I have spent thirty years making things look like they belong in the world. That skill still matters. It is no longer the center of the job.
An AI model does not know what any human wants better than a human. It predicts. It has seen more examples than I will ever hold in my head, so it can guess the next click, the average preference, the likely headline. That is useful. It is not knowledge.
People do not fully know their own wants either. They change, contradict themselves, say one thing and buy another. The model is a sharp statistician of that mess. It is not the person in the situation. Treat it as a forecast, not a verdict.
Design beyond the object
That same gap is why an experience designer beats a graphic designer who only designs the object when the work has to reach different audiences.
Graphic design, practiced the old way, starts with the artifact: grid, type, brand, “make it feel right.” After enough years, you learn that “it looks right” is what clients say when they cannot tell whether it works.
Audiences are not palettes. They are people under constraints: time, language, device, distrust, bad light, fat fingers, a form they did not ask to fill.
Experience design starts there: the job to be done, the journey, the test, the edge cases. Different audiences are not a color swap. They are different first questions. The craft still finishes the work. Research and use decide whether the craft is aimed at anyone but us.
AI as a cheap rehearsal
Partnering with AI applications sits in the middle of those two facts. AI does not give me better taste, and it does not give me the audience’s inner life. It gives me volume and a cheap rehearsal.
Variations before I pick. Captions and proposals at the speed of the layout. Tags and culls for photographs. A rough stand-in for the user I have not met yet: “Explain this to someone who is not a designer.” Video and mockups a one-person shop used to need a team for.
That is how retainers stop being discounted hours. The application takes the repetitive middle. I keep the brief, the cut, the politics of the room, and the last look.
Three principles
Do not outsource wanting.
Do not design only for the file.
Do not confuse generation with judgment.
Use the model to see patterns and make options. Use experience design to find out whether those options survive contact with real people. Use thirty years of graphic craft so the thing that survives is not only usable but worth putting in front of them.
The product is not the picture, the prompt, or the persona slide. The product is whether the right people can do the right thing without me standing there explaining it.
Everything else—AI included—is staff. I am still the art director.
Send me one report, deck, or page and the audience it has to reach. I will tell you where it holds up with a real reader, where it does not, and what I would change. If it needs more than that, we scope a bigger project from there.
Start with one piece →