The conversation about AI in creative work has two dominant registers.
The first is panic. AI will replace writers, designers, strategists, researchers. Human creativity is a temporary condition, soon to be made redundant by a model that produces faster and cheaper and without the inconvenience of having opinions about the brief.
The second is naïve enthusiasm. AI will handle everything tedious, freeing humans for purely creative work, in a frictionless collaboration where everyone's best ideas get executed immediately.
Both of these are wrong in the same direction: they misunderstand what creativity actually is.
Creativity is not the production of novel output. A language model produces novel output constantly. It produces sentences no one has written before, images no one has made before, arrangements that are technically unprecedented. This is not creativity. This is recombination at scale.
Creativity is the prior step: the decision about what question is worth asking. What problem is actually interesting. What would be genuinely new rather than statistically improbable. What matters and why.
This is a values question. Values require a position, a history, a body, a specific location in time and culture that generates a perspective. A model trained on human output can approximate the surface of this. It cannot have it. This is not a limitation to be corrected in the next version. It is structural.
What AI is, in practice
A thinking partner. Exceptional at accelerating the middle of a process — once a direction exists, once a question has been formed, once a judgment has been made about what matters and why.
It synthesizes faster than any human. It holds more reference simultaneously. It drafts without ego, revises without resentment, generates variations without the particular fatigue that makes a human on the third iteration increasingly attached to whatever they just produced.
These are genuinely useful capabilities. They compress the distance between idea and execution in a way that changes what small teams can produce.
What AI is not
AI cannot tell you which problem to solve. It can map the problem space with impressive comprehensiveness. It cannot locate the problem that is actually worth your attention, because that determination requires judgment about what the world needs, what your particular perspective can contribute, and what is worth making exist that does not currently exist.
This is the creative act. Everything downstream of it is, to varying degrees, execution. AI is extraordinary at execution. It is categorically incapable of the thing that precedes it.
The risk is not replacement. The risk is delegation. The creative professional who uses AI to accelerate their thinking is more capable than they were before. The one who uses AI to avoid the discomfort of forming their own position is producing output without authorship — a different thing from creativity wearing the same clothes.
The question the brief cannot answer, the direction that has to come from somewhere real, the perspective that makes a brand recognizable and a strategy coherent — these still have to come from a person. From a history. From a set of things you actually believe.
The model can help you say it better. It cannot tell you what to say. That part remains, stubbornly and I think permanently, yours.