Advice for automating AI agent QA post-deployment?

I’m at a mid-sized SaaS with a team of six. We’ve been doing manual testing for three years and we’ve gotten good in the way that anyone does with experience. Pattern recognition, intuition, and tribal knowledge basically. The problem is that all of the knowledge lives inside our heads. Test coverage decisions are essentially vibes. We trust things that haven’t broken recently and test things we’re scared of lol. Last quarter there were two production inc