For years, the bottleneck for AI in healthcare has not been the technology. It has been the billing code. Original Medicare pays for visits, procedures, and a defined set of activities — none of which describe what an autonomous system actually does when it manages a patient’s blood pressure for a month between appointments. A provider deploying AI to keep patients out of the clinic was, in effect, paying to reduce its own revenue.
On July 5, 2026, that changes. CMS launches the Advancing Chronic Care with Effective, Scalable Solutions Model — ACCESS — a ten-year, voluntary payment model that pays providers a fixed amount per Medicare patient for managing a qualifying chronic condition, with full payment contingent on whether the patient actually hits clinical targets. It is the first federal payment instrument shaped to fit AI-delivered care.
The mechanism
The core construct is the Outcome-Aligned Payment, or OAP. Participating organizations receive a fixed annual amount per enrolled beneficiary, paid in quarterly installments, with up to half withheld pending demonstrated results. Full payment is earned only when patients hit measurable targets — blood pressure control, HbA1c reduction, depression or pain scale improvement — relative to their own baseline. For the 2026 cohort, hitting targets for at least 50% of patients earns the full OAP; the threshold rises each year.
Three design choices make this a vehicle for autonomous care specifically:
The payment is modality-agnostic. CMS does not prescribe a care model. Care can be in person, virtual, asynchronous, or fully software-mediated. An organization that manages a hypertensive patient with a wearable, an algorithm, and a once-a-quarter human check-in earns the same payment as one delivering weekly in-person visits — provided the blood pressure comes down. The only thing that matters at reconciliation is the outcome.
The payment is per condition, not per total cost of care. Unlike an ACO, ACCESS participants are not on the hook for the patient’s hospitalizations or specialist visits. They are accountable for the specific outcome in their clinical track. This lets AI-native organizations enter without the actuarial infrastructure required to bear full risk.
The data plumbing assumes software. Participants must use FHIR-based APIs to report outcomes and share clinical updates. That is exactly the integration scaffolding autonomous systems need to read and write structured data at scale.
Why this changes the economics
Under fee-for-service, the marginal chronic-care patient costs you a clinician slot and earns you a visit fee. Economics scale linearly with labor. Under an OAP, the marginal patient costs you whatever software and supervision it takes to keep them at target, and earns you the same fixed payment as every other patient in the track. Economics scale with software margins.
This is the condition under which autonomous care becomes a business rather than a research project. A continuous glucose monitor plus a model that flags emerging hyperglycemia has high fixed cost and near-zero marginal cost per additional patient. A behavioral health platform that triages symptom reports with NLP and routes only the urgent ones to a clinician has the same shape. Both can manage thousands of patients with the same core system that managed a hundred. Neither has had a clean revenue line in Original Medicare. Under ACCESS, both earn the OAP if they produce the outcome.
The commercial alignment matters too. Private payers covering roughly 165 million lives across Medicare Advantage, Medicaid, and commercial coverage have committed to aligning with the OAP approach. That begins to standardize a payment shape across the system, not just within CMS.
The caveats
ACCESS is voluntary, narrow, and conditional. It covers four clinical tracks, not the full spectrum of Medicare spending. It expands only if the CMS Office of the Actuary certifies it improves quality without raising costs, or reduces costs without harming quality. And outcome-based payment can create its own distortions — particularly the incentive to avoid harder-to-reach patients in whom targets are harder to hit. CMS has built in baseline-relative measurement and public reporting to mitigate this, but it is not solved on paper.
What this actually is
The dollars in the pilot are modest. The template is not. ACCESS is the first time Medicare has paid for a result rather than an activity in a way that is structurally compatible with autonomous delivery. If it works, the case for extending OAP-style payment to other conditions becomes hard to refuse — and the question of whether AI can deliver care cheaply stops being theoretical, because for the first time it has a payer willing to pay for it that way.