> Surely anomaly detection is better done by a machine?

I've worked with machines. I don't think this is a good idea.