Clinical AI: Pittsburgh pilot will take on challenges of diabetes care. Kