My mother, of all people, sent me an alarming story from her favourite newspaper informing the world that “UK ad agencies undergo their biggest exodus of staff as AI threatens industry”.
Oh dear, it’s happening. The singularity is upon us. Bots are replacing all human endeavour. In fact, I think I remember there being a prequel to The Terminator in which Skynet won Sales Team of the Year at the Media Week Awards.
Or maybe… this story is a load of garbage?
Not a conspiracy. Just some really simple maths
Even the standfirst is wrong: “Number of employees declined by more than 14% to 24,963 last year, with fall greatest among younger workers.”
Except, according to the IPA Census on which the story is based, UK agency headcount fell from 26,787 in 2024 to 24,963 in 2025 — roughly a 7% drop. Not 14%. (And if you think 7 percentage points is splitting hairs, that’s 900 people’s jobs you’re casually dismissing as nothing — you monster…)
If we’re being charitable, let’s say the publication is playing fast and loose with the term ‘ad agency’, which is what people often think of when they think of a ‘creative’ agency, whose staff the IPA says are down 14%.
Media agencies (you know, the ones historically looked down on for being spreadsheet monkeys and sales oiks), actually saw their staff members increase by 2%. Despite, you’d think, their businesses being more vulnerable to AI and automation. But we’ll get to that.
Here’s what just under 25,000 looks like compared to the same Census data going back to 2017. Show me the exodus.
To make it even easier, here’s the same data presented as a year-on-year percentage change:
If this is an exodus, it’s an oddly tidy one. And it conveniently ignores the fact that IPA member agencies grew by 19% in 2022. Or the fact that 2025 employment is almost exactly where it was before Covid. The real outlier in that dataset isn’t the decline — it’s the surge between 2022 and 2024. Agencies went on a post-lockdown hiring spree. Now they’re reverting.
That looks less like technological displacement and more like a cyclical correction.
And it gets worse.
How can you report on job figures in the ad agency sector without even mentioning that two of its biggest employers merged and shed literally thousands of jobs?
In December, we learned Omnicom would cut more than 4,000 jobs globally as it integrated Interpublic Group. That followed earlier IPG reductions of around 3,200 staff and prior Omnicom cuts in the thousands.
And then there’s WPP, which had axed 4,000 jobs in 2025 before the new school year had even started. Rather than “AI” being the culprit, this company lost its CEO and is about to announce the close of even more agency brands because it had lost a series of international media accounts in recent years.
I can’t give you exact numbers for how many of these global holdco cuts are responsible for the 1,800 that were lost in the UK last year. But I reckon it’d be worth a mention.
And it’s certainly more likely an explanation than “AI”.
The real story is not a new story. Which means, on planet Earth, it’s not a story
You could spin the same arithmetic the other way: despite the largest network consolidation in modern agency history, total UK headcount is broadly back to its pre-pandemic norm.
But that wouldn’t fit the ‘AI-is-eating-adland narrative’.
To be clear: there is an important trend around entry-level roles, because AI tools can automate some junior tasks. I hear enough from industry leaders to accept that this is directionally true (or, at least true-ish). But young people are always at the sharp end of labour-market shifts. In 2007, before the financial crisis was fully visible, graduates could already feel the air coming out of the job market. Downturns hit juniors first. That isn’t new.
The more uncomfortable truth is this: the holding-company model is broken.
Because in a commoditised market, scale does not create defensible value (a “moat”) — it creates pricing pressure. When media buying, social production, retail media planning and influencer management are interchangeable line items on a procurement spreadsheet, consolidation does not make you more differentiated; it makes you easier to benchmark. Clients see large agency groups making “synergies” and demand lower blended rates. Public markets see softening revenue and demand higher margins. And when margin becomes the objective function, headcount becomes the most convenient lever.
None of that requires “AI”.
Meanwhile, brands are in-housing. Smaller independents are growing. Work is fragmenting. Some of those companies may, ironically, be using AI more deftly than the legacy networks and AI could be helping create jobs in advertising and media. Some may be creating more output with fewer people. Some may even be hiring. But they may not sit inside the IPA’s membership universe. They may not show up in the census. A shift in where work is done is not the same as work disappearing.
This is why this story matters.
The rise of AI ‘vibe reporting’
There is a growing genre of what the author and computer science academic Cal Newport calls “vibe reporting”: identify a culturally resonant fear — in this case, AI taking white-collar jobs — then assemble loosely related data points and suggestive quotes until the narrative feels inevitable. No explicit causal claim is made. But that’s fine, none needs to be: it’s like the Only Connect game where you guess the well-known phrases with no vowels: such as WHT LD F BLLCKS.
If you’re hipster enough to subscribe to Ad-verse Reactions, you’ll know already that virtually all consumer newsbrands stopped long ago trying to report seriously on advertising. I really should be commending The Guardian for even bothering to read an IPA press release, let alone take the time to write something (anything!) about what’s actually happening in the industry that keeps its journalists in gainful employment.
But when we in the industry start accepting correlation dressed up as causation, we are doing the machines’ work for them.
If AI really is replacing creative labour at scale, show the campaign it replaced. Show the P&L line where human cost was permanently substituted by machine output. Show the structural break in the time series.
A single-year dip back to pre-pandemic headcount levels is not that.
Misdiagnose the cause and you misdiagnose the solution. If agencies convince themselves that AI is the primary threat, they will invest in tools instead of confronting broken economics. If young people believe the industry is already automated, they may exit prematurely. If we accept suggestive narrative over demonstrable evidence, we erode the very human skillset we claim to value.
And that leads to the uncomfortable conclusion.
If we can’t distinguish correlation from causation (or even do basic maths), then maybe we deserve to be automated.
This article first appeared in Ad-verse Reactions, a newsletter written by independent journalist and consultant Omar Oakes, covering the economics, power structures and unintended consequences shaping advertising and media. You can subscribe to Ad-verse Reactions for regular analysis at omaroakes.substack.com.
