The AI Agency Delusion

On WPP’s glossy vision and the messier reality ahead

The trouble with the future is that, in the right hands, it can be made to sound frictionless. WPP’s new whitepaper, The AI-Empowered Agency, is a case in point: a smooth, confident rendering of what happens when artificial intelligence seeps into every corner of the agency model. Six neat “principles” promise transformation without trade-offs. Deadlines that melt away, budgets that magically stretch, personalisation perfected en mass, talent that shapeshifts into multi-skilled polymaths. It reads like a permission slip to believe that everything we’ve ever wanted from this business is finally within reach.

The vision is alluring precisely because it removes all of the mess. The slow, stubborn parts of creativity that don’t compress well under the pressure of “faster, cheaper, more.” In WPP’s telling, these bottlenecks simply dissolve. AI expands what agencies can take on and shortens the time it takes to deliver, as if scope and scale could grow without limit, and without cost, whether in people or in overhead. It’s a future where the machine runs so smoothly you hardly notice it’s there.

But creativity has never been a smooth-running machine. It’s a field littered with false starts, inexplicable leaps, the sudden click of something that didn’t make sense for weeks. The moments worth having are often bought by the hours worth wasting. And those hours are rarely optimised. They are the coffee breaks that spark a sideways idea. The late-night rewrites that happen because something in the back of your mind wouldn’t let it go. The tangents, detours, and arguments that leave you exhausted but also, somehow, holding the right answer. These are not inefficiencies to be ironed out. They are the conditions for originality itself. Erase them in the name of acceleration, and you risk erasing the work’s very reason to exist.

And look, I understand why WPP has gone this way. There’s a high-profile change in leadership with a new CEO at the helm. Increasing pressure from the likes of Publicis, loudly touting their technological muscle. A significant account loss in Coca-Cola’s media buying. The need to come out swinging with clarity and a strong direction is obvious. But this, despite its polish, isn’t that.

So let’s break it down.

The first promise WPP makes is one of boundless capacity: AI allowing them to widen the funnel of what agencies can take on, while shrinking the time it takes to deliver. It’s the dream of doing more with less, but they’re calling it ‘transformation’. On the surface, who wouldn’t want that? But the danger in this idea is that it quietly redefines “more” as an unquestioned good. Agencies already have a habit of measuring value by output rather than impact, filling decks and feeds with a constant churn of deliverables. If AI grants near-infinite capacity without any guardrails, it risks supercharging this pathology, turning “more” into a substitute for “better” and leaving teams busier, more brittle, and further away from doing the kind of work that actually moves the needle.

That fantasy of infinite capacity blends seamlessly into their second claim: personalisation at scale. The language here is pure seduction - imagine a world in which you every audience, every segment, every individual gets a tailored message, perfectly matched to context, taste, and timing. Sounds great, right? But personalisation is not persuasion. A thousand algorithmically tuned micro-variants of an idea do not necessarily add up to a bigger cultural footprint. In fact, they can fragment the impact, creating a blizzard of “almost-right” executions, none of which land with the force of a singular, culture-shaping concept. And there’s an even subtler risk: when every message is personalised, nothing feels shared. The great campaigns of the past worked partly because they gave us common reference points. Personalisation done without a unifying narrative erodes those points of connection, leaving only isolated, forgettable fragments. There must be a balance.

From there, WPP moves to “unlocking polymaths” - this idea of arming creative talent with AI so they can work across disciplines, roles, and skill sets. It’s a vision of endlessly adaptable creatives who can shoot, edit, write, design, and strategise all in one breath. The problem is that this collapses the difference between capability and capacity. Just because you can do everything doesn’t mean you should, or that it will produce your best work. Specialisation exists for a reason. It’s how deep craft develops, how instinct is honed, how mastery is passed down. Agencies have already lost much of the old apprenticeship model; asking everyone to shapeshift until no one has a centre of gravity risks flattening the work into a series of shallow passes. And it overlooks the human cost: constant context-switching erodes focus, drains energy, and makes genuine creative immersion that much harder to achieve.

