There’s a very particular kind of agency presentation doing the rounds right now. The fonts are clean, the diagrams glow and somewhere around slide 17 there’s an impressive-looking ecosystem showing how strategy, creativity, data and production now flow seamlessly through a proprietary AI-enabled workflow. Words like “accelerate”, “augment” and “orchestrate” appear with suspicious frequency. Somebody has named an internal chatbot. Somebody else has probably put a little sparkle icon next to it.
It all looks very 'modern'. The problem is that it also looks remarkably similar to the presentation from the agency down the road.
That might be the more interesting problem facing AI in creative industries right now. We’ve spent the last few years asking whether artificial intelligence will make agencies faster, cheaper or smaller, whether it will replace jobs and whether clients will eventually decide they can make everything themselves. All perfectly legitimate questions, obviously, but perhaps there’s another possibility sitting quietly underneath all of them. What if AI isn’t necessarily making agencies worse, but is making some of them harder to tell apart?
That’s a much less dramatic proposition than the death of the advertising agency, but it might prove considerably more useful. AI can make an agency ferociously productive. It can help people research faster, prototype ideas, generate variations, organise information, localise campaigns, automate repetitive production and get a surprisingly respectable first draft onto a screen before somebody has finished saying, “Shall we get a coffee and kick this around?” None of that is trivial. The difficulty is that everybody else is gaining access to much the same capability.
The UK creative industries are already adopting AI faster than the wider economy, with 51% of creative businesses now reporting some form of AI use, while research in the agency sector shows similarly widespread adoption. What was genuinely distinctive capability a couple of years ago is rapidly becoming ordinary infrastructure. And once a capability becomes ordinary infrastructure, simply possessing it stops being much of a competitive proposition.
So if everybody has the same new superpower, it isn’t really a superpower anymore. The interesting question becomes what your agency chooses to do with it.
How AI Is Changing the Way Creative Agencies Work
We Are Batch
The AI impact on advertising agencies is already much deeper than adding Midjourney to the art department and ChatGPT to the strategy team. AI is moving into almost every layer of agency life, from research and strategy through ideation, visualisation, production and delivery. Research documents can be compressed and interrogated in minutes. Meeting transcripts become usable summaries before everyone has left the room. Concept routes can be visualised long before a production budget exists, while copy can be adapted across channels and audiences almost instantly. Storyboards, treatments, layouts and prototypes can now be created quickly enough to turn a vague conversation into something tangible before enthusiasm has had time to evaporate.
Production is changing just as dramatically. Versioning, localisation, clean-up, resizing and routine asset creation are all becoming increasingly open to automation, while some of the administrative sludge surrounding creative work, from project documentation to internal knowledge management, is becoming easier to shift away from expensive human attention. Most of this is undeniably useful. There’s very little romance in forcing a talented designer to spend half a day resizing near-identical assets just because somebody thinks repetitive labour demonstrates creative integrity.
We’ve already explored why sensible adoption depends on understanding what AI should be allowed to do and what should remain firmly human, and the same distinction applies at agency level. If AI can remove repetitive production friction, compress tedious research tasks and help people arrive at a workable starting point faster, it should. The more difficult question is what happens to the time, attention and money that have been saved.
There are two possible versions of that future. In the optimistic version, AI removes repetitive work and creates more room for research, experimentation, craft, cultural exploration and properly ambitious thinking. The agency becomes faster at the boring things so it can spend more time being interesting. In the less inspiring version, every efficiency gain is immediately converted into greater output. The team that once produced 50 assets is now expected to produce 300. The strategist who gets a day of research back is handed twice as many briefs, while the creative team that can visualise 20 territories by lunchtime is congratulated and asked whether it could make that 40 by Friday.
Both models are more productive. Only one is obviously more creative.
And that distinction matters because a faster conveyor belt is still a conveyor belt.
When Everything Gets Easier to Make, What Becomes Harder to Value?
Pernod Ricard
There’s no point pretending the productivity benefits aren’t significant. Creative businesses have always contained an enormous amount of labour that sits somewhere between craft and logistics. People have spent decades manually performing tasks not because those tasks required rare human imagination, but because no sufficiently capable machine existed to perform them. Now one increasingly does, and that is fundamentally changing the economic value of execution.
