It doesn't work. Too many variables in life to capture all of them.

I'm imagining 20 years from now someone making IntervieweeBot, who will pass all the tests and STILL not get the job because the interviewer didn't like that indignant look in the robot's eye. Or MLHire, a machine-learning algorithm that reliably fails to predict what the employee will want to be doing at the end of their contract. It will make millions in a Series A, become hugely popular, then promptly forgotten after [Elon Musk-like in 20 years] fails to deliver.

EDIT: 80% chance of male pronoun considering the industry, but wasn't sure.