As someone who works in or around the creative industries, you probably nod along when you hear that AI is “transforming the space.” Everyone talks about disruption, new workflows, the death of certain roles. Everyone is on the same page.
But last week I was at a closed-door meeting — platform owners, startup founders, production studio heads. The kind of people who talk about transformation at conferences and talk about margins among themselves. And the conversation was different.
==It wasn’t about the future of creativity. It was about where you can stop spending money.==
In this article, I’ll explain why I think the real driver of AI adoption in creative industries isn’t automation — it’s the elimination of transaction costs. I’ll show why that’s rational, why it changes the product, and why the people doing the optimising may not fully understand what they’re giving up.
The Real Reason AI Gets Adopted
Most public conversations about AI in creative industries orbit the same topics: copyright, aesthetics, whether AI will replace jobs. These are real questions. But if you look at what’s actually driving adoption at the business level, the answer is less philosophical.
The key driver is transaction costs — everything a business spends not on the work itself, but on organising that work with people. Finding someone, negotiating terms, explaining the brief, reviewing the output, managing the conflict, replacing the person who just quit. In manufacturing, you can standardise most of this. In creative work, you can’t, because almost everyone involved brings subjectivity, unpredictability, and their own negotiating position to the table.
In that back-room conversation, this was put plainly: human specialists are a continuous source of operational uncertainty. They get sick, burn out, push back on briefs, quit at the worst possible moment. The infrastructure around them keeps getting more expensive — offices, benefits, corporate therapists. From the perspective of a financial model, none of this is easy to plan for.
I don’t think this logic is wrong. It’s rational. But it leads somewhere worth paying attention to.
AI Doesn’t Reduce Transaction Costs — It Eliminates Them
This is the part that makes AI genuinely different from previous waves of automation.
You don’t hire a model, onboard it, motivate it, or retain it. It has no negotiating position. The cost of coordination drops to essentially zero. From a theory-of-the-firm standpoint, an entire class of costs that was previously considered unavoidable in creative production simply disappears.
Which is why adoption isn’t happening through strategic announcements. It’s happening quietly, through budgets. First the freelance pool gets cut. Then contracts don’t get renewed. Then positions that used to open automatically just don’t get posted. Nobody declares they’re replacing people — hiring just becomes economically unjustifiable.
Classic substitution effect. Nothing surprising here, if you’re watching the numbers.
What Happens to the Product
Here’s what barely came up at that meeting, and what I keep thinking about.
Reducing the cost of producers changes the product. When the main optimisation criterion is cost-per-content-unit, quality becomes the first variable to sacrifice. Not intentionally — just by the logic of the system.
AI is good at solving the problem it’s given. The issue is that it’s being given the wrong problem. Clickthrough, retention, completion rates are measurable, and the model optimises them well. To do that, it leans on the most reliable behavioural levers: anxiety, FOMO, fast dopamine hits. Not depth — button-pressing.
The result is already visible: structural homogeneity. Identical grabby headlines, cloned formats, copy-paste content. Each piece technically does its job — captured attention for a few seconds. In a spreadsheet, that looks like a win. At the market level, it’s the slow devaluation of the product itself. When everyone produces the same thing, differentiation disappears — and so does the reason an audience chose a particular brand or publication in the first place.
It’s a classic race-to-the-bottom trap: every player behaves rationally, the aggregate result destroys the market for everyone.
The Server Rack Where the Office Used to Be
The logic I heard at that meeting, taken to its conclusion, draws a simple picture: not a team in an office — a server rack. Not an editorial room — a generation pipeline. Some companies are already moving there. Quietly.
Short-term, it works. Long-term, it’s less obvious. Business history has plenty of examples of aggressive cost optimisation producing things that are technically flawless but stopped being needed. Companies that removed humans from customer touchpoints and discovered that loyalty left with the headcount.
Human “inefficiency” in creative work — disagreement, subjectivity, reframing the brief — isn’t a bug. It’s the mechanism that produces differentiation. It’s what makes work feel like it came from somewhere. AI removes that mechanism along with the costs.
What’s left on the other end of this optimisation, I’m genuinely not sure. And I suspect the people doing the optimising aren’t entirely sure either.
