I have consistently been trying to use AI for the actual tasks I need to complete for work over the last year or so (we are gently encouraged to try it, but thankfully not forced). I have found it to be “successful” at maybe 1 in 10 tasks I give it. Even when successful, the code quality is so low I edit heavily before it’s pushed and attributed to me.

I think the problem I have is I rarely work on boilerplate stuff.