I think it's simpler than that. I think the AI snake oil sales people realized that corporate executives want to give themselves bonuses. And if you propose this magical technology, that's going to change everything, they can pretend for several years but it's actually going to change everything, and pay themselves a lot of money in the meantime.

It's greedy people relying on other greedy people so that some of them can get richer.

No, making the customer think they've got the magic bullet doing all they need is a valid one too. The only problem is that it's repeating a job already done in areas where you need to do jobs not previously done, ideally. It's very useful for all wrong applications of tech where it was honestly not needed (things easily repeatable done by people who are honestly not that much smarter than haircutters). It's useless for the skeleton applications.

It's a very smart Ponzi scheme, one where much of the funds collected gets spent on maintaining real usefulness, at least visible one. Unlike most Ponzi schemes, it's able to achieve unimaginable scale due to that.

Like, say, plague has low survivability rates, and doesn't spread too easily, meaning that it's not very dangerous in the modern world. While a disease with very high survivability rates and easy spread has far bigger reach. Or a very slow incubation period with carriers being infectious long before detection.

OK, I suppose everyone likes Plague Inc.

IRL it's no good as something everyone uses, because current token costs are subsidized and because what's solvable with it for most people is what's been done many times before.

When things balance out, I'd expect token costs to be like 4 times what they are now. And people using these things to be mostly engineers.