Medicare is launching a pilot program that tests whether health tech companies — including those that heavily rely on AI — can lower medical costs while improving care.

Why it matters: We've discussed why many experts think AI will actually increase health care costs in the near term in part by generating more health care and making spending levels even more unsustainable.

• But there's a genuine belief that emerging technologies can improve people's health if we fundamentally change the way the system pays for care.
• How to pay for Doctor AI is exactly what Medicare is experimenting with.

Driving the news: CMS' ACCESS payment model launches Sunday, with more than 150 participating health care organizations.

• Medicare will pay providers based on whether patients reach specific, measurable outcomes, like lowering the blood pressure of a patient with hypertension — not on how much care they receive.
• The focus is on chronic conditions that CMS says affect more than two-thirds of Medicare enrollees, including high blood pressure, heart disease, diabetes, chronic pain and depression.
• And the low fixed payments — which translate to upfront payments as low as $7.50 per patient, per month — almost ensure that providers will make heavy use of technology and automation to reach desired outcomes.

The big picture: Moving away from paying for individual health care services has been a white whale for policymakers and the health care industry for as long as I've been covering it.

• So far, there's been very limited success, and national health care spending reached $5.7 trillion in 2025.
• But the market is increasingly warming to performance-based contracting, according to a recent Peterson Center on Healthcare report.
• Backing from Medicare — the nation's largest payer — will almost certainly add momentum to its use if the model is successful.

Yes, but: There are multiple ways that AI and other emerging technologies influence health care spending, and this only addresses part of the overall concern.

• AI "bot wars" over payment rates and care approvals are already having an inflationary effect.
• "It's very clear on the administrative side, AI is increasing spending. On the clinical side, it could go either way," Peterson Health Technology Institute executive director Caroline Pearson told me.

The ACCESS payment rates are much lower than the industry was expecting when the program was announced last December — which may have scared off some digital health companies that require intensive human labor.

• "It's very clear that CMS is trying to encourage the use of very tech-heavy solutions — solutions that are leveraging tech rather than humans — and that's the only way you can succeed under this model," Pearson said.
• Some of the biggest names in the digital health world are staying on the sidelines, possibly because they're designed for fee-for-service payments and can't generate enough cash flow per person.
• "Right now the digital health space is a mix of companies that were built as tech-first and those that were built as human-first, tech-enabled," Pearson added, and they have "very different underlying cost structures."

What they're saying: "I think if you are delivering care in a way that's software-first with doctor supervision, you can totally make the rates work," said Brandon Ballinger, co-founder of Empirical Health, one of the ACCESS participants.

• "If AI can do things that would normally take a physician a lot of time, you can deliver outcomes more efficiently," he added.
• "I think you see that effect in sort of the previous generation of digital health companies who say the payment rates are too low. Implicitly, that's because they're doing stuff with human time."

Yes, but: Low reimbursement rates like these may benefit tech-heavy companies with small staffs and low marginal costs, not necessarily the ones that deliver the most value.

• "What I think will be important is, as the evidence base in ACCESS improves, can they make the pricing more dynamic," Pearson said.
• And all of this requires some level of trust in the care AI can provide.
• "We are putting a lot of trust into AI bots replicating the judgment and trust of humans," Venrock partner Bob Kocher said.

By the numbers: Peterson Health Technology Institute has been measuring the impact of various health technologies on health spending for several years now, and the results are all over the place.

• Remote patient monitoring for diabetes is estimated to increase commercial spending by around $2,000 per user each year, while gastrointestinal management services reduce spending by nearly $1,900 annually.
• Other similarly ambiguous assessments underscore ACCESS' central question: whether paying for measurable outcomes, rather than digital services themselves, can produce better health at lower cost.

What we're watching: Commercial insurers have pledged to incorporate outcomes-based payments aligned with ACCESS in their coverage by the beginning of 2028.

• That could extend the experiment well beyond Medicare — but without knowing whether the new arrangements are workable.