This is how I have found AI to be very powerful. If my tests pass, I know my code won't break. I spend more time on tests than any other part of the code-base. I am curious who else is building in this manner? I use the statement in the title because I feel that is the bar you have to meet to get the AI to work well, but once you meet that bar lately it has been working pretty flawlessly for me.
But when I have to work in an area, the first thing I look for is coverage, and if it's not there, I add "fixing tests" to enforce the existing behavior before making changes so I'll know when I broke something.