That probably means there wasn't a good testing process for patching and there wasn't adequate redundancy. In theory, if a patch breaks one server it shouldn't matter.
In reality, patch testing stacks up and gets behind and redundancies are rarely tested. That is expensive, time consuming work which probably isn't worth the time of someone who is already underpaid and overworked. And fuck! If patch and redundancy testing ever breaks anything prod for whatever reason, the person who was testing everything gets blamed and fired so nobody is going to volunteer for that.
As with most things, I expect it'll help the guys who know what they're doing do their thing faster and more efficiently.
I don't expect it to replace nor be a effective substitute for a properly trained pen tester.
It might be helpful to developers to fast track security testing, but I think there's already a wide array of "non-AI" tools that accomplish that? Don't know a lot about how it.couod affect that side of things.