Talent: The Media Lens - 4As

Developing Next-Gen Talent

An agent can run the auction. It can't explain itself to a regulated client.

Machines are great at recognizing patterns and running tests. But a campaign involves more than a bid: an audience, a market, a set of exclusions and sometimes a client who needs to know why a decision was made before it ships. In pharma and other regulated categories, "the model decided" isn't an answer. Advertisers seem to agree. Only 6% say they almost always act on AI recommendations surfaced in their platforms.

Comfort with AI falls from 90% when it recommends and humans decide to 50% when it acts on its own, even with proven performance (StackAdapt, The AI Delegation Gap(opens in new tab)).

There's another way to look at it: if these systems were truly doing the math better on their own, performance would plateau far less often than it does.

That changes the advice for someone starting in media in 2027. The tedious work is going away, and that's a good thing, but that work was also how people learned to read a system well enough to question it. The goal for new talent is to understand as many parts of marketing as possible, because that breadth is what makes someone capable of directing agents or teams of people and agents working together. If you can't explain what the machine did and why, you can't really say you're managing it.

Back: Policy Watch

Next: What's Next Now for Agencies