The ability to produce a competent first draft, visual concept, social asset, product image or piece of campaign copy is no longer particularly scarce. As we’ve explored previously, AI is commoditising average creative output at extraordinary speed. Work that once required specialist skills, expensive software or significant production resources can increasingly be achieved by a relatively small team using readily available tools. That is liberating for smaller agencies because it allows them to punch far above their historical production weight, but it is also commercially uncomfortable because much of what agencies once sold as specialist capability is gradually becoming accessible to everybody.
This is where scarcity begins to move. If making something competent becomes dramatically easier, then competence itself becomes less valuable as a differentiator. A beautifully rendered key visual can now appear in seconds. A convincing campaign manifesto can be produced before lunch. A deck can be filled with credible-looking ideas at a rate that would once have required days of studio time. None of those outputs are necessarily worthless, but their existence tells us less and less about the quality of the thinking behind them.
That means value starts accumulating elsewhere. When making becomes abundant, selection becomes scarce. When options become practically infinite, taste becomes useful. When 100 plausible campaign routes can be generated before the first proper review, the person adding the most value isn’t necessarily the person who can create route 101. It might be the person who understands why route 37 contains something the others don’t, why route 12 is visually seductive but strategically meaningless, and why the least immediately impressive option might actually be the one capable of becoming distinctive.
This is why curation is becoming as important as creation across the creative industries. In a world where agencies can generate huge volumes of plausible creative work, the number of options presented to a client is no longer compelling evidence of value. If anything, the agency that can intelligently reduce those options may become more useful than the one proudly producing hundreds of them.
Agencies have traditionally been very good at demonstrating how much they’ve done. AI makes those quantities almost meaningless. Perhaps the more interesting future skill is being able to demonstrate why so much work was deliberately left undone.
When Everyone Has the Same AI Advantage, What Makes an Agency Different?
PepsiCo Design & Innovation
Here’s the slightly awkward part. Your AI capability probably isn’t as unique as you think it is.
That doesn’t mean your workflows aren’t clever. Some agencies are undoubtedly building genuinely sophisticated internal systems around proprietary information, custom models, specialised agents and more elegant connections between strategy, production and delivery. Those systems can create real operational advantages, particularly while clients and competitors are still working out what good AI implementation looks like.
But operational advantages have an annoying habit of spreading. Software gets copied. Features become standard. Staff move between businesses. Vendors package clever internal workflows into products. Clients gain access to the same platforms.
What looked like genuine sorcery in January can become a reasonably detailed YouTube tutorial by October.
If an agency’s differentiation depends entirely on having discovered a smarter way to use broadly available technology, that differentiation may come with an expiry date. That is why a genuinely future-ready agency needs something deeper than technical fluency. It needs to know what it believes about creativity, brands, business and culture. It needs distinctive people with particular obsessions, recognisable taste and enough confidence to approach the same problem differently from everybody else with access to precisely the same machines.
That kind of agency personality used to develop almost accidentally. Some agencies became famously strategic, some chaotic but culturally electric, some beautifully crafted, some blunt, weird, funny, provocative or gloriously commercial. You could occasionally look at a piece of work and have a fairly decent chance of guessing where it came from because the agency itself seemed to have a character.
There were obvious problems with that mythology. “Culture” has occasionally been used in agency land as an excuse for everything from ludicrous working practices to truly catastrophic furniture. But the underlying point still matters. Agencies with strong identities tended to make choices. They had preferences. They admired certain kinds of work, hired particular kinds of people and viewed clients’ problems through a recognisable lens.
AI doesn’t remove any of that automatically. The concern is that agencies may voluntarily sand away those edges while chasing optimisation.
Could AI Be Making Creative Work More Predictable and Less Original?
AKQA
The uncomfortable answer is that it can.
One of the more interesting findings emerging from research into generative AI and creativity is that the technology can improve an individual’s creative performance while simultaneously making the wider pool of ideas more homogeneous. In experimental creative tasks, people using AI have produced ideas judged stronger on average while the diversity of ideas across the group has narrowed. That should concern agencies just enough to be useful because advertising doesn’t necessarily need everybody in the industry becoming slightly better at producing the same answer. It still needs somebody to find the answer nobody else saw.
AI is extraordinarily good at plausibility. It understands the patterns that make something look like an advert, sound like a strategy, read like a manifesto or resemble a premium brand film. That capability is incredibly useful when you need speed, volume or a starting point. But plausibility has a gravitational pull. If you’re not careful, everything begins drifting towards the centre, where work is beautifully coherent, professionally executed and almost impossible to remember ten minutes later.
