Medicare's ACCESS Program: Payment Model Built for AI Agents, Details Inside

Medicare's ACCESS program (Advancing Chronic Care with Effective, Scalable Solutions) is a 10-year CMS pilot that pays for health outcomes, not activities. Traditional Medicare reimburses based on clinician time — no mechanism to pay for an AI agent that monitors patients between visits. ACCESS creates that mechanism. The first cohort of 150 participants goes live July 5, 2026.
How the Payment Model Works
Participating organizations receive predictable payments for managing qualifying conditions (diabetes, hypertension, chronic kidney disease, obesity, depression, anxiety). They earn the full amount only when patients meet measurable health goals: lower blood pressure, reduced pain, etc. This is a shift from fee-for-service to value-based care.
Pair Team's Implementation
Pair Team (raised ~$30M from Kleiner Perkins, ~850 clinical staff) deploys a voice AI agent called Flora as its primary patient interface. Flora handles intake, referrals, and check-ins. A peer-reviewed study in the Journal of General Internal Medicine showed Pair Team's model reduces avoidable ER visits by 50% and inpatient visits by 25%. The company claims one in four hospital visits and one in two ER visits don't happen when a patient is under their care.
Real Patient Interaction
A 67-year-old woman living out of her car, managing PTSD and congestive heart failure, spoke with Flora for over an hour. Hourlong conversations are now routine. Batlivala: "That's the companionship piece. And it turns out that is truly an intervention."
Who It's For
Developers building AI agents for regulated industries (healthcare, insurance) or working on voice-based patient engagement tools. The program's designers (Abe Sutton, Jacob Shiff) are former startup operators.
📖 Read the full source: HN AI Agents
👀 See Also

The AI Dependency Trap: Why Over-Reliance on LLMs May Erode Core Skills
A contrarian take arguing that heavy reliance on AI chatbots will lead to atrophy of critical thinking, writing, research, and learning abilities.

1.2B Local Model Beats 1T Clouds in Poker: Aggression Trumps Knowledge in Shove-or-Fold Format
A 1.2B Liquid model won 2 of 5 Texas Hold'em tournaments against models up to 1T parameters, because in a short-stack format, never folding earned more chips than smart play.

Claude Shannon's 1950 Chess Paper Predicted GenAI's Core Problem: Guessing vs. Knowing
Shannon's 1950 chess paper framed the core challenge of AI: making 'tolerably good' decisions under uncertainty—exactly the problem generative AI faces today when it produces polished but wrong answers.

BMW Dealership Revokes Buyback Offer After AI Chatbot Mistake, Precedent from Air Canada Case
A Toronto BMW dealership revoked a buyback offer generated by its AI chatbot, sparking legal questions. The Air Canada precedent holds companies liable for chatbot errors.