Well, maybe you've been blessed with a better pool of juniors than the ones I've worked with. The ones I've worked with require exactly what an AI would require -- completely defined tickets with no ambiguity, lots of deliberate review, no guarantees that what they produce is bug-free.

At least with AI if it's not working out you just shut it off and it's done. If it's a person's livelihood that process takes a huge burden on all involved.

If you do your hiring (and then mentorship) right, even such juniors can skyrocket in 1-2 years. AI won't do that. So, in case of willing juniors, you get much, much more "out of them" soon, if you invest a little. They will be able to tackle problems you can't solve by "throwing more AI at it"