I think it is better to use real humans for checking violations and sending emails. Yes, humans inevitably will make mistakes. However, unlike a neural network algorithm they can be reasoned with and can correct their mistakes later. I'd take much longer issue resolution with hired operators over the fast AI violation check.