutes. For a personal site or a small project, that can be all you need. What it does not cover is everything a real project for a large organization depends on, who the audience is, what the business is trying to achieve, what the stakeholders need, the brand understood properly, and a design system underneath it all. An AI has no taste of its own and cannot set the standard it is judged against, so without those things it fills the gaps with its own guesses, and what comes back is convincing but wrong.

That distance between convincing and correct is where a designer earns their place, and a real system is what keeps the machine from guessing across it. Anyone can generate output now; the rare part is telling whether it is any good and why. The making has been automated. What is left is strategy, direction, and taste, which is not a smaller job than before, if anything a more demanding one.

The tools have changed, but the skill they amplify has not, and the designers who saw that first are the ones out in front.

How I actually work

Everything I know and everything I am working on lives in one structured system that carries its own context, so an agent, in my case Claude Code, knows who I am and how to work inside it. Instead of starting from nothing, it starts from everything I have already thought through, reads across every project at once, and writes its work back into the right place rather than a chat window I will close in an hour. The decisions I make often I package as reusable Skills, so a good call made once does not have to be made again. The system holds the context, and I keep my attention for the judgment.

This is the same thing I argue for with design systems. A good system has always helped people work with consistency and speed. When it is also readable by a machine, it gains a second kind of power: an agent can build from it directly, not only a person. My own knowledge is structured the same way, so it would be strange to argue that to organizations while working the old way myself. And it is not theory here: everything on this site was drafted by agents reading that system and writing back into it, with me directing.

The tools will keep changing, but the shape will not: put the context where the machine can reach it, and keep your own time for the decisions only you can make.

Contact

Email: pierre@bremell.com
Phone: +
46 709353902
LinkedIn: bremell

© Pierre Bremell
Experience Designer