That's why alerts can have different priority levels. So, less serious issues can be addressed during normal working hours (eg. disk is 80% full). Maybe LLMs will figure out the correct P level for something like disk usage, but it's unlikely to get it right for things that are particular to your application. Maybe use an LLM when you're creating the alert to auto fill the priority level. That can then be verified by someone. Don't silence an alert based on what an LLM thinks though.