Has anyone integrated AI tools into their code review process?

I've been using tools like CoPilot for code reviews, and my experience has been pretty mixed. It occasionally points out logical flaws but often proposes suggestions that go against what I intended. Sometimes, it feels like I'm fighting against it. Overall, I find it useful but definitely not reliable.

We actually use AI tools at work for our PRs. They catch a lot of small errors that can easily be overlooked, like naming inconsistencies or comments that need updating. While it's not perfect and sometimes gives random feedback, it's definitely helpful, especially in DevOps where we deal with config files that have fewer checks available. Still, we always make sure a human does a final review before approving anything.

I've tried a couple of AI tools for PRs and to be honest, the results are pretty mixed. They can help with minor issues but also generate a lot of noise. So, it's kind of a hit or miss.

We implemented Coderabbit alongside our manual review process, and it has been mostly beneficial. It helps identify minor bugs and violations in naming conventions, which saves time. But on bad days, it can be a real headache, especially when its suggestions don't make sense. But those discussions about its findings do lead to productive outcomes sometimes!

We've had good success with AI tools as extra reviewers. They help identify simple errors like typos and improve our overall workflow. Sure, you need to validate their suggestions, but they really speed things up when used correctly. It's good to remember, though, human oversight is still crucial.