We’ve already explored how creative work can acquire a recognisable AI aesthetic, and the problem extends well beyond glossy skin, dramatic backlighting and suspiciously atmospheric fog. The deeper issue is conceptual. If thousands of creative teams are drawing from overlapping systems trained on vast amounts of existing culture and asking them similarly phrased questions, it isn’t difficult to imagine those teams beginning to orbit similar creative territory.
Human creativity has always had its trends, of course. Agencies didn’t need machine learning to make every start-up look identical in 2017. But generative AI can compress the cycle. An aesthetic, phrasing style or creative convention can spread quickly, be generated at enormous scale and harden into default before anybody has had time to become properly irritated by it.
That doesn’t mean AI inevitably destroys originality. It means originality may require more active resistance than it used to. Before generative AI, an agency could become generic through laziness. Now it can become generic through efficiency.
The Efficiency Trap
Mantra
Efficiency is one of those words nobody likes to argue against. Imagine walking into the boardroom and confidently announcing that your strategic objective for the year is to make the agency slower and more expensive. You probably won’t be given the laser pointer again.
But creative businesses should maintain at least a little suspicion around efficiency because not every inefficiency is waste. Research into agency AI adoption has repeatedly shown that increased productivity and lower costs are among the most obvious reasons businesses are embracing the technology, while the more difficult questions centre on whether that same focus can diminish differentiation and creative value.
The danger isn’t efficiency itself. The danger is assuming that everything capable of being optimised should be.
Creative development has always contained moments that look terribly inefficient from the outside. Somebody goes down the wrong research rabbit hole. A designer spends half an afternoon exploring a route that ultimately dies. A strategist introduces an apparently irrelevant cultural reference that suddenly unlocks the brief. A junior creative asks an awkward question that everybody more experienced stopped noticing years ago. None of these moments can be guaranteed to generate value, but neither can they be comfortably replaced by a process designed exclusively to reach an acceptable answer faster.
Even mistakes have creative value. There’s something revealing about creatives deliberately putting mess and imperfection back into work that has begun to feel too smooth. That isn’t nostalgia for poor craftsmanship. It’s a response to a world in which flawless technical polish is becoming so readily available that irregularity, idiosyncrasy and evidence of an actual decision can feel unexpectedly alive.
The same principle should apply to agency process. Remove production friction wherever that friction contributes nothing. Automate the tedious steps. Compress the repetitive work. Make administrative systems faster and more sensible. But be much more cautious about removing creative friction, because creative friction is where disagreement, surprise and unexpected connections often happen.
Those two kinds of inefficiency might look similar on a spreadsheet. They are not the same thing.
Don’t Optimise the Weird People Out of the Building
Juice
There’s also a human version of this problem.
Agencies have always depended on people whose value is frustratingly difficult to represent in a neat efficiency model. There’s the strategist who remembers a completely irrelevant article from seven years ago and somehow connects it perfectly to today’s brief, the designer with an unhealthy obsession with 1970s Japanese packaging, or the creative who spends far too much time in obscure corners of the internet but seems to know six months before everyone else that a particular cultural shift is gathering momentum.
These people aren’t necessarily efficient in the conventional sense. They can still be immensely valuable.
AI is already changing workforce economics in ways agencies will understandably find difficult to resist. If routine production, research and administrative tasks can be automated, fewer people may be required to perform certain categories of work. That can make agencies leaner and commercially healthier, but it also raises a less glamorous question about how creative talent is developed in the first place.
Junior agency roles have never consisted exclusively of glamorous creative revelation. Young creatives learn partly through doing work that senior people no longer want to do. They sit in reviews. They produce the variants. They watch experienced people dismantle their first ideas and discover, occasionally painfully, the difference between something that merely looks finished and something that actually works. They absorb the culture of creative judgement through proximity as much as through formal training.
If AI strips away much of that entry-level work, agencies need to replace the learning rather than simply celebrate the saving. Otherwise they may find themselves in five or ten years with highly efficient senior teams and a worrying shortage of people who ever got the chance to become senior.
That is why the new skills agencies need to build cannot be reduced to prompt engineering and workflow automation. Judgement, commercial curiosity, cultural literacy, critical thinking and the ability to recognise non-obvious solutions become more important when obvious ones are available instantly. If agencies optimise every awkward, inefficient, exploratory part of junior development away, they might become leaner in the short term while quietly eroding the very talent pipeline that makes them interesting over the longer term.
Your Clients Probably Don’t Want to Buy Your AI Stack Forever
Yarza Twins
There’s a fairly predictable cycle with new agency technology. First, it’s mysterious. Then it’s impressive. Then every agency buys it. Eventually, clients start wondering why they’re paying extra for something they also have.
