For years, the biggest blocker to deploying AI in healthcare wasn’t the technology. It was the billing. Medicare pays clinicians for time spent with patients. There’s no code for an AI agent that calls a patient at 2 a.m. to check on their blood pressure, coordinates a housing referral, or reminds them to pick up their meds. That just changed.
A new Medicare program called ACCESS (Advancing Chronic Care with Effective, Scalable Solutions) goes live July 5. It’s a 10-year CMS program that pays based on health outcomes, not activities. That one shift creates a reimbursement path for AI agent workflows for the first time in federal healthcare. If you operate in the healthcare AI payment model space, this is the story you should be paying attention to.
What happened
- CMS selected 150 participants for the ACCESS program, which launches July 5, 2026.
- The program covers six chronic conditions: diabetes, hypertension, chronic kidney disease, obesity, depression, and anxiety.
- Participants receive predictable payments for managing qualifying conditions but only earn the full amount when patients hit measurable health goals (lower blood pressure, reduced pain, etc.).
- Traditional Medicare reimburses for clinician time. ACCESS doesn’t care whether a human or an AI agent does the work, as long as the outcome improves.
- The first cohort includes AI doctor startups, virtual nutrition therapy providers, connected device companies, and wearable makers like Whoop.
- The program was designed by Abe Sutton and Jacob Shiff at the CMS Innovation Center. Both are former startup operators (Sutton was a healthcare VC, Shiff was a healthcare founder).
The numbers
- 150 organizations selected for the first ACCESS cohort.
- 10-year program duration.
- Pair Team, one participant, claims 1 in 4 hospital visits and 1 in 2 ER visits are avoided when patients are in its care.
- Pair Team has partnerships giving it access to roughly 500,000 potential patients, with a goal of 1 million within three years.
- Pair Team employs roughly 850 clinical professionals and reports revenue above nine figures.
- A 2023 Congressional Budget Office analysis found the CMS Innovation Center increased federal spending by $5.4 billion in its first decade rather than producing projected savings.
- Digital health funding hit its highest Q1 total since the pandemic this year, with AI companies capturing the bulk.
5 things healthcare operators need to know about ACCESS
- Outcome-based pay means AI-first operations win. CMS is reportedly paying less per patient per month than many participants expected. That’s intentional. As Pair Team CEO Neil Batlivala put it: “If you want to build a model that truly incentivizes the use of AI, the reimbursement rates have to be low.” The economics only work if you’ve automated most patient interactions.
- The six covered conditions are massive. Diabetes, hypertension, chronic kidney disease, obesity, depression, anxiety. These aren’t niche. They cover a huge share of the Medicare population. If you’re already managing patients with these conditions, you’re sitting on an ACCESS-eligible workflow.
- AI agents can now do the between-visit work. One participant, Pair Team, deployed a voice AI agent called Flora nine months ago. Flora handles intake, coordinates referrals, and runs check-ins 24 hours a day. Patients routinely talk with Flora for over an hour. Under old Medicare rules, none of that was billable. Under ACCESS, it is.
- Data privacy risks are real. Patients in ACCESS are sharing deeply sensitive information (conversations about housing instability, mental illness, substance use) with AI systems feeding into federal infrastructure. CMS has a documented history of data breaches, including exposed Social Security numbers. If you’re building for this population, your security posture has to be airtight.
- Most of the tech world hasn’t noticed yet. ACCESS has barely registered outside health tech trade press. That’s a window. The organizations that move fastest to build compliant AI workflows for these six conditions will have a significant head start before the broader market catches on.
The hot take
This is the most important thing to happen in healthcare AI in the last five years, and it’s not close. Forget the model benchmarks. Forget the chatbot demos. The reason healthcare AI has underdelivered isn’t capability. It’s that nobody could get paid for deploying it. ACCESS fixes that. It tells every healthcare organization in the country: if your AI agent keeps a diabetic patient’s A1C down, Medicare will pay you for it. The organizations that built AI-first clinical operations before this announcement (like Pair Team, which has been at it for five years) are about to look like geniuses. Everyone else is about to scramble.
The Agency OS play
If you run a healthcare practice, a home health agency, or any organization managing Medicare patients with chronic conditions, here’s what to do this week. First, pull a list of your current patients with diabetes, hypertension, chronic kidney disease, obesity, depression, or anxiety. That’s your ACCESS-eligible population. Size it. If it’s meaningful, you have a business case to build around.
Second, audit your between-visit workflows. What happens after a patient leaves your office or hangs up the phone? If the answer is “not much until their next appointment,” that’s the gap ACCESS is designed to fill. Start mapping the specific touchpoints where an AI agent could check in, coordinate referrals, confirm medication pickup, or flag a deteriorating metric. You don’t need to build Flora overnight. You need to know exactly where automation would improve an outcome you can measure.
Third, get serious about your data infrastructure. ACCESS pays on outcomes, which means you need clean, connected data showing that your interventions (human or AI) actually moved the needle on blood pressure, pain scores, or hospital utilization. If your EHR data is messy, your reporting is manual, or your patient records live in five different systems, fix that before you try to layer AI on top. The reimbursement only flows when you can prove the result. Start there.