Alfredo Roque Gameiro,The Conquest of Lisbon, 1917
A shorter version of this piece first appeared in Campaign on 07.10.26.
The argument is familiar. AI can generate content at scale, but it lacks the sense that a human with a pulse is behind the work.
"AI might handle 80% of the grunt work, but that last 20% — the strategic, emotional, creative layer — is all human."
"AI can scale content. But only humans can scale connection."
"Brands are more than visuals; they are personalities. And personalities cannot be automated."
"The human touch is essential in cutting through the noise and keeping audiences engaged."
"Clients still want authentic, people-generated ideas."
"Human-led storytelling is still key when it comes to creating emotionally resonant, on-brand content."
"It's presence — the felt sense that someone with a pulse is actually behind the words."
The ceiling and the floor
Every version of this argument defends “human creativity” by invoking its ceiling — AI cannot do match the best this industry has ever produced — Just Do It, Think Different, A Diamond Is Forever, We Try Harder, Vorsprung durch Technik, Smell Like a Man… 1984, Launderette, Double Life, If You Let Me Play, Cog… This, however, is like pointing to the likes of Shakespeare, Austen and Borges and arguing that AI can never replace the human author. It’s a sleight of hand that passes off the exception as representative of the whole, and forgets (or chooses to ignore) where the bulk of human activity and output actually lies.
Instead, it invokes the exceptions and declines to look at what makes up most of the volume: the derivative, the safe, the generic, the formulaic. Literature has its Dostoevsky, its Eliot, its Joyce, its Márquez. But the best-selling genre in American fiction is romance — nobody is going to build a new theory of the novel’s potential from it, and the chances of these books being read in two hundred years’ time are, I’d wager, slim.
Of course professional inspiration, hope, and motivation are to be found in what is excellent. But the case for the importance of the human touch in advertising seems to float in a kind of parallel reality, untethered from the vast majority of work that actually runs.
On the tube platform I stare at a cross-track poster: the kind of gurning young man who only exists in ads gazes with orgasmic joy at a pint as someone tops it up. The line says something banal about simple pleasures and I wonder where the felt sense that someone with a pulse is behind the work has gone.
I search for a tutorial video on how to fix a Mac glitch and am forced to watch a skull-crushingly awful ad for a toilet cleaner that is so devoid of the human touch essential to keeping me engaged that its makers have ensured I cannot skip it.
I keep reading that the risk is AI will enable us to produce mediocrity at scale. We already do.
Ninety percent of everything
Excellence is, by its very nature, unevenly distributed. Theodore Sturgeon, the science fiction writer, observed in 1958 that ninety percent of everything is, in his words, “crud”. He was talking about his own genre, defending it against literary critics who pointed to its worst examples, and his point was that the same ratio applied everywhere — to film, to fiction, to music, to art.
Sturgeon’s Revelation, as it became known, has held up rather well, and there is no evidence that advertising is exempt from it. Mostly crud is not a bug but a feature, the inevitable outcome of creative output at scale. If anything, the scale has made it worse. Digital channels have multiplied the volume of advertising many times over, but there is little to suggest they have multiplied the amount of genuinely excellent work. Indeed, the suspicion is that the same small amount of brilliance is now spread across a vastly larger pile of output. Sturgeon’s ninety percent may well be a generous estimate.
Claiming creative excellence on the basis of the minuscule percentage that rises above the average is like an indifferent partner claiming to be a loving one because they remember Valentine’s Day. The gesture is real, the roses are real, but one annual act of attentiveness does not describe a relationship. It describes the exception that the other three hundred and sixty-four days contradict.
The Golden Age that never was
AI then does not threaten to accelerate a decline from some prior pinnacle of excellence. And let’s say it and get it out of the way now. There was never a Golden Age. Sturgeon’s Revelation applied to the sixties (or seventies or eighties or nineties — pick your decade) as reliably as it applies to today.
The difference is not that the advertising industry produced more exceptional work then and less now. It’s just that the work that survived from those decades survived precisely because it was exceptional — because Bernbach and Gossage and Krone and Lois were operating so far above the average output of their era that their work was worth keeping.
What didn’t survive, because nobody valued, archived, or posted about it, were the utterly formulaic print ads, the forgettable television commercials, the generic radio spots that surrounded the exceptional and vastly outnumbered it.
