There's to me a sole unimpeachable use case... Two SMEs of adjacent fields can work with an LLM SME to find overlaps in research. This is a manual slog of evaluating hints at patterns and holistically trying to discover something and the LLM is only a part of it. All of the other use cases I've heard of have some kind of religious undertone or assumption and leads to people wanting to make blanket summarizing statements to avoid digging into inconvenient realities.