I don't know exactly how to describe this, we've all heard the bonkers stories about AI leaderboard gaming that happened. It's similar. There are a lot of orgs that have been blasting a lot of data at LLM, both via API and to racks of local GPUs, and leadership is happy to just "have a lot of AI happening" without measuring results. The benchmark so far has been "are you using AI" not "what is the [cost | value | benefit]" or even "how will maintaining this look in the long run?". Literally everyone I know across multiple industries has been pressured to "try this specific tool at... stuff" and not asked "do you think this is a problem this kind of tool could help with?" There's a guy (and it's almost always a guy) in your org chart whose greatest joy is to flip-flop models all day and jerk off to how different ones produce different results, and he's spent the last two years excitedly telling leadership that some great breakthrough is right around the corner, that "Blurst-3.9 is just AMAZING, and we'd have been able to do that thing we didn't do last year if we'd only had it at the time, but it just came out, and I heard Chudvom-5 is gonna blow it out of the water!"