The Future for the Ad Agency Strategist

From Lone Wolves to SWAT Teams

This is a story about advertising, but it is more than that.

It is about what happens to a profession when the word it uses for its most important work drifts free of the duty the word was invented to describe, and what happens when a new technology arrives to audit that drift.

The advertising industry ran this experiment on a smaller stage than most, and on a shorter clock.

Which is why the collapse is already further along there than in most of the places it will eventually arrive.

I. The Severing

The story begins in London in the late 1960s. Stephen King at JWT and Stanley Pollitt at BMP were trying to solve the same problem from opposite ends of the same office.

Agencies knew things about consumers, but the knowledge lived in research departments that only surfaced when someone remembered to ask. It was backwards-facing. Briefs were written in the language of the client’s marketing plan rather than the language of the person the advertising was supposed to move.

King’s answer was a new kind of brief. One that forced the agency to say, in a sentence, what the advertising had to make the audience think, feel, or do. Pollitt’s answer was a new kind of person in the room. Someone whose only job was to represent the consumer against everyone else’s assumptions.

They called that person the Account Planner.

A modest title on purpose. It did not claim authority. It claimed a duty.

When planning crossed the Atlantic in the early 1980s, it arrived as a specific, disciplined practice. Then the title flipped. Planning departments were renamed Strategy. Planners became Strategists. The stated reason was positioning. Strategy sounded like the room where decisions got made, rather than the room where the briefs got written.

The real reason is that “Strategist” can be priced higher.

By 2003, the American industry had almost completely swapped the vocabulary.

What the flip actually did, the part the industry spent twenty years refusing to notice, was sever the word from the duty.

Pollitt’s planner existed to represent a specific party against a specific pressure. The consumer against the room. A strategist, in the inflated sense, represented nothing in particular and therefore everything in general.

The title grew. The responsibility dissolved. Agencies began hiring Associate Strategists and Group Strategy Directors in numbers that bore no relationship to the amount of actual strategic work in the building. Decks replaced briefs. Frameworks multiplied. Insight came to mean any statement about consumers that sounded slightly surprising, whether or not it was true.

II. The Audit

Then the floor went.

Tens of thousands of people are now leaving agency holding companies in a contraction deeper than the 2008 cuts. AI is the accelerant, not the cause.

The cause is thirty years of title inflation, meeting the first technology that can cheaply produce the specific artefact the inflated role produced. The plausible strategy deck. A modern language model can generate a competitive audit, a consumer segmentation, and a brand-architecture diagram in roughly the time it takes a human to open Keynote. The output is reliably good enough to expose how much of the human version was also merely good enough.

For the strategists whose job was the deck, AI is the solvent dissolving the role in real time. It is not a threat. It is an audit. The industry is discovering which parts of its work were load-bearing and which parts were theatre. The theatre is what the machine replicates first.

This is a useful thing about being audited by a language model.

A language model is a machine for producing the generalized mean of everything that has already been written.

Whatever it can produce was already in the mean.

Whatever it cannot produce is, by definition, the part a client pays premium rates for.

III. The Return of Ground Truth

What the machine cannot produce is anything that existed only in a specific room.

Citrini Research recently sent someone to the Strait of Hormuz with fifteen thousand dollars in cash because they wanted to know what was actually happening there, rather than reading the third-hand version on a Bloomberg terminal.

The report they filed was not a scoop. It was a first-hand description of the texture of a place.

Expensive, slightly dangerous, almost entirely unautomatable.

This is the move the advertising industry taught itself to stop making.

There was a time when planners went to people’s homes and stayed until the wine worked and the respondents stopped performing.

The two hours were the cost of being in the room for the ten minutes that mattered.

You cannot prompt that out of a language model.

The absence of cost is the absence of ownership.

When you haven’t sat in the living room yourself, you have no skin in the game and no conviction to defend a leap against a skeptical client.

Ground truth is not always dramatic. Most of it is boring. Riding along on a sales call, sitting through a long interview, spending a day at a retail location, watching the thing the deck claims is happening actually fail to happen.

The Hormuz trip is the extreme case.

The mundane case is just as load-bearing and more common.

What they share is the part that the machine cannot reach.

Specificity that could only have come from being there.

IV. The New Unit

The strategist whose job was making the deck is over.

The strategist who goes to the living room is not.

But neither is the cultural reader, or the numbers person who knows which metric is load-bearing, or the brand integrator who holds the whole picture in one coherent frame.

All four of them survive the audit.

None of them survives it alone.

This is the part the industry keeps missing.

The answer to the collapse is not a better solo strategist. It is not a returning hero.

The solo role is what the machine replicates first, because a solo role is a single stream of output, and output is the part a language model is built to produce.

A unit is different. A unit produces something no single stream can. Four angles on the same problem, argued out in real time, with a first-hand witness in the room.

That is not a workflow.

It is a kind of conviction the machine cannot simulate, because the conviction comes from the friction between four people who each know something the others don’t, defending it to each other before anyone defends it to a client.

The Data person knows which number matters. The Cultural Strategist knows what the number means in the wider weather. The Brand person knows how it fits the thing the company actually is.

The Why? person knows what the number feels like on the ground because they were there when the wine was working.

A room with all four in it is producing something the machine cannot reach.

A decision that has been stress-tested by four different kinds of ownership before it leaves the building.

That is the artefact clients will pay premium rates for, because that is the artefact they cannot get anywhere else.

Pick a lane.

Be honest about which one.

Nobody is all four, and the people who still claim to be are the ones who were already being audited when the audit arrived.

Find the other three.

They are already in your network.

The problem has outgrown the solo role because the solo role was always a performance of integration that one person could not actually perform.

The unit is not a workaround for that failure.

It is the form the work should have taken all along.

The audit is clearing the stage.

What walks onto it next is the SWAT.

---