AI could be ‘the best thing we have’ for health care’s biggest problems,’ former CMS chief says

Even with seemingly rapid adoption of some artificial intelligence (AI) tools, American health care still is in the “early innings” of using it, former CMS chief Seema Verma said.

The technology already is easing administrative work and could soon reshape everything from coding to prior authorization. That promise comes as an aging population, a strained workforce and rising costs converge on physician practices, Verma said. And some clinicians remain wary of using AI at all, she added.

“There’s a lot of fear. There’s a lot of concern, and I think some of that is valid. But I’m really encouraged about AI because when I think about some of these long-term problems that we have in health care, that there’s a real incredible opportunity to start addressing those problems,” said Verma, who served as Centers for Medicare & Medicaid Services administrator in the first term of President Donald J. Trump. She is now executive vice president and general manager of Oracle Health and Life Sciences.

Verma spoke Sept. 29 in a keynote fireside chat, “From Burden to Bandwidth: AI That Simplifies Care Delivery and Operations,” at the Medical Group Management Association (MGMA) 2026 Annual Conference in San Antonio. The conversation was moderated by Ranga Bodla, vice president of field engagement and marketing for the NetSuite Global Business Unit at Oracle.

Why is this moment different?

Looking back on decades of legislation and policy changes, Verma said health care still hasn’t cracked many of its core problems. Care is getting harder and more expensive, and burnout is growing. Part of the reason is that health care never embraced technology the way banking or retail did. Some communication methods still used in health care seem primitive now — she joked that if not for health care, “there would be no fax machine.”

Electronic health records (EHRs) didn’t deliver what was promised and in many cases created burden, she said, largely because the industry simply digitized paper processes. What’s different now is progress on interoperability combined with AI tools that she said are “perfect for medical care and for medical administration.” Coding, for example, exists to translate physicians’ notes into payment. With ambient listening that records a visit and generates the codes, Verma asked whether that step will even be needed.

What pressures are practices facing?

Verma described “the perfect storm” affecting health care right now: demand rising with an aging population, too few workers, mounting cost pressures. Workforce wasn’t really discussed during her time in Washington, she said. Patients now struggle to get primary care and specialty appointments. She described her own mother being told to wait three to four months for an appointment after losing about 40 pounds, quickly and unexpectedly.

Meanwhile, Washington isn’t expanding coverage, benefits or rates. Instead, it is cracking down on fraud and abuse and changing Medicaid, all of which will hit front-line providers, she said. AI isn’t a cure-all and may create new problems, she acknowledged.

“But for right now, I think it’s the best thing we have to solve some of the big problems,” Verma said.

What could AI do for physicians’ time?

Verma described a prior authorization process in which, once a physician orders a test, medication or procedure, AI starts the request, and predicts whether it will be approved. The physician signs off, the insurer’s automated system responds in real time and the patient is scheduled before leaving the office.

“That’s not why nurses and doctors went to medical school to be typing away at a computer and passing on all these policies. They want to treat patients,” she said.

In the short term, she expects that freed time to translate into more capacity and higher revenue for practices. Some data show physicians using ambient listening tools get better documentation, more accurate coding and a revenue uplift, she said. But practices must measure results over time, and the investment should pay for itself. Otherwise, she said, “you’re just layering on more technology.”

How should practices buy and manage AI?

Verma warned against repeating the EHR experience of spending heavily without getting results. AI is only as good as the data it can reach, she said. A prior authorization tool without access to the clinical record and the insurer’s requirements won’t do a good job, and incomplete data can lead to errors.

“You can’t have an AI strategy without having a data strategy, which means how am I going to bring the right information at the right time to the AI model so that they can work? And that's where you're going to hear a lot about AI data platforms,” she said.

Verma urged caution about “bolt-on” point solutions that do one task in isolation. Physicians and practice leaders should ask vendors how their products were tested, whether recommendations are transparent and traceable, and whether a human must sign off before a workflow proceeds. Once a tool goes live, practices should keep monitoring how it performs with their patients, she said.

What about skeptical clinicians?

Verma said she is increasingly concerned about a public backlash against AI. At one hospital deployment, some nurses said, “I don’t do AI,” and refused to try the tools. Others watched colleagues use them and concluded the technology would help rather than replace them.

Organizations can’t simply install a tool and expect results, she said. They need training, adoption plans and governance that answers practical questions, such as who acts when AI flags a patient at risk of sepsis and what happens when staff disagree with an AI recommendation. As for jobs, Verma said demand for care is so high that workers can be redeployed.

“Health care is a very human thing. I think it’s going to help the humans do their work and serve the other humans in a much better way,” she said.

What should leaders do next?

Her advice was to pause, resist the pressure to buy every new product and build an enterprise strategy for how data, AI tools and staff will work together. Vendors should be able to explain how they deliver AI safely and produce a return on investment, she said. Her time in government showed her that more money isn’t the answer.

“We don’t have enough money to keep putting money at the problem, and our processes have become so convoluted that they require a lot of humans, and that’s just not sustainable considering where we are with the workforce. So I’m excited for the future,” she said.

Medical Economics is in San Antonio at the MGMA Annual Conference, Sept 27-30, celebrating 100 years of MGMA, attending sessions and speaking with industry leaders. Follow our coverage on our MGMA conference page.