This slow rate of adoption of new technology is not unusual in healthcare, which has a long tradition of skepticism. The history of novel technology in healthcare is one of great promises to 'transform care' followed by failure to deliver meaningful outcomes for patients. There is an existing system in place for evaluation of new healthcare technologies. A core principle of this system is general unwillingness to trust anything until it's gone through rigorous clinical testing. Widespread adoption occurs after expert panels examine the available evidence and make summary recommendations. This method is imperfect but protects patients against harm posed by new technologies. Zachi Attia, director of AI at Mayo Clinic and Hamid Ghanbaro, clinical lecturer in internal medicine at the University of Michigan, discuss the principles needed for building, validating and implementing machine learning algorithms for healthcare. We will use an algorithm used for prediction of congestive heart failure as a case study to illustrate the challenges and opportunities facing AI algorithms and presented from a physician and AI engineer point of view.