It's just marginal at most anything it does. I use it to monitor my inbox and chats. It misses stuff all the time. I've actually had to use claude to create very specific prompts for Copilot agents. It helped a lot, but overall, it's just not very good.

I tried to get my company to buy my security team Anthropic tools, but they all fought me on it. My boss, my employees, everyone.

I told them that while we're here trying to make Copilot help us work through things, the bad guys are out there using anything they can get their hands on. In response, my boss said that we should focus on patching and not spend time concerning ourselves with what bad actors are doing to leverage AI.

I never understand the current anti-AI view for a lot of people. When regular people use it to try and help them do their jobs, they turn out slop, when adversaries use it to successfully and rapidly mount cyber attacks, all of a sudden it is scary effective? Which is it, because it can't be both. Either AI is shit and makes slop or it is effective in the hands of people that learn how to use it.