This is absurd. What you want for any critical system are human interpretable rules which tell the operators how the system is currently functioning. Gathering

This is absurd. What you want for any critical system are human interpretable rules which tell the operators how the system is currently functioning. Gathering that information is obviously useful, but if the human staff isn't constantly interpreting it, something is going very wrong. I can not imagine this AI ever having any use, personell failures will not be mitigated by an LLM and routine operations need to be human supervised. LLMs are not useful here, they are useful if the rules making up a system can not be formalized and you need statistical methods to extract information.