The Era of Agencies Is Not Over: It Is Being Edited

Every generation of agency leaders has watched a moment that was supposed to be the end. Digital was going to make agencies irrelevant. Programmatic was going to cut them out entirely. Social media was going to democratize creativity past the point where professionals added value. Each time, the prediction landed somewhere between half right and half wrong — in the way that matters. The agency as a specific configuration of people and processes changed. The underlying need it serves didn't.

This moment is different in one real way: the technology is more capable. AI can now produce competent work across a broader range of functions than any previous tool. That's genuinely new. But it surfaces a question that has always been the most important one in this industry: what is the difference between competent and exceptional? And who is responsible for the latter?

There is a thought experiment worth sitting with. An agency made a widely covered announcement that they had trained their AI model on 200 years of advertising history. The implication was that this produced something uniquely powerful. The reality is more complicated.

This is not a dismissal of what AI can do. It's a precise description of what it can't. A model trained on everything the Renaissance produced — every painting, every technique, every compositional convention — would give you the best possible version of what was already known. It would not give you The Last Supper. The deviation, the break from what existed, the thing that had no precedent in the training data — that's the part that changed art history. And it's the part that no amount of historical data can generate.

What AI trained on existing advertising will reliably produce is existing advertising, optimized. For much of what agencies do at volume, that's genuinely useful. For the work that changes categories, repositions brands, or creates the cultural moments that drive long-term business value — the work that clients actually remember paying for — it's a starting point, not a destination.

The human advantage in creative work isn't romantic. It's structural. The breakthrough — in science, in art, in strategy — almost always comes from outside what's already known. It comes from a connection that didn't exist in the dataset, from an observation made in the real world rather than derived from historical patterns, from someone who looked at the category and saw something everyone else had normalized.

That capability — what might be called the synapse leap — is what the agencies that survive this period will protect and invest in, even as they automate everything around it. The logic is not sentimental. As AI makes competent work faster and cheaper to produce, competent work loses its value. What becomes more valuable, not less, is the work that AI can't produce. The insight born from a real human experience. The strategic move that reframes the category. The creative decision that makes a person feel something they didn't expect.

The problem with craft is that it doesn't develop automatically. It gets built slowly, through years of immersion in the work, through proximity to people who are better at it, through making things that fall short and understanding why. If agencies eliminate the conditions for that development — by cutting entry-level roles, removing mentorship, turning creative and strategic work into a series of prompts — they will find themselves with efficient tools for producing competent work and no one who knows how to produce exceptional work at all.

The agencies that are thinking clearly about this are not choosing between AI and craft. They're choosing to use AI to eliminate the work that was never craft to begin with — the production, the formatting, the scheduling, the research synthesis — and to reinvest that time in the thinking that actually differentiates them.

The agencies that look back on this moment as a turning point will share a characteristic that has nothing to do with their AI stack. They will have been genuinely good at thinking like their clients. Not thinking about their clients — modeling their competitive reality, their organizational dynamics, the decisions they face that have no clean answer. That capability is what makes an agency indispensable, regardless of the technology environment around it.

The agencies without it — still selling time, still competing on output volume, still defining their value by the size of the team — face a version of this moment that is genuinely more threatening. Not because AI is exceptional. But because they have never clearly articulated why they are.

The era isn't ending. It's editing. And the agencies that understand what needs to stay are the ones the next version is written around.

Back to Blog