>A human SRE never manually detects an error - we already have well-established anomaly detection implementations and wiring them to some ticket generation tool is also an established pattern.

I'm currently dealing with fallout at job because we were doing all this with humans with no alerts and we missed a couple major issues. This product could have prevented a lot of stress in my case, but it'd be a bit like a bandage on a missing limb.

That still begs the question though: There are existing tools and solutions that do this. Why not, and would this being AI make a difference?

"My boss would be more likely to approve it" is a cynical but valid answer.