More like nothing
The fact of the matter is that AI is just making newly visible (and vulnerable) the vast ocean of work that never needed a human mind to produce it. The primary challenge AI presents to the advertising industry is not one of technological adaptation and adoption. The machine is simply revealing what a capable machine can now conceive and produce, cheaper and faster. The displacement of that work is not a threat to the creative heartland — it is, at most, an efficiency correction.
Mediocrity — by which I mean the predictable, formulaic, genre-bound, cliché-recycling, expectation-confirming, best-practice copying-and-pasting — has no defence against the machines. Because that’s precisely the property a machine can replicate: pattern-completion is what the machine does.
Consider that in June 2025, an AI-generated band called The Velvet Sundown released two albums of vintage-sounding folk-rock on Spotify. Within weeks, listeners were reporting its tracks on their Discover Weekly playlists, with no tours, no press and no visible human presence behind it. At its brief peak, swollen by the controversy it had caused, The Velvet Sundown had around 1.4 million monthly listeners.
It’s not a story about AI producing art of genius — The Velvet Sundown are nothing more than the kind of competent, perfectly listenable music that fills the median of every genre, gets added to playlists, is played on Sunday morning and is utterly forgotten by Monday. As Ian Bogost, writing in The Atlantic, put it, “You know, it’s not bad. It’s not good either. It’s more like nothing — not good or bad, aesthetically or morally.” He could have been writing about advertising.
Everything’s gonna be all right?
George Carlin got it right when he wrote “The real cynics are the ones who tell you everything’s gonna be all right.” The human element is the advertising industry telling itself that everything’s gonna be all right. It is the most comforting and the least useful thing it could say.
Something worth defending
But framing the encroachments of AI as simply an efficiency correction keeps the argument inside the AI debate when the tougher (and more ambitious, more exciting) question is what it would take to build something worth defending.
I’m a humble brand strategist, not a business strategist, not an organisational transformation consultant, not a commercial model architect, and have no expertise in restructuring agencies, repricing retainers, or redesigning operating models. But it strikes me there are some useful questions to ask — ones I think an agency serious about this should be willing to ask itself honestly.
The audit
What if we took the last twelve months of the agency’s output — all of it — and with ruthless honesty asked of every piece — what did this require that a capable machine could not have provided? Or will not be able to provide in six months’ time?
Applied honestly, the audit is likely to be clarifying and/or uncomfortable. The temptation will be to automate what it exposes. But the point is to close off every avenue towards producing it.
The conditions
What if we then interrogated what survives the audit — what conditions produced it? The close reading of a real person’s lived experience we call ‘insight’? Client trust accumulated over years rather than goodwill extended over quarters? The refusal of the brief as delivered? A team willing to walk in stupid? A director who made the ordinary extraordinary?
Reengineering the conditions is no guarantee. Changing them can enable latent or overlooked talent to rise to its full potential, but the audit will also reveal that genuine creative excellence (as opposed to capability) is spread considerably thinner than the headcount suggests.
The hours
What if we then dared to ask if AI frees up human hours, who decides what they’re for — and if we can afford the answer?
In a time-based model the hours the machines free are not the agency’s to reinvest, but the client’s to reclaim. Which makes an outcome-based model the precondition for doing the work differently — as well as a fairer way to charge for it. Without it, the efficiency gains flow to the client as shrunk timelines, shrunk fees — and nothing about the quality of the work changes.
“Procurement won’t buy it” will be the last words of the agencies that waited too long to find out.
The thinking
And what if we then asked if we are using AI to think differently, or just to make the same things faster? Not just in the development of strategy, but in the development of creative ideas.
The question requires abandoning the most comfortable version of the AI story in which the machine handles the grunt work and the human handles the ideas. The story is reassuring because it changes nothing. And efficiency will not survive as a moat.
But put AI in the hands of a talented creative mind as a real thinking partner and their ceiling moves. An agency that invests in that builds what the time-based model never could keep — a practice that deepens with use, and an advantage that grows the longer it runs.
The difference
The questions guarantee nothing. But they would, I suggest, expose what the human-element argument conceals — that defending the heartland against the machines is protecting a vacancy. Understanding why so much of our output never needed a human and being willing to change the conditions that produced it is how we might build something in its place actually worth defending — and vastly more valuable to client organisations.
The inconvenient truth
The machines are not coming for the exceptional work that could only have sprung from the weirdness and wildness of an embodied human imagination operating in the Premier League of creativity, but for the work that never needed a human in the first place. The question is whether we had the honesty, before they took it, to know the difference.