Unwarranted variation and overprescribing of outpatient opioid medication have long been a problem in healthcare. With opioid related deaths accounting for more than half of all the drug overdose related deaths in the United States and the average duration of opioid prescriptions increasing, healthcare organizations have the responsibility to reduce the opioid utilization. Locally, Advocate Aurora Health (AAH) has made an effort to reduce opioid medications prescribed; however, has struggled to make any significant impact. Traditional analyses and summary statistics were not sensitive enough to detect clinical variation within a large, high-performing organization. Machine learning (ML) is effective at uncovering subtle patterns within large amounts of data and identifying unwarranted variation in opioid prescribing (provider and specialty level). Using unsupervised ML, opioid prescriptions for 56,000 patients from 1,500 providers were evaluated, and an algorithm that detected unwarranted variation in prescribing was developed.