I get the feeling that this is one of those things where you s/AI/statistics/g. Doctors using a predictive statistical model trained on thousands of patients' worth of data faring better than doctors using the seat of their pants makes total sense.