If you want a snapshot of the fundamental move of the intersection between the "AI economy" and the "creator economy" — air quotes very much intended — this recent BBC article is as good as anything else.

• Company uses a team of writers (sorry, "content creators") to write articles ("generate content") to post.
• Company hooks up a generative AI API.
• Company fires everybody except a single person to polish the generated texts into something somewhat usable, paying them less because it's not writing, it's editing.
• Profit

None

From the point of view of the writers this straddles the line between business history and horror story; a couple of years ago I wrote a short story, The Writer's Room outlining this precise scenario, with a side order of neurotech. And I do have concerns about the aesthetic and intellectual implications of a shift to this model, for example in the context of journalism as an activity with irreplaceable positive externalities.

But let's put that aside — let's pretend we can — and focus exclusively on the business case. What companies gain in the short term is obvious: hugely increased margins and practically unbounded scalability.

The cost is the very possibility of quality. Not on an absolute scale as generative models get better almost month by month. But whatever model a company is using their competitors can too. Whatever prompt or trick they figured out, somebody else will reverse-engineer in an afternoon or two. Once you've outsourced creation to somebody who's happy to sell to anybody you can no longer compete on the quality of what's created.

By itself this isn't necessarily bad. Companies selling low-cost commodity products at scale are seldom famous but can be very profitable. They don't have quality advantages but what they do require large capital investments, complex and competitive distribution chains, or access to specific locations that can be made exclusionary by state support; this is true mostly of things like soy or copper, but applies also to some manufactured products.

Content generation is a very different industry. It has low entry costs, but also, and critically, distribution, surfacing, monetization strategies, customer relationships, and market data are all under the tight control of a handful of oligopolistic platforms. You can't build customer loyalty because what most people see most of the time is an algorithmically "curated" feed: when people subscribe to your channel or follow you they are really completing an invisible survey telling the platform what else they can show them. Algorithms use this almost instantaneously — making your content harder to stand out in the feed — and shortly afterwards, as this information percolates down the content generation chain, it nudges more content towards whatever you did that seemed to work.

Let's pull back slightly away from the specifics of algorithms and platforms. Standard economic question: there's a market with absolutely dominant oligopoly intermediaries almost impossibly expensive to compete with and a supply side with low barriers of entry (posting is easy, posting is cheap). Add to this a technological shift that lowers marginal costs to nearly zero for all suppliers. Who captures the new value?

The platforms, of course. The more content there is the less they have to even gesture towards getting people paid for it. Even if they aren't directly in the monetization path, they do control how much attention each content creator gets. The more content there is, the less negotiating power has any creator, and the more the platforms themselves can get from them.

The consequences of the shift in the supply cost function are as simple as they are concerning to the supply side: costs drop but there's no barrier of entry or capex minimum, so margins crash as well and there's no way to lock in demand. You can't compete on quality (everybody uses the same tools), uniqueness (you fired any human that might have some knowledge inside their heads that's not yet baked in the data the models were trained on), or relationships (everybody is on an algorithmic feed).

It doesn't mean no creator or content generation company will make money. Some will, by a combination of grit, skill, and luck, but grit and skill won't be enough — they seldom are — and the luck you'll need in the exponentially more competitive and slippery market will also have to be exponentially better than before. Some people and companies love this kind of challenge: some will take a stab at it, a very few will get rich, and we won't really hear a lot from the vast majority that didn't.

It's usually like that.

There's no such thing as the AI economy. Nvidia makes so much money because competing with them requires levels of investment even the Chinese government has to consider with care. Calling a model API is at best an industry-wide reduction in cost with no impacts on margin and at worst a suicidal sacrifice of hard-won expertise advantages.

The moral, anyway — or rather the business strategy takeaway, as we put aside that other sort of concern near the beginning of the article — is twofold:

• If you thought being a content organization in online platforms was already hard and chancy, and you were right of course, it's going to get much worse.
• Capturing the benefits of generative AI requires some degree of control over your relationship with your audience of the sort that simply doesn't happen in the large platforms.

In fairness, it might be useful even there to generate at lower cost the checkbox content that for traditional reasons has to be written but neither harms nor benefits a company by it being read and therefore isn't handicapped by the lack of any competitive advantage; the Kafkian logic of this use of new technology would be hauntingly unsettling if it weren't already so familiar.