To me, there is no good or bad method. Probably a mix of Artificial intelligence and fighting services would be the best.
I am sure the groups dealing with bad behavior have already some protocols established and some of those could be programmed in bots using some AI to do the job. More sophisticated cases would still have to be managed manually but there would be less work for the groups which would be able to focus on those specific cases where human intervention is a must to take part in the final decision.