Dr. Karthik Ravindran's primary care practice in Framingham, Massachusetts, became something of a laboratory last year. While his patients came in for routine checkups and chronic disease management, software agents were quietly working alongside his staff—identifying diabetics overdue for blood work, flagging missing diagnosis codes worth thousands in Medicare risk payments, triaging after-hours messages.
Ravindran isn't just testing this technology. He's the chief medical officer of the company that built it.
Beacon Health, a New York startup that emerged from Y Combinator's Winter 2026 cohort, launched publicly on February 16 with what might be the healthcare industry's most casually radical pitch: autonomous AI agents that live inside electronic health records, doing the administrative grunt work that has made primary care nearly unsustainable. No dashboards. No new logins. Just software executing tasks where clinicians already spend their days.
The company claims it's already operational at an independent physician association serving roughly 40,000 patients. Whether that deployment is primarily within Ravindran's own network remains unclear—Beacon's materials describe him as both "first customer and CMO"—but the assertion that this isn't vaporware, that the agents are actually running, sets it apart in a market suddenly thick with "AI agent" announcements.
Four Workflows, One Question
What do these agents actually do? Beacon focuses on the unglamorous mechanics of value-based care contracts, where physician groups take on financial risk tied to quality metrics and documentation accuracy.
The agents identify care gaps: patients who've missed screenings, labs, or preventive visits that affect quality scores. They either automate outreach directly or queue up orders for staff review. They hunt for undocumented hierarchical condition categories—the diagnosis codes that determine risk adjustment payments in Medicare Advantage plans. Miss a patient's chronic kidney disease in the chart, and you're leaving money on the table. Document it properly, and the payment follows.
After business hours, Beacon's software triages incoming messages and follows up to reduce morning backlogs. It also tackles prior authorization requests and unspecified "front-office tasks," according to company descriptions.
The technical approach is what Beacon is betting on. Instead of asking overwhelmed practices to adopt another software layer, the agents operate inside existing EHR systems—executing clicks, populating fields, navigating screens the way a human administrator would. The company hasn't revealed which EHRs it supports or whether it's using API connections, FHIR standards, or robotic process automation under the hood. That opacity might signal competitive sensitivity, or it might suggest the integrations are narrower than the pitch implies.
The Founder's Credential: A Childhood Spent in a Clinic
CEO Mark Pothen has an unusual origin story for a tech founder. He grew up watching his mother run a primary care practice, absorbing the realities of patient panels and insurance billing from an early age. Before launching Beacon, he spent six months embedded in a clinic, observing workflows and pain points firsthand.
His co-founder, CTO Obinna Akahara, comes from a different trajectory: stints at Conversion.ai and LinkedIn after studying physics at the University of Texas at Austin. It's a pairing that mirrors many health tech startups—one founder steeped in clinical context, the other in engineering firepower.
Ravindran's dual role as both CMO and inaugural customer is common enough in early-stage healthcare software, where founder-friendly deployments accelerate iteration. It does raise the question of validation, though. How well does the product perform outside networks where founders have operational control? The answer will matter.
A $1 Trillion Opportunity in a Very Crowded Room
The market Beacon is entering has genuine urgency. The Association of American Medical Colleges projects a physician shortage of up to 86,000 by 2036. Primary care will bear the brunt of it, even as over 100 million Americans already lack consistent access to a family doctor, per Kaiser Family Foundation data.
Value-based care contracts, which shift financial risk from payers to providers, represent what Beacon frames as a $1 trillion opportunity. Independent practices are increasingly taking on these arrangements, where profitability hinges on closing quality gaps and capturing diagnosis codes accurately—exactly the workflows Beacon automates.
The competition is real and well-funded. Aledade, the largest network of independent primary care practices in value-based arrangements, serves more than 2,400 organizations and generated $801 million in Medicare Shared Savings Program returns in 2023. Aledade's model blends technology with hands-on services, a fuller-service approach than Beacon's software-only bet.
On the pure software side, Notable Health and Skypoint have both rolled out AI agents for value-based care workflows. Helpcare AI, another Y Combinator company, is focused on autonomous scheduling and revenue cycle agents. Oracle Health recently deployed what it calls a "Clinical AI Agent" for ambient documentation. The label "AI agent" is attached to everything now, and distinguishing real workflow automation from repackaged alerts will require scrutiny from buyers who've been burned before.
What Proof Looks Like
Beacon is actively courting independent physician associations, accountable care organizations, and primary care networks. The company's marketing materials suggest potential upside of "up to $250,000 per primary care physician" from optimized value-based contracts—a figure that originates from Beacon itself, not from independent validation.
What comes next will be about proof. Customer references outside the founding team's orbit. Measurable outcomes: quality scores lifted, hierarchical condition category capture rates improved, net financial impact documented. Early-stage healthcare software lives or dies on those metrics, and skepticism among clinicians toward AI promises is well-earned after years of overhyped automation tools.
For now, Beacon has cleared a meaningful first hurdle. It's live. It's processing workflows for tens of thousands of patients. And it's addressing tasks that practices genuinely need automated, not inventing problems to justify solutions.
Whether that translates to traction in a market suddenly crowded with agent-labeled competitors—well, that story is still being written. Perhaps more than Beacon's founders expected when they started building six months ago.
