If the demand is for business use cases... its more so that yes the demand is there, or at least was, but the returns are not.

Every industry projection of productivity gains from implementing LLMs into their businesses was wildly, wildly overestimsating how much it would actually help any given business.

So its basically all a massive leveraged bet on the idea that LLMs will be able to usefully automate tasks that used to be done by people... but for the most part, it can't.

And that's partially because the capabilities of LLMs are absurdly overhyped, and partially because most businesses internal software set up is a clusterfuck nightmare of smashed together contractor modules and services that barely works, because management tells developers to just keep putting bandaids on things snd writing spaghetti code, never letting them do a proper refactor/restructure.