In my opinion, the uncomfortable answer is: a lot of it.

And maybe we're asking the wrong question.

AI can already write the copy, create images, make videos, generate dozens of concepts in minutes and produce hundreds of personalised variations. And it's getting better frighteningly fast.

If we define a creative agency as a team that turns a brief into creative executions, then yes, AI is coming for the job.

But that's not quite what a good creative agency does.

The hardest part of advertising was never producing the ad. It was figuring out what the ad should say.

Why should it matter? What's the human truth? What makes someone stop? What's culturally relevant? What's the idea a competitor would never think of?

And perhaps most importantly: which of the 10,000 possible ideas is actually worth making?

AI will make creative production incredibly cheap. When it costs almost nothing to produce 1,000 ideas, ideas are no longer the scarce resource.

Judgement is.

Taste. Strategy. Originality. Knowing what not to do. Knowing when a technically brilliant execution is completely wrong for the brand. And having the guts to say: β€œNo. It's not good enough.”

There's another uncomfortable consequence. If AI makes production 10x cheaper, clients probably won't keep paying agencies simply to produce more stuff. The economics will change. Junior production roles will evolve. Agency pyramids will flatten. Agencies built around hours, headcount and production volume will be under pressure.

But I don't think creative agencies disappear. They evolve.

Less factory. More creative brain trust.

Less: β€œGive us the brief and we'll make the campaign.” More: β€œLet's figure out what the brand should actually be saying.” Then use AI to execute that idea at a speed and scale that would have been impossible five years ago.

Which leads to a strange conclusion: AI may make great creative agencies more valuable, and mediocre ones less necessary.

The creative agency of 2031 will look very different from the creative agency of 2026.

And I'm not sure that's a bad thing.