This sounds like actually useful AI - maybe.
The AI chatbots that make up random stuff when they don't know the answer, like some bullshitter down the pub or a verbally incontinent politician, are not useful. Unless you are interested in randomly fabricated lies mixed with truth, which almost no-one is, they are pointless. Even propagandists, marketers, etc, want some chosen lies mixed with the truth. There's no point in having propaganda about your Glorious Leader that randomly adds untrue good things and untrue bad things about him. You just want untrue good things.
AI systems that summarise existing information, and don't try to make things up when they don't know, are much more useful. Zoom taking a transcript of a meeting and working out what the major topics were and any decisions stated explicitly in the meeting would be incredibly useful.
Remember the Librarian in Snow Crash, our science-fiction prototype for all this: it cannot guess what people are thinking, it cannot make leaps of intuition, and it resists speculation on such matters because it cannot know if what it is guessing is true. That is a useful AI to help humans process information.