AI | Skynet never came. Something quieter did.

Joshua Barnes, at Connected Media Group, on why the real AI story in marketing isn't a takeover - it's a slow handover.

I grew up with big Arnie and Sly Stallone the action heroes of a generation. A time when Skynet was sending machines back through time to wipe us out and Sly’s character John Spartan thawed out of cryo-freeze to fight Wesley Snipes in a future so sanitised that swearing gets you fined. That was our idea of the AI future: dramatic, explosive, unmissable. One day humans are in charge, the next day we're not.

Twenty-odd years later, I'm sitting in marketing meetings watching the actual thing unfold, and it looks nothing like the movies. There's no single moment where AI "takes over." No war-room countdown. Just a thousand small decisions, each one sensible on its own, each one a little more efficient and automated, handing more of the steering wheel to a system rather than a person.

I'm a genuine supporter of AI and what it offers this industry, enough that we've made a serious investment of time and resources to ensure we're solidly on the ai train, not watching it leave the station.

But we need to be careful about what we're actually racing toward. The sci-fi version of this story is a takeover, one dramatic moment where AI seizes control. The real version is quieter, and in some ways harder to spot: a race to the middle. When every business has access to the same tools, trained on much the same data, producing much the same "good enough" output, the risk isn't that AI takes control of anything. It's that everyone starts to look the same. Acceptable becomes the default setting, and acceptable, dressed up as efficient, is genuinely hard to tell apart from a competitor's acceptable. That's what disempowerment actually looks like here: not one moment where authority changes hands, but a slow erosion of the thing that made you distinct in the first place, because sameness is what the fastest, most efficient path produces when nobody's pushing back on it.

We keep the job titles, the sign-off meetings, the org charts. But more and more of the actual decisions, what gets bought, what gets built, what gets said to a customer gets made a layer or two below where a human is actually looking. That becomes more likely, not less, if we stop bringing in new, ambitious, thoroughly trained marketers into this industry, the people who'd otherwise be the ones pushing back. And that's getting harder to do, because the entry-level processing jobs that used to train them up are now done faster and cheaper by our best mates, Claude and ChatGPT.

None of this is intentional. But in an industry full of constant time pressures, deadlines and cost structures tied back to pitch processes and procurement-driven "value," it doesn't need to be intentional to happen. The version of the business that removes the human bottleneck moves faster than the version that doesn't, and slower simply loses. No dramatic moment required. Just competitive pressure, one delegation at a time.

That's the sci-fi we should actually be paying attention to. Not Skynet. The version where nobody loses control on purpose, it just quietly stops being anyone's job to hold.

What we’re already getting right with AI

Let's not pretend it's all downside, because it isn't. Right now, in this industry, AI is genuinely buying back time:

• Less time drafting and re-drafting thought pieces, decks and first cuts of copy, more time actually thinking about them.

• Bulk data sorted, cleaned and structured in minutes, not days, freeing up hours for judgement instead of admin.

• Faster first passes on competitive insights, research and reporting so strategists spend more time on "so what" and less on "gathering."

That's real. Anyone telling you AI hasn't made parts of this job easier this year hasn't been paying attention.

It's also why, even as we've found genuine efficiencies through AI, we've made a point of keeping our smart, senior core team fully intact. More automation doesn't mean less need for sharp people, if anything, it means more. Getting the best out of AI still depends on someone with the judgement to challenge what it produces, spot when it's average, and know what better actually looks like.

We're already seeing exactly this shift play out in something as tangible as campaign bidding. Automated bid management now handles most of the moment-to-moment optimisation that used to fill a media team's day, and it does it better than we ever could manually. But that just moves the human value upstream. It has to sit in the inputs we feed the system, the creative testing frameworks that tell us what's actually resonating, and the brand and positioning strategy that gives a campaign something distinct to say in the first place. Automate the bidding, not the thinking.

The bill that comes with it

But there's a cost sitting alongside the benefit, and it's not the one people talk about most.

The obvious one is investment - the money, time and internal politics of building the infrastructure, the workflows, the governance to actually use this stuff properly. Most businesses are still building and refining this, which means living with the awkward bit: overpaying for capability you haven't fully utilised yet.

The quieter one worries me more. A generation of grads and new starters are coming into this industry learning to brief AI to do the job, without ever having done the job the slow way first. There's a difference between knowing the value of a good brief and having actually written forty bad ones before you understood why the good one worked.

It's a bit like being taught to change a tyre by watching someone else do it on video. You get the theory. What you don't get is the stubborn lock-nut that won't move without something to leverage it, or the ten minutes afterwards getting the grease off your hands before it ends up on the upholstery of a car you actually care about. Nobody teaches that part. You only learn it by kneeling on the bitumen and doing it wrong once.

Strategy is the same. The slow, arduous version, digging through the data yourself, writing the bad brief, watching it fail, is where the instinct comes from. Skip straight to "AI, do this" and you get someone who can operate the tool, but has never felt why the tool needed operating that way in the first place.

Where we stand on AI in Marketing

A few things I think are worth holding onto, whatever else changes:

• Keep questioning and challenging what AI-driven work actually produces - don't wave it through because it looks efficient.

• Before anything goes into a prompt, know what "good" looks like coming out. If you can't explain the result, you haven't briefed it properly, you've just automated the guessing.

• A bad brief in, a bland strategy out. That's not a new problem. AI just makes it faster to produce bland at scale.

That last bullet is the race to the middle I mentioned earlier, playing out at brand level. We've watched independent agencies rise over the past decade precisely because they thought differently, took bigger risks and approached challenges differently. AI doesn't threaten that by producing bad work. It threatens it by producing acceptable work, cheaply, at volume, that looks a lot like everyone else's acceptable work. You've seen the ad on social. The one that says "yes, we're expensive, but". If everyone can generate a passable version of that ad in ten minutes, the "but" stops meaning anything.

In an industry that supports clients who need to differentiate and cut through to succeed in an increasingly competitive environment, efficiency is essential, but we can’t forget an agency's value comes from its ability to challenge, think differently, and create the difference clients themselves desperately need in this AI-focused economy.

The actual question

So, here's the one I keep coming back to: how do we keep training and rewarding independent thought, investing time in graduate and executive training, celebrating the strategists and thinkers who differentiate brands, in an industry that has always, if we're honest, wanting independent thought, but making decisions which creep towards a cheaper version of "acceptable"? I hope we hold ourselves to a higher standard.

We already know the brands that break that pattern. The ones who commit to something genuinely different in market, not just louder, but different. These are the ones who hold their prices and grow while the price-competitive middle grinds itself down.

The question isn't whether AI is going to change this industry. It already has. The question is whether we let it quietly average us all towards the same bland middle, one efficient, defensible, "good enough" decision at a time, or whether we use the time it's buying us to actually get better at the one thing it can't do for us: thinking differently, continually challenging and learning, on purpose.

No Skynet required. Just a choice, made slowly, one brief at a time.