AI is racing through those stages.
Recent advertiser research suggests generative AI adoption is already close to universal in many marketing organisations, with 99% of surveyed advertisers engaging with it in some form. That changes the agency proposition because the client is no longer sitting on the opposite side of the table wondering what this strange new technology can do. Increasingly, they’re using the same underlying tools, hiring their own AI specialists and developing internal systems of their own.
That doesn’t mean they no longer need agencies. It means “we use AI” becomes a progressively less interesting answer to the question of why they should use you.
We’ve already asked whether brands will still need agencies when AI can make the work, and the answer depends heavily on what agencies believe they’re selling. If the agency is essentially an external production department with some expensive strategic language wrapped around it, growing client capability is a clear threat. If the agency provides external perspective, cultural intelligence, challenge, specialist judgement and the ability to solve problems the client struggles to see clearly from inside the organisation, the proposition remains far healthier.
Clients are unlikely to remain permanently impressed by the existence of an AI stack. They might remain impressed by what your people see that other people miss.
That distinction is going to become increasingly important as technology becomes less visible. Nobody currently chooses an agency because it has access to Adobe Creative Cloud. Eventually, access to powerful generative systems will feel just as ordinary. The competitive advantage won’t come from admitting AI exists. It will come from using it in ways that sharpen rather than flatten the agency’s point of view.
So What Is an Agency Actually Valuable For Now?
Name and Name
This question keeps returning because AI has exposed something agencies have occasionally managed to avoid discussing for years: what exactly is the client paying for?
Historically, the answer bundled together an enormous number of things. Clients bought access to talent, ideas, production infrastructure, specialist software, strategy, project management, relationships, scale, craft, experience and whatever collection of intangible qualities made one agency feel more capable than another. AI is beginning to pull some of those strands apart.
Certain production capabilities are becoming cheaper and easier to reproduce. Routine execution can happen internally. Specialist knowledge is easier to access. That is one reason the traditional agency business model is under increasing pressure, because clients can separate tasks that genuinely require outside expertise from those that historically required an agency simply because the client lacked the infrastructure to do them.
Paradoxically, that may make the genuinely valuable parts of an agency easier to see. The agency can stand outside the organisation and recognise assumptions that have hardened internally into facts. It can connect one client’s problem with lessons from entirely different categories. It can create a space in which ideas survive long enough to become bolder than internal politics would normally allow. It can bring together strategists, designers, writers, technologists and makers with radically different ways of seeing the same problem.
Most importantly, it can have a point of view.
Yes, AI can generate a point of view in seconds. Ask it to behave like a provocative challenger brand strategist, a sceptical creative director or a culturally savvy Gen Z consultant and it will happily oblige. But a genuine agency point of view has consequences. It influences the people you hire, the work you admire, the clients you attract, the briefs you reject and the ideas you’re prepared to defend when a meeting becomes uncomfortable. It shapes what comes out of the business even when nobody explicitly asks for it.
It might even cost you business occasionally.
That’s often how you know it’s real.
The Danger of Letting the Tool Become the Culture
Yolo productions
There’s something slightly ironic about agencies trying to differentiate themselves through a technology specifically designed to make advanced capabilities more accessible.
Every agency should understand AI. Every agency should be exploring it seriously, building sensible governance around it and training people to understand both its possibilities and limitations. Refusing to engage with generative technology in 2026 isn’t some romantic defence of creative purity. It’s rather like refusing to use email because handwritten correspondence has more soul.
But AI literacy and AI identity are very different things.
The danger begins when the tool stops supporting an agency’s culture and begins replacing it. You can already hear some of this happening in the language. Everybody is “AI-native”. Everybody is combining “human creativity and machine intelligence”. Everybody has discovered “augmented ideation”. Everybody promises the apparently revolutionary ability to produce more content faster.
Fine. But once every agency can say broadly the same thing, those claims stop telling clients much about you.
The more interesting questions are harder to automate. What do you care about? What kinds of clients bring out your best work? What makes your creative directors argue? Which part of culture do you understand better than the agency sitting three streets away? Which designers, directors, writers and strategists do you find exciting? What would you make if nobody had searched for it yet?
Those questions sound softer than workflow efficiency, but they’re also far more difficult for competitors to copy.
That is why the argument around human creativity becoming more valuable in the age of AI was never really about proving that humans can outperform machines at every creative task. Increasingly, they won’t, and they don’t need to. The interesting question is which human capabilities become more valuable precisely because mechanical competence is becoming abundant.
