People, process, technology: AI success takes more than software, former CMS chief says

Medical groups that pair artificial intelligence (AI) with a clear data strategy and redesigned workflows will pull ahead of those that simply buy and “bolt on” tools, a former top federal health official told practice leaders.

With demand for care rising, staff hard to find and costs climbing, AI can create badly needed capacity. But it takes planning around people and processes, and some staff will resist using it.

“I think those institutions that are doing both the people, the process, and the technology all together have a better chance of being successful, more competitive,” said Seema Verma, executive vice president and general manager of Oracle Health and Life Sciences and former administrator of the Centers for Medicare & Medicaid Services (CMS).

Verma spoke Sept. 29 in a keynote fireside chat at the Medical Group Management Association (MGMA) 2026 Annual Conference in San Antonio. Ranga Bodla, vice president of field engagement and marketing for the NetSuite Global Business Unit at Oracle, moderated.

What should you ask before buying?

Verma discussed the state of AI technology ranging from big-picture trends in the nation’s capital, to implications for medical practices and offices.

Health care organization leaders may be experimenting with AI for single tasks, such as revenue cycle work, clinical notes, coding or summarization, Verma said. The next challenge is making those purchases work together. If the revenue cycle team buys one AI agent and physicians buy a separate coding agent, leaders need to know how the products will share data and operate at the enterprise level.

She suggested asking how a model was trained, because training on older medical records may not reflect current care, and whether it can pull real-time information. An AI tool that lacks a patient’s full history, such as an allergy, can produce flawed recommendations. A prior authorization tool without access to both the clinical record and the insurer’s requirements “it’s not going to do that great of a job,” she said.

Verma cautioned against “bolt-on” point solutions. A tool that calls patients to prepare them for a preoperative visit may work well, but if it can’t confirm that an operating room, staff and supplies are available, it’s doing only one piece of the job. She compared bolt-on tools to the plug-in GPS units drivers once used; they gave directions but couldn’t work with the rest of the car the way more current built-in systems do.

Flexibility matters, too. Leaders should look for an open architecture that lets them build their own agents or add best-of-breed products instead of getting locked into one vendor, she said.

How do you measure the return?

Organization leaders should decide up front what they want from each tool, Verma said. With ambient listening, early indicators may be physician satisfaction and retention. Longer-term measures include whether physicians have capacity to see more patients and whether better documentation leads to more accurate coding and higher revenue.

The practices that realize value fastest plan for it from the start. She described organizations that gave physicians time to get comfortable with ambient listening, then adjusted schedules to use the capacity it created. Without that step, she said, practices may not see the return they expected.

The work doesn’t end at go-live. Leaders should keep monitoring how each tool performs with their own patient population and whether it is paying for itself.

“But it’s not something like you’re going to put it in and then just walk away,” Verma said.

How will roles and processes change?

AI will change how work is organized, Verma said. A task that once required 10 people might need only two to oversee the AI, leaving leaders to decide how to redeploy the others. Alternatively, the same team could handle many more patients. She doesn’t expect widespread layoffs, because demand for care is so high.

Processes need to be rethought as well. If an AI tool flags that a prior authorization request lacks enough data to be approved, someone has to decide what happens next. If staff disagree with the tool and want to submit anyway, the organization needs a governance process for resolving that conflict. Verma noted that AI governance was one of the most popular topics at Oracle’s health care technology summit held the previous week.

How do you win over reluctant staff?

Adoption is its own project, Verma said. Not everyone wants to use AI, and leaders shouldn’t assume that implementation alone will make things work. Training and comfort with the tools take deliberate effort.

At one hospital deployment, some nurses refused even to try AI tools for ambient listening, orders and summarization. The organization didn’t force the issue. Other staff were more interested, and watching the tools at work convinced them AI would act as an assistant rather than a replacement, she said.

Media coverage of AI is shaping how employees react, and fears about job loss are common. Verma said leaders should be ready to explain how the tools work and how safety is handled.

“They need to be answered by the vendor, but I think for organizations, they need to be able to explain that to their end users because there’s going to be questions about this,” she said.

What’s the payoff for getting it right?

Organizations that use the right tools with an enterprise strategy will have “a strategic advantage,” Verma said, through more appointment capacity, shorter waits, fewer denials and cleaner claims. Those that don’t may fall behind. The next gains will come from agentic workflows, in which AI handles an entire process, such as generating the bill after a visit and identifying what needs to change before submission to prevent denials on the front end.

Her advice to leaders heading home from the conference was to pause before buying more products, think through how data and tools will work together, and hold vendors accountable.

“So I think it’s incumbent on the vendors out there to explain how they’re doing AI safely and how are they going to deliver a return on investment for their customers,” she said.

“I think you've got to think about this more holistically,” Verma said. The last couple years have brought some AI success stories, she said.

“That's great. We're experimenting, we're getting to know AI,” she said. “But as you start to think of more enterprise, how am I going to use AI across the enterprise with all of our tools? Take a moment to really think about that overall strategy. I think there's some foundational things that need to be done.”

Physicians Practice 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.