Branding the mistakes LLM's get as hallucinations was sort of brilliant in my opinion, it was a good way to disguise the fact that LLM's are mostly just really lucky. In my anecdotal experience it hasn't worked out too well though, so maybe it wasn't that brilliant? Anyway, parts of how the AI impacts our business (solar energy + investment banking) has been through things like how Microsoft Teams was supposed to be capable of transcripting meetings with AI. Now, we're an international organisation where English is at best the second language for people, and usually the third, so this probably has an impact on it, but it's so bad. I know I'm not personally the clearest speaker, especially if I'm bored, but the transcripts the AI makes of me are so hilariously bad that they often become the foundation for the weekly friday meme in IT. Which may be innocent enough, but it hasn't been very confidence building in the higher ups who thought they wouldn't have to have someone write a summary for their meetings, and in typical top brass style didn't read the transcripts until there was some contract issue.
This along with how often AI "hallucinates" has meant that our top decision makers have decided to stop any AI within the Microsoft platform. Well every thing except the thing that makes power point presentations pretty. So our operations staff has had to roll back co-pilot for every non-IT employee. I don't necessarily agree with this myself, I use GPT quite a lot and while github co-pilot might "just" be fancy auto-complete it's still increased my productivity quite a lot as well as lowering my mental load of dealing with most snippets. That isn't how the rest of the organisation sees it though, they see the mistakes AI makes on areas where no-mistakes are allowed, and they consider it untrustworthy. The whole Microsoft debacle (I used chatGPT to tell me how to write "debaclable") where they wanted to screenshot everything all the time really sunk trust with our decision makers.