Somerville, Mass.-based Mass General Brigham is spending more on AI than it’s getting back in a straightforward financial sense — and its CFO said that’s by design, not an oversight.
Asked whether AI is currently a cost drain, a long-term investment or a source of savings, Mass General Brigham CFO and Treasurer Niyum Gandhi didn’t hesitate: “We are investing in it net.” But he was quick to explain why.
“That’s actually not because we haven’t yielded returns yet,” he told Becker’s in a Sept. 23 interview at the health system’s World Medical Innovation Forum in Boston. “It’s actually because a lot of the places that we’re investing, we’re not trying to generate a financial return.”
Mr. Gandhi said he sorts AI spending into three categories — a framework he was careful to describe as his own thinking, not Mass General Brigham’s official policy.
The first category is using AI to do things that weren’t possible before, largely in research. “That’s not a productivity play,” he said. “That’s an outcomes play, whether it’s research or clinical care.” That kind of spending, including AI-assisted protein-folding research, “is always going to be a net investment because we aren’t looking for an ROI there,” he said. The same logic applies when the goal is a better clinical result rather than a discovery: “The money’s only there so we can deliver the outcomes.”
The second category covers AI applied to something the system was already doing, done more accurately — an upgraded predictive model for which patients are likely to decompensate, for instance, so care managers can intervene sooner. Those investments sometimes generate a return, Mr. Gandhi said, particularly when they help the system avoid Medicare readmission penalties. But even without one, he said the math still works: “Sometimes they have no financial return, but we decrease the likelihood somebody’s going to die in the hospital — pretty worthy thing to spend money on.”
The third category — using AI to do existing work more efficiently — is where Mr. Gandhi’s numbers get concrete. He pointed to ambient documentation, which Mass General Brigham scaled from an early pilot to hundreds of physicians. When the technology’s price tag came back as a seven-figure annual investment, Mr. Gandhi said he didn’t wait for a full ROI case. “Just do it,” he said, “because if we can’t find that money to save somewhere in the organization, shame on us — myself and our COO.” Part of his reasoning: Physicians who report being burned out commit significantly more medical errors, he said, and the tool has been associated with a meaningful drop in self-reported burnout among Mass General Brigham clinicians. He argued that kind of return doesn’t show up cleanly on a balance sheet — fewer errors, fewer malpractice claims, physicians choosing to stay in medicine longer — but it’s real.
That reinvestment logic extends to how Mr. Gandhi thinks about margin generally. Mass General Brigham has a target operating margin it hasn’t yet hit, he said, attributing the gap largely to inflation. But if AI ever helped the system add several points of margin, he said, that wouldn’t necessarily mean lower prices or fatter reserves. “If our margin went up 3% on a $25 billion organization, that’s $750 million,” he said. “I have an enormously long list of places where I would like to lose money to further our mission.” Because Mass General Brigham has no shareholders, he said, any such gain would likely go toward the system’s own version of a dividend: community health, care for the underserved, research and education.
Mr. Gandhi said he’s less certain what happens to overall healthcare spending if AI succeeds at lowering the cost of care, pointing to the Jevons paradox — an economic principle in which making something cheaper increases demand for it enough to offset the savings. If AI-driven chronic disease management becomes dramatically cheaper and more convenient, he said, more people might use more of it rather than the same number of people using less. “We are too early to tell whether that will happen in healthcare,” he said, adding that it could mean the health system ultimately needs more clinicians, not fewer, to meet the resulting demand.
