Some medical technologies that use artificial intelligence might benefit patients but result in a drop in health system revenues. This could make widespread adoption of AI in medicine a tough sell.

Healthcare is a world of perverse incentives. One person’s waste is another person’s annual bonus. If you have shares in Novo Nordisk, you don’t really want people to eat more fruits and vegetables, because then sales of diabetes meds might drop. The same is true on a larger scale for hospital systems that get paid to conduct tests and perform surgeries instead of keeping people healthy.

It doesn’t pay—yet—to keep people healthy. Many entrepreneurs and government leaders are dragging the American healthcare system away from the fee-for-service model, but the road to pay for performance medical care is long and winding.

HeartFlow is the perfect example of how a positive change for patients could mean a drop in health system revenues. The company’s goal is to improve care for heart disease by avoiding both under- and over-treatment. HeartFlow takes data from a CT scan and turns it into a 3D model of the veins and arteries that supply blood to the heart.

“If our system is adopted, the net annual savings would be $2 billion for one commercial payer,” said John Stevens, MD and president of HeartFlow. “We could avoid 250,000 unnecessary procedures and even save 30,000 lives.” Stevens shared these shocking numbers at HLTH 2018 in early May in Las Vegas.