The Industry's Consolation Word
AI is taking over execution, and the creative industry has settled on an answer: taste is what remains for us. That answer does not survive scrutiny.
2026 is the year AI took over a visible share of creative execution. Copy, layouts, campaign variants, interfaces: what used to cost days now costs minutes. The industry has found a consolation word for this, and it shows up in every second conference panel: taste. The machine produces, we curate. Execution gets cheap, so taste gets valuable.
It sounds reassuring. It has just two problems: it underestimates the machine, and it overestimates what taste ever carried as a business foundation.
Trainable Taste Is Consensus Taste
What emerges through training can, in principle, be learned by a machine. Taste that can be learned from data is average. From the machine we call it slop, from the agency we call it experience.
Taste is not a talent that falls from the sky. Taste is a trained muscle: you see a lot, compare a lot, discard a lot, and develop a judgment from it. That is exactly why the position is fragile. What emerges through exposure to many examples is precisely what language models are built on. They have seen more than any creative director who ever lived.
Today, the models are strong in statistics and analysis and weak in taste. Ask ChatGPT or Claude about design, strategy, or your own brand, and you get the average of every answer ever given: competent, plausible, interchangeable. But this state is a snapshot, not a protective wall. The trainable part of taste will be learned.
And here lies the uncomfortable core: taste that can be learned from data is consensus taste. It is the middle of what many consider good. When that middle comes from the machine, we call it slop. When it comes from the agency, we call it experience. The mechanism is the same; only the sender differs.
Honestly, taste would not have carried the industry even without AI. Taste cannot be checked or handed over. It has no address: it lives in people and leaves the building with them. A client who buys taste buys a claim.
What Clients Actually Pay For
When execution costs nothing, value shifts to the decision. And to the person who stands behind it.
When the tenth variant costs nothing, the variant is no longer the scarce good. Scarce is the judgment of which one is right. And scarcer still: someone who makes that call and stands behind it when it turns out to be wrong.
The industry's next word will therefore not be a capability but an address: accountability. Someone you can point to.
A judgment call is different from an opinion. You voice an opinion and move on. You make a judgment call, justify it, and carry the consequences. The difference is the business model: opinions are free, including from machines. Accountable decisions only come from someone with something to lose.
Responsibility Needs a Place
For accountability to be more than a professional attitude, decisions must become explicit: justified, dated, versioned, with a named sender.
Responsibility that is merely claimed is indistinguishable from taste. It becomes visible only when the decisions themselves become visible. A decision someone stands behind is justified, dated, and has a sender. And it is phrased so that you can later check whether it was followed.
Take tone of voice: "We sound premium" is a mood. "We say 'precision,' never 'quality,' for this reason, decided by this person, effective from this date" is a documented judgment call. Any machine can plausibly simulate the first. It can apply the second, but it cannot be accountable for it.
This is exactly where the circle closes with the machine-readable brand. A brand specification, as I describe it in Brand Infrastructure, is more than a format for systems: it is the place where judgments become explicit, versioned and with a named sender. Every rule in it is a documented decision. When AI takes over execution, the specification becomes the place where responsibility becomes visible.
What remains of brand work is not the better taste. It is the party that stands behind the decision.
The Consequence
For brands, this means: whoever documents their judgments can delegate without diluting. To teams, to agencies, to machines. The brand stays consistent, not because someone reviews every output, but because the decisions against which every output can be measured exist explicitly.
For everyone selling brand work, it means: taste as a promise becomes interchangeable the moment everyone promises it. Responsibility cannot be copied, because it is bound to an address. From now on, the question for every service provider and every internal brand function is: Who stands behind which decision, and where can I read it?
AI does not make judgment obsolete. It makes unaccountable judgment worthless.