Where it usually goes
Most studios meet AI at the end of the process. The idea is set, the layouts are approved, and a model is brought in to fill the gaps: more variations, a background, a cutdown, an image nobody had time to shoot. That is useful. It is also the least interesting thing these tools can do.
Speeding up production makes the cheap part of a project cheaper. The expensive part of creative work has never been making the thing. It is deciding what the thing should be, and finding out too late that the decision was wrong.
Strategy: more directions, compared honestly
Strategy is where a project usually has the least room to explore. There is a week, maybe two, to land a position before the work has to start, so most teams test two or three directions and defend the one that feels safest.
AI changes the economics of that stage. I use it to draft far more territories than a team could write by hand, to argue against the one I like best, to read a positioning line back in the voice of a sceptical customer, a competitor or a journalist, and to find the places where a strategy sounds good and says nothing. None of those drafts is the strategy. They are a way of seeing the whole shape of the options before committing to one.
The model does not know the client, the market or the politics in the room. It does not need to. Its job at this stage is to widen the field quickly, so that human judgement has more to judge.
Design: the prototype becomes the argument
In design, the shift is from describing an idea to showing it working. A concept used to arrive as a deck: a mood board, a few frames, a paragraph about how it would feel. Now it can arrive as something you click, scroll and react to, built in hours rather than weeks.
The small interactive line drawings on this site were made that way. For the home page I went through four versions with an AI model. The first three, a stack of layers, a product on a turntable and a projection floor, were well made and wrong. The fourth, a tray of case cards that riffle through on their own, was right. I would not have known it was right without watching the other three move.
That is the real gain. You stop debating a sketch and start reacting to the thing itself, so decisions get made on evidence, earlier, while they are still cheap to change. I wrote about the same pattern in Claude’s new motion design ability: the drafts were cheap; deciding was not.
A brand now has a second audience
There is a strategic shift here that is easy to miss. People increasingly ask an AI assistant who to hire, what to buy and what a company actually does. The contact form on this site asks how people found me, and ChatGPT and Gemini sit in that list next to Google and LinkedIn, because that is how people find people now.
So this site is built to be read by machines as well as people. Every page is available as plain text, the case studies and services can be read through a small public API, and there is a guide written for AI assistants. For a brand, that is a strategy question before it is a technical one: when an assistant describes you, is it describing the company you mean to be? Clear positioning used to help people remember you. Now it also shapes what the machines repeat.
What does not change
AI makes the first draft nearly free. It does not make the decision. Give it a vague brief and it returns a confident, polished, generic answer, fast. Taste, a point of view, knowing what to leave out, and the nerve to hold one idea when a hundred plausible ones arrive at once still belong to the people doing the work.
If anything, craft matters more. To direct a model well you have to know what good looks like and be able to say exactly why something is wrong. The people who get the most out of these tools are the ones who could do the work without them.
What to ask when you hire creative leadership
The useful question is no longer whether someone uses AI. Almost everyone does. It is where they use it: as a shortcut to finished assets, or as a way to explore, test and decide before the expensive part begins.
Used at the start, it means more directions considered and fewer defended by default, working prototypes instead of promises, and a brand that reads clearly to people and to the systems they now ask. That is how I bring it into a project.