GPT-LLM-ML: Use as a natural-language interface to a reliable and ideally well-structured database is, I think, one of the use cases for LLM's that is actually likely to be very useful soon. And Dave Guarino is on it!:

Dave Guarino: Using AI to make sense of policy documents <daveguarino.substack.com/p/using-ai-to-…>: ‘I've been experimenting with using AI to do "policy reasoning." That’s because fundamentally this is what staff have to do to answer questions about often complicated and complex rules.…. I think… pairing a policy question in an end-user’s own language with the specific policy language that a government staff member would use to answer it… [is] of promise…. I am working on experimenting and evaluating more. There are definitely ways that this would need to be hardened to be used in something closer to a production (real users) environment. Though my first step would most likely be to have a human in the loop, with something like an email support desk or a content web site as the service channel. (That mitigates a lot of the risk of having it be instantaneous)…