Your personal failure to use ML successfully has no bearing on its wider deployment and success. Overfitting strongly implies that you or whoever was doing it just didn’t know what they were doing.
Dozens. I’m a consultant and it’s what I do. The modeling is very easy. The operational changes are the hard part. Basically any random task of. A sort of maintenance, production, replacement, procurement or scheduling task can save 10-30% with a simple model. There’s a long way to go but this stuff is everywhere.
The world hasn’t even fully realized the productivity value of spreadsheets and email.
ML is used everywhere.