Answer
Users are randomly split into groups, each interacting with a different model version, and results are compared using real usage metrics to decide which model to keep.
Real-world example A company A B tests two versions of its AI recommendation model to see which one leads to more customer purchases.