UpDoc, a startup led by a Stanford primary care physician, has received FDA clearance for what it calls the first agentic clinical AI platform designed to support doctors. The platform, now being deployed at Stanford Health Care, Cleveland Clinic, Allegheny Health Network, and UCSF Health, automates chronic disease management tasks between patient visits under physician supervision. The company also announced an $18 million oversubscribed seed round from investors including the American Diabetes Association and Mayo Clinic.
What the platform does
UpDoc's system delivers what CEO Sharif Vakili, M.D., M.B.A., M.S., terms remote patient intervention (RPI). It acts on remote patient monitoring data to adjust medications, order follow-up tests, and communicate with patients, all within boundaries set by the physician. "What has historically taken place in an encounter with me is now able to happen at the patient's home at the cadence of when it's clinically indicated, not at the cadence of the appointment with me," Vakili said.
The AI agent operates only after the doctor pre-authorizes specific actions for each patient. Vakili described it as a clinical AI that does "something that would have historically taken place in an encounter with the licensed clinician." He emphasized that the physician remains involved, neatly bounding what the agent is allowed to do so there are "no surprises."
How it changes diabetes care
Vakili offered a concrete example: a patient with diabetes who needs insulin. In the traditional model, the doctor starts insulin, asks the patient to check blood sugars at home, and waits three months for the next appointment to review numbers and adjust the dose. This cycle can stretch more than a year before the patient's condition is controlled.
With RPI, the physician prescribes a treatment plan, and the clinical AI implements it. The agent contacts the patient, reviews continuous glucose monitor data, and makes medication adjustments as permitted. "By the time I next see that patient in clinic, in my next available slot in three months, the patient may already have had those meds dose-adjusted and all the follow-up tests collected," Vakili said.
Evidence and early results
A randomized controlled trial conducted six years ago by the co-founders at Stanford, before UpDoc was formed, tested the care delivery model on patients with uncontrolled diabetes. After two months, 81% of patients in the RPI arm achieved glycemic control, compared with 25% receiving standard care. The RPI group also had five times more prescription adjustments. "If you get five times more care delivery, you're going to end up with better outcomes," Vakili said.
The platform is now being integrated into electronic health record systems at the partner health systems. Vakili said about 90% of the EHR integration is consistent across sites, with the remaining 10% tailored to each practice environment. The company has worked closely with health systems on governance and oversight frameworks, with physicians double-checking the system's actions during the initial rollout.
Why this matters for healthcare
Chronic disease management is labor-intensive and demands frequent follow-up that overburdened primary care teams often cannot provide. UpDoc's approach shows that clinical AI, when deployed with physician oversight, can compress the time to treatment adjustments and improve outcomes without adding to clinician workload. For health systems, the model offers a path to scale chronic care delivery without hiring more staff-a critical need as the primary care workforce shrinks. The platform's early adoption by major institutions signals a growing willingness to let AI agents perform tasks that once required a licensed clinician's direct involvement, provided safety boundaries are clear and the physician remains in control.
Your membership also unlocks:
