Yeah, my thoughts are that the best usecase for LLMs is as an input to a process that is then controlled by a human For the summary thing, I think we eventuall

Yeah, my thoughts are that the best usecase for LLMs is as an input to a process that is then controlled by a human For the summary thing, I think we eventually settled on a workflow that generates the summary using LLM, then we store the output in our dataset for the report. This also allows the human building the report to modify it if necessary