Taste is one. Conviction is another. Curiosity matters. So does the willingness to disagree.
And being interesting definitely helps.
How Can Agencies Use AI Without Losing What Makes Them Interesting?
Groove Jones
The answer isn’t to use less AI simply to demonstrate creative virtue. That would be wonderfully tidy and almost completely pointless.
If anything, AI should gradually become less conspicuous inside good agencies. It should sit within creative workflows in much the same way Photoshop, cloud software or search engines do now, helping wherever it makes sense without requiring everyone to pause and admire the fact that it’s there.
Use it for compression, prototyping, variation, repetitive production and the hundred tiny pieces of friction that prevent talented people spending time on higher-value work. Let it interrogate research, challenge assumptions, visualise speculative ideas and help teams test possibilities before significant money is committed.
But don’t allow it to become the default source of every first thought.
If every brainstorm begins with somebody asking the same model for 20 ideas, the machine has already established the territory before the humans have had a chance to explore it themselves. Instead, agencies should be creating more space for references from outside the obvious data stream, for research with actual people, for strange cultural interests, archives, travel, conversations and personal obsessions. In other words, all the messy human inputs that make one group of creatives see something another group might completely miss.
As we’ve explored in looking at the skills that make creative work distinctive in an AI-saturated market, speed is unlikely to remain much of a creative advantage once everybody has it. Curiosity, divergent thinking and the ability to avoid the obvious answer become considerably more important.
That should change how agencies judge AI-assisted work too. Asking whether something is “good enough” is an increasingly weak standard when the technology itself is becoming remarkably good at achieving exactly that. Instead, ask whether the idea could have come from anybody else. Does it reveal something particular about the brand? Is there an actual observation underneath it? Has a creative person made a choice the model wouldn’t naturally have made? Is there cultural specificity, tension, humour, craft or surprise?
The growing interest in “Made by Humans” as a potential marketing signal comes partly from this same anxiety. It isn’t necessarily an anti-AI movement. It reflects a growing desire to know whether somebody actually meant the thing we’re looking at, whether there was intent behind it and whether a human being is prepared to stand behind the decisions that produced it.
The clever agency won’t respond by banning AI or fetishising analogue imperfection.
It will simply make sure its people are still adding something worth crediting.
Maybe AI Hasn’t Made Your Agency Less Interesting
VILKA Agency
Perhaps that’s the more optimistic way to see all this.
AI hasn’t necessarily made agencies less interesting. It has simply made being uninteresting harder to hide.
For years, an agency could compensate for a relatively generic point of view with scale, production infrastructure, access to specialist capabilities and enough bodies to service enormous volumes of work. Those things still matter, but technology is distributing some of those advantages far more widely. Smaller teams can now achieve what once required entire departments, while clients themselves can handle tasks that previously had to leave the building.
That means the parts of an agency that can’t simply be downloaded, licensed or recreated from a tutorial become more visible. The people, culture, judgement, relationships, weird references and accumulated experience of getting things spectacularly right and occasionally horribly wrong all start carrying more weight.
So does the willingness to tell a client that the brief isn’t actually the problem. So does the ability to recognise when something technically immaculate has no pulse whatsoever. So does the confidence to kill the efficient answer because the difficult one has more potential.
AI should make all of that more powerful. It should give a brilliant strategist more territory to explore, a brilliant creative more ways to test an idea and a brilliant designer more freedom to prototype something impossible before anyone spends a penny making it real. It should clear away enough of the mechanical work that interesting people have more time to do interesting things.
But that only happens if agencies remember which part of the equation they’re trying to accelerate.
The machine shouldn’t supply the personality. It should create more room for the personality to show up.
Because when every agency can generate, automate, iterate, localise and optimise, efficiency stops being particularly interesting. The tools become invisible and the capability becomes expected. At that point we end up roughly where the creative industries have always been.
Clients still need somebody to notice what other people didn’t.
They still need somebody capable of challenging the obvious answer, understanding why a technically perfect idea feels completely lifeless and recognising the strange little insight that might turn into something nobody else could have made.
So if AI has made your agency less interesting, the problem probably isn’t that you adopted AI.
It’s that somewhere in the rush to become faster, leaner and more efficient, you started optimising the very things that made you worth hiring in the first place.
And that might be the greatest irony of the whole AI revolution. The technology was supposed to remove all the boring work so humans could spend more time being creative.