AI and the brand designer

A year ago I would not have written this post. Six months ago I would have written a hot take. Now I have used the various AI tools enough in real client work to have something more measured to say, which is also the post I would have wanted to read a year ago.

What AI tools currently help with, in my work.

Drafting alternative copy for a piece that is almost there. Not writing the copy from scratch. Taking a paragraph that the client has written and is unhappy with, and producing five variations to react to. The variations are usually mediocre. Reacting to mediocre variations is sometimes the fastest way to find out what the original needed.

Generating placeholder images at the wireframe stage. Specifically when I want to communicate “an image of a person holding a ceramic mug” to a client at week one and do not want to license or stage a real photograph yet. The AI version is good enough to communicate intent, bad enough that nobody mistakes it for the final.

Naming things. Domain searches, working titles, file naming for case studies. Mediocre at all three, but mediocre is faster than my own attempts when I have nothing.

What AI tools currently do not help with, in my work.

Real visual design. None of the image generators produce work I would put my name to, at any stage of the process, beyond placeholder. The output looks like AI. The output is, in fact, AI. This is fine for placeholder. It is not fine for delivery.

Brand identity work. The tools have no idea what they are doing at the level of a coherent system. They produce things that look like brand identity work superficially and fall apart on the second look.

Typography decisions. The tools cannot actually pick a typeface for a brief. They will return Inter and Playfair Display, in that order, almost regardless of what is asked.

Saying no to clients. The thing the tools are worst at, paradoxically, is the thing most studios most need to be better at.

Two pieces of writing on the topic that have actually shaped my view, rather than the dozens that have not. Frank Chimero’s “The Inventory” is about how the tools change what we are willing to keep and what we are willing to throw away. Robin Sloan on AI is the most generous interpretation I have read of what these tools are actually for.

The middle position I have settled into: I treat AI tools the way I treat Photoshop’s content-aware fill. It is a useful piece of plumbing for a specific kind of nuisance task. I do not pretend it is doing the design. I would not be embarrassed to mention it in a case study. I would not credit it.

For tools the studio actually uses, sparingly: Claude for copy variations, Midjourney for placeholder images at the wireframe stage, Adobe’s generative fill for the boring restoration jobs that used to take an hour. None of these are doing the design. All of them are doing the kind of work the studio used to outsource or skip.

The studios I see panicking are mostly panicking about a kind of design work that should have been priced lower or refused, not about the kind of work they are still being paid well to do.