journal · on craft
Creative direction in the age of AI
Agents can make almost anything now. They still cannot tell you what is worth making. That gap is the whole argument, and it is where creative direction in the age of AI stops being a cost and becomes the point.
The cheap thing is the work. The rare thing is the judgment.
For most of the last decade, making the work was the bottleneck. You wanted ten cuts of a campaign, you paid for ten cuts of a campaign, and the budget set the ceiling on how much you could try. That ceiling is gone. An agent will give you the ten cuts, then forty more, faster than you can watch them.
Which means the scarce resource flipped. Volume is now free. Taste is not. The question stopped being can we make it and became should this exist at all, and is it the best version, and does it sound like us. None of those are production questions. They are direction questions, and they are the only ones left that are hard.
Do founders still need a creative agency when AI can make anything?
The honest answer is that founders need the agency for a different reason than before. You no longer need us for hands. You can generate a hundred posts in an afternoon, and many of them will be competent. Competent is the new floor, and the floor is crowded.
What a founder needs now is someone to stand at the other end and say no. No to the eighty that are merely fine. No to the on-trend thing that is not yours. The agency that earns its place is the one selling the decision, not the deliverable. Yes, or no. In writing. That sentence was always the product. AI just made it obvious.
More output, without more people
The old way to make more was to hire more. More designers, more editors, more coordinators to keep the designers and editors pointed in the same direction. Output and headcount rose together, and so did the drag.
Agents break that line. The volume comes off the same brief, run by a system that does not get tired and does not drift, so output grows without the team growing with it. What stays human is small and deliberate: someone sets the direction, someone signs the work. A media department that never sleeps, governed by a few people who decide what it is allowed to mean. That is not a smaller team doing more. It is a different shape entirely.
The machine has no point of view, and it never will
It is tempting to wait for the model that has taste. It is not coming, not in the way people imagine. A system can learn the average of everything that has worked and hand you the center of that distribution on request. The center is exactly where nothing memorable lives.
Taste is a series of refusals. It is the willingness to throw out the clean, correct, reasonable option because it is forgettable, and to defend the stranger choice because it is true. That is a human act, made by someone with something at stake and a reason to care. The machine optimizes toward agreement. Direction is the discipline of disagreeing with the average on purpose.
The human role is the luxury, not the apology
For a while the industry treated human involvement as a thing to explain away, a cost you tolerated until the tools caught up. We think that is backwards. When the work is abundant and nearly free, the human at the front becomes the rarest part of the whole arrangement, and the most valuable.
This is the inversion worth sitting with. The agents are the scale. The person deciding what the agents are for is the luxury. Humans set the direction, agents do the volume, and the reason that order matters is that it is the only order where the work still means something. We keep a select few clients for exactly this reason. Direction does not divide cleanly across a hundred accounts.
What this looks like in practice
None of this is theory for us. It is how the studio runs. Agents make the volume on a single brief, every channel fills from the same source, and every step is on the record, yours to read at any time. The taste lives where it always did, in the people who decide what gets made and what gets killed.
If you want to see what that division would actually do for your business, we built a quiet way to find out. We point our agent system at your company and report back, plainly, what is worth making and what is not. Read by Luka, every time. No pitch attached to it.
When you want a real read on what agents should make for you, and what they should not, we wrote one down.
start the audit