The company that I work for is using Microsoft's generic LLM, but has supplemented it with a huge amount of it's own corporate data to help people get accurate answers for company-specific questions. We've been told that there is a moat around the company data that outsiders won't get access to (ie, it won't be used to train the generic model).
There's still lots of unanswered questions in the internal data protection space, but it's super interesting to see.
That's cool! And probably also the future. For example, in order for the AI to give the correct answers in a given game context, it needs more custom training data. One of the reasons why I wanted to get practical with ai agents - getting beyond the hype so to speak.