There is a conversation happening across every marketing team and every agency in the industry. It usually goes something like this: AI will automate the repetitive work, free up time for creativity, and make everyone more productive. Human judgment stays at the top. Execution moves to the machines.
It's a comfortable story. It's also incomplete.
The honest version is harder to say out loud.
AI is not just automating execution. It is getting better, faster than anyone predicted, at language. At text. At code. At reasoning. At the visual languages that designers have spent careers mastering. The roles that required a human because they required fluency in a tool — Photoshop, Illustrator, a content management system, a web development environment — are being hollowed out. Not eventually. Now.
This is not a crisis for the industry. It is a reckoning.
Once the machines handle everything that can be expressed in language, what is left that is worth a human being in the room?
The interface layer is collapsing
One thing does seem clear, even now: the roles most at risk are not only the obviously automated ones. Any work that is essentially about being the human interface to a tool — translating a brief into a platform, moving assets between systems, knowing how software works so others don't have to — is exposed.
If your value is that you know how to use something, and that something is becoming usable by everyone, the value disappears. The interface layer is collapsing.
We don't know what fully replaces it. Nobody does. But we have some intuitions, and we think they're worth saying out loud, even if we hold them loosely.
Five things that still feel human
1. Cultural sensitivity
There is something in the act of reading a moment — understanding what an audience is ready to hear, what tone will land, what would feel wrong even if it's technically correct — that seems to require having actually lived in the world. Not processed descriptions of it. Lived it. Whether that's something AI will eventually replicate, we genuinely don't know. For now, it still feels distinctly human.
2. System thinking in complex contexts
Real marketing problems are rarely clean. They come loaded with factors that were never written down anywhere: a CFO who distrusts the brand team, a market that's shifting for reasons that aren't in any brief, a product with a complicated history that shapes how everything lands. AI can only work with what it's been told. Human judgment navigates what it hasn't been told — and often doesn't even know it's doing it.
3. Ethical and strategic judgment
Knowing when to stop, what not to say, when the brief itself is the problem. AI optimises toward a goal. It doesn't question whether the goal is right. That capacity for refusal, for redirection, for recognising the thing that shouldn't be done — we think that still belongs to humans, at least for now.
4. Relationship and trust
The work that matters tends to emerge from something that was built over time between people. The ability to walk into a room, read the dynamics, earn confidence, manage conflict, hold a client through uncertainty — none of that is expressible in a prompt. Whether it ever will be feels like the wrong question. It's human in a way that goes beyond capability.
5. Synthesis under pressure and ambiguity
Not the optimal answer with complete information, but the best available answer with incomplete data, limited time, and people in the room who disagree. That capacity for judgment in the fog, for committing to a direction when no direction is obviously right, seems irreducibly human — at least in the contexts we work in.
We offer these not as certainties, but as working hypotheses.
The honest position is that we don't know where this is going. Nobody does. The tools are moving faster than our ability to understand what they're changing, and any confident prediction about where the human ends and the machine begins is probably wrong already.
But if you work in this industry, or want to, these might be worth investing in. Not because they guarantee anything, but because they seem to be the least replicable things we have. The places where, for now, a human still makes a difference that a machine doesn't.
What seems certain is that there will be far less work in these industries than there was before. Not eventually. The contraction is already underway. The roles being lost are not coming back, and the ones that remain will be fewer, more demanding, and harder to define.
The industry that is being dismantled was, in many ways, a noise machine. Thirty years of internet-fuelled content production, built on the idea that more was always better: more formats, more channels, more output, more reach. It produced an enormous amount of work. It produced far less meaning.
If AI forces a contraction, that might not be a loss.
It might be a correction.