Their fourth pillar, “radical collaboration,” promises to dissolve silos through shared, AI-enabled workflows. It’s a nice vision, but silos in agencies are not simply structural; they’re cultural, political, and bound up in incentive systems that reward territory-holding. The idea that a new set of tools will fix that is optimistic at best. Without changing how success is measured, how credit is shared, and how leadership reinforces or dismantles fiefdoms, the technology will simply speed up the handoffs in the same old territorial game. And if you’ve ever watched “collaboration” in an agency under deadline pressure, you know that speed without alignment doesn’t lead to magic; it leads to friction and rework, just at a faster clip. i.e. clusterfuck.

The fifth claim - “data-driven creativity” - is positioned as the sweet spot where human imagination and machine intelligence meet. But there’s a gravitational pull here that shouldn’t be underestimated. AI thrives on pattern recognition, on drawing from vast back catalogues of what has worked before. The danger is that this can subtly nudge creative decisions toward the median, toward the safe centre of what the data says is likely to work. But the best creative work often comes from breaking patterns, not reinforcing them. We’ve already seen what happens when streaming platforms optimise for completion rates: content starts to converge into a narrow band of predictable beats. Advertising is just as vulnerable to that flattening effect. Unless leaders actively defend the space for genuine leaps of imagination, “data-driven” risks becoming “data-determined.”

Finally, there’s “always-on optimisation”. The promise of campaigns that learn, adapt, and evolve in real time. In performance marketing or iterative design, this can be incredibly powerful. But when applied universally, it risks trapping creative work in an endless state of tinkering, where nothing is ever finished, nothing is ever allowed to breathe, and no one ever gets to make the bold bet that can’t be A/B tested into safety. Creativity needs endings. It needs the moment of release, the exposure to risk, the chance to succeed or fail in the wild. Without that, you don’t get work that defines an era; you get work that just drifts endlessly in a shallow loop of micro-adjustments.

…And then there’s the part they don’t talk about at all - the underlying dependencies. Every one of these principles assumes that agencies will be in control of their AI destiny. But the truth is, the big AI players own the rails. Microsoft, Google, OpenAI, Anthropic; these are not vendors you just “partner” with. They’re infrastructure owners. The minute you build your workflows, your IP, and your client deliverables on top of their platforms, you become a renter, not an owner. Your margins and your capabilities are contingent on someone else’s roadmap, pricing model, and terms of service. The whitepaper doesn’t touch this, perhaps because it’s harder to wrap into a slide-friendly promise, but it’s the foundation everything else rests on.

And even if you solve for dependency, you run straight into the data problem. AI is only as good as the data you feed it, and most clients do not have the clean, structured, integrated datasets this vision assumes. They have fragmented CRM systems, half-finished tagging taxonomies, and years of ungoverned content that no one has had the time or budget to wrangle. Without addressing that, “personalisation at scale” is just a nice headline. You can’t optimise what you can’t measure, and you can’t measure what you can’t even find.

Underneath it all, what sits behind these six principles is a bigger, quieter assumption: that friction is the enemy. The pauses, the blind alleys, the deep silences, the imperfect human hand-offs - these are all cast as inefficiencies to be eliminated. But in reality, they’re the conditions under which originality, depth, and cultural resonance emerge. Strip them away, and yes, you can speed up the process, but to what end? The only one I can see in this scenario is that you turn the work into something predictable, weightless, and interchangeable - a fast track to the complete annihilation of everything this industry promises.

………….

This is not to say that AI has no place in the agency model. In fact, used well, it can be genuinely magical; expanding creative range, unblocking production bottlenecks, and opening new ways to work with ideas. But magic is dangerous when it’s unbalanced. The right question isn’t “how can we use AI everywhere?” It’s “where should we let AI in, and where should we protect the friction, the slowness, the human awkwardness that makes the work worth anything?” That question doesn’t have the same glossy certainty as WPP’s six principles. But it has the advantage of being real.

This is existential work. It’s messy, it’s hard, and it refuses the comfort of frictionless futures. Overly simplified, hyped-up promises won’t cut it. The agencies that survive what’s coming will not be the ones who buy every tool and dissolve every boundary, all in the name of efficiency and optimisation. They’ll be the ones who treat this moment not as a race to automate the shit out of every possible step, but as a fight for the soul of the work itself.