How do you determine which test cases are the best candidates for automation versus manual execution?
