The constraint on creative production has flipped. For years the limit was capacity: how much a team could design, write and produce. Generative tools have largely removed that limit, and the new scarce resource is judgment — the direction, taste and brand discipline needed to keep a flood of machine-made content distinctive and on-brand. Producing the asset is no longer the hard part. Making sure it is the right asset, and unmistakably yours, is.

This is most visible at the enterprise end, where the tools have matured fast. Adobe’s Firefly has passed 29 billion generations, and its newer products are aimed squarely at producing brand-consistent content at volume rather than one-off images — a shift that reframes what a creative team is actually for.

What the tools now do

Adobe’s Firefly Foundry lets a business train private generative models on its own proprietary assets, producing brand-protected outputs intended to be ready for external use. Its GenStudio for Performance Marketing pairs those custom models with brand guides to generate campaign content at scale, and a new Firefly AI Assistant pushes toward what Adobe calls creative agents. The company’s own framing of the new division of labour is blunt: marketers provide the direction, creatives own the craft, and agents handle the grind.

Adobe is not alone — Google’s Asset Studio and Meta’s generative ad tools chase the same goal — but the pattern is consistent across all of them. The selling point is no longer just generation; it is governed generation: custom models trained on brand assets, outputs validated against brand guidelines, and provenance metadata attached. The vendors have understood that producing content was never the real problem for established brands. Producing content that stays on-brand, at volume, was.

Why the bottleneck moved to governance

Tools like Adobe’s Brand Intelligence exist precisely because infinite generation creates a new risk: not too little content, but too much of the wrong content. When anyone can produce a thousand variations in an afternoon, the danger is a flood of material that is subtly off-brand, factually loose, or so generic it could belong to a competitor. Volume without governance is a liability, not an asset.

That makes the brand system itself the critical input. A model can only generate on-brand work if the brand has been defined clearly enough to train on and validate against — codified guidelines, a coherent visual and verbal identity, an organised library of assets. Brands that have done that groundwork can point the tools at it; brands that have treated their identity as something held loosely in a few people’s heads have nothing precise for a model to learn. The unglamorous work of writing the brand down has become a competitive input.

What it means for a creative team

The instinct to read this as the end of creative roles is the wrong one. Production tasks do compress, but the work that decides whether output is any good — setting the direction, exercising taste, governing the brand, catching the off-brand or the bland before it ships — becomes more valuable, not less. The skill moves up from making the asset to judging it.

For a B2B team, the practical priorities are clear enough. Invest in codifying the brand so the tools have something precise to work from. Put human curation and direction at the centre of the workflow rather than at the end of it. And treat distinctiveness as the thing to protect most fiercely, because the real hazard of an industry all drawing on similar models is not bad content but sameness — a market of brands that have each generated their way into looking like everyone else.

AI has removed the excuse that there was no time to make the content. What it cannot supply is the judgment about which content is worth making and whether it actually sounds and looks like you. The open question is whether brands use that freedom to be more distinctive, or quietly converge on the same machine-made middle.