Medicare's Hidden AI Payment Model Most Tech Builders Are Missing
A new Medicare payment model is quietly creating the first reimbursement pathway for AI agents that monitor patients between clinical visits, coordinate care, and manage medication adherence. Most of the tech industry hasn't noticed it exists.
Original sourceMedicare's new Principal Illness Navigation and Caregiver Training payment model contains a largely overlooked mechanism: it can reimburse for non-physician care coordination activities that AI agents are technically capable of performing. For the first time, there's a credible path for a software system to perform billable work inside the U.S. healthcare reimbursement structure — not by replacing a doctor visit, but by filling the massive gap between them.
The core problem the model addresses is structural. A patient discharged after a cardiac event might not see their care team for weeks, but the highest-risk period is often the first 72 hours at home. Today, no payment mechanism exists for an AI system that calls to check in, confirms medication pickup, or coordinates a social services referral during that window. This new model creates the billing codes and organizational requirements that could make such an agent reimbursable.
The policy details are dense enough that most startups building AI health tools haven't mapped their products to these codes — and most investors evaluating those startups haven't asked. But for the handful of companies that do understand the reimbursement architecture, this represents a rare instance of government policy moving ahead of commercial technology awareness rather than lagging behind it.
The catch is real: the model requires participating providers to meet specific organizational criteria, and the reimbursement rates are modest. But as a proof of concept for AI agents embedded in federally reimbursed care pathways, the signal is significant. Healthcare AI that can bill is categorically different from healthcare AI that can only be sold as software.
Panel Takes
The Founder
Business & Market
“The buyer here is the provider organization, not the payer, and the budget is the new billing codes themselves — that's the wedge. Any AI health startup that hasn't mapped its product to these CPT codes is leaving a distribution channel on the table and building a pure SaaS sale into one of the hardest institutional buyers on earth. The companies that figure out 'our agent is the thing that makes you eligible to bill this code' will have a moat that's literally written into federal regulation.”
The Futurist
Big Picture
“The thesis this policy bets on is specific and falsifiable: AI agents performing longitudinal patient monitoring will be more cost-effective than human care coordinators at the between-visit layer of care, and CMS is pricing that in before the technology is mainstream. The second-order effect that nobody is talking about is credentialing — once an AI agent is a reimbursable care actor, the question of what it's allowed to do and who's liable for it stops being a product question and becomes a regulatory one, which creates enormous barriers to entry for late movers. This is the healthcare AI infrastructure layer, and it's being built in policy documents, not GitHub repos.”
The Skeptic
Reality Check
“The reimbursement rates in these models are historically low enough that the unit economics only work if the AI handles volume at near-zero marginal cost — which is the optimistic case, not the base case, for anything running inference at clinical-grade reliability requirements. What kills this in 18 months isn't a competitor, it's the audit: the first time CMS reviews claims submitted under these codes and finds AI-generated contact logs that don't meet documentation standards, the billing pathway closes faster than it opened. The tech industry not knowing about this is actually the correct state of affairs until someone proves the compliance layer works at scale.”
The PM
Product Strategy
“The job-to-be-done for a provider organization adopting this isn't 'monitor patients better' — it's 'generate new reimbursable revenue from care coordination we're already doing inconsistently.' That framing completely changes what the product needs to do: the AI agent isn't the product, the billing workflow is, and the agent is the delivery mechanism. Any team building in this space that leads with the AI capability rather than the reimbursement outcome has the product strategy backwards.”