I'm optimistic that AI solutions can really do great work to improve false negative/positive rates in highly manual systems like this.

With that said, it's a slippery slope, and will quickly lead to different types of inconsistencies in policy, enforcement, and all the other problems that come with taking humans out of the loop.