In contemporary medicine, many activities that resemble scientific work are classified as quality improvement (QI), which has fewer ethical obligations than clinical research. Yet in practice, the boundary between these categories is a historically porous one, involving similar biomedical expertise. This project traces the evolution of QI from its origins in management science to use as a category for governing AI in healthcare. Early twentieth century statistical techniques for measuring quality were pioneered at Bell Labs and later applied to Japanese industrial production, before eventually making their way into corporations including Ford Motor Company. In the 1980s, shepherded by Donald Berwick of the Harvard Community Health Plan, ideas about QI catalyzed the Institute for Healthcare Improvement and were institutionalized federally with the 1999 Healthcare Research and Quality Act. Today, much AI work in medicine is classified as QI, raising questions about what constitutes ‘research’ and how to govern medical activities.