AI will not democratize creativity. It will concentrate it.

Every AI tool launch arrives with some version of the same sentence: now anyone can make a film, build a brand, ship a product, run a studio out of their bedroom. The sentence is technically accurate. It is also the most strategically misleading thing being said about AI in creative work right now, because the part it gets right is the part that was never the bottleneck.

The bottleneck was, and still is, judgment about which output to keep. Nothing in the new tooling moves that.

The teenager and the director

A teenager in Lagos and a creative director in London both have Midjourney now. Both can generate a thousand images in an afternoon. Six months from now one of them is running a studio and one of them is not. The tool did not pick which one. The tool was identical for both of them. The thing that decided was the editorial floor under the work — the ability to look at the thousand images and know which three are interesting, and why, and what to do with them next.

The director had that floor before the tool existed. She built it over fifteen years of looking at things, getting things wrong, arguing about taste with people who were better than her. The teenager does not have it yet. He might build it. He might never. The tool does not help him build it. The tool helps him produce more of whatever he can already see.

The democratization story keeps refusing to name this. The cost of the artifact has collapsed; the cost of the decision has not moved. What used to be expensive — the rendering, the versioning, the cleanup — is almost free now, and what used to feel like a small thing on top of it — the editorial call — is now most of the value in the room.

What AI actually compresses

To work in creative production for any length of time before 2023 was to spend most of your time on things that were not the work. Rendering. Versioning. Cleanup. Rounds of “can you make the logo bigger.” Reformatting the deck for the third stakeholder. The actual decision-making — what should this be, what should it not be — was a small percentage of the hours, sandwiched between the crafts that took the time.

AI compresses the crafts. The rendering is faster. The versioning is faster. The cleanup is almost free. The reformatting is something the model can do while you sleep.

The decision-making does not compress. If anything, it inflates. Because the number of options the team is now being asked to choose between has gone from three to three hundred, and every option looks plausible enough to ship.

The studios that come out of this well are the ones that learn to say no faster. The ones that don't are mostly the ones that hired a person to operate the tool, shipped the first plausible output, and called it work because the dashboard could not tell the difference.

The hiring mistake

Most organizations I have watched adopt AI in the last eighteen months have made the same hiring mistake. They have hired the AI operator. The person who knows the prompts, the LoRAs, the workflow, the latest model. This person is real and useful. They are also the wrong centre of gravity for the team.

The role the team actually needs is the AI editor. The person who looks at what the operator produces, throws out 90% of it, sends the rest back with notes, and is responsible — career responsible, money responsible — for whether the final cut is any good. This role does not exist on most org charts. It is being absorbed, badly, by overworked art directors and creative directors who do not have the bandwidth to triple their throughput.

The cost of the missing role is the thing every leader can feel but most are not naming yet. The output is faster. The taste is decaying. The brand looks more like every other brand using the same model. The work is plausible everywhere and distinctive nowhere.

Why this concentrates rather than democratizes

The democratization frame imagines a flat distribution. Everyone gets a tool. Everyone produces. The middle layer of professionals disappears because the tool ate their job. The reality is the opposite shape.

The tool eats the middle of the production craft. It does not touch the top — the editorial call, the cultural read, the decision to make this thing and not that thing. It does not touch the bottom — the access to the audience, the trust capital, the distribution relationships that make any of this matter.

What happens, then, is that the people who already had the top and the bottom — the established studios, the founders with audiences, the institutions with distribution — now produce ten times as much for the same cost. And the people who only had the middle — the production talent, the executors — find that the market for what they sold no longer exists.

What gets called democratization is mostly consolidation dressed in the language of access.

What I tell people who are starting out

I get this question every week. Should I learn AI tools or should I learn fundamentals? It is a false binary, but the answer is heavily weighted in one direction.

The fundamentals, mostly. Composition, typography, hierarchy, rhythm in a sentence, why one shot of a Wong Kar-wai film carries the whole emotion and the next shot is just connective tissue, why a joke that should work on paper dies in the read-through. The model can produce a thousand images in an afternoon. The fundamentals are how you know which one of them is the picture and which 999 are exhaust.

Learn the tools. They are not hard. A weekend of focus and you are operating them at a working level. But do not confuse fluency with the tool for the thing the tool was supposed to produce. The cursor on the canvas is not the painting.

The honest version

AI is restructuring creative labor. What it is making more accessible is the production of plausible output, which is not the same thing as making creativity more accessible, although it gets sold as if it is.

The people who were already doing the editorial work — who could already look at a hundred options and pick the right one and defend the pick — are about to have a very good decade. The people who were doing the production work that AI has now absorbed are going to have a harder one, unless they move into editorial themselves. Some of them will. Most won't, because the move is harder than the seminars make it sound.

Democratization is the marketing claim. What is actually happening looks more like consolidation in slow motion — every studio quietly tripling its output, the production roles that used to make that output possible quietly disappearing, and the people who already had the audience getting a bigger one. I am not certain this is a stable equilibrium. I am reasonably certain it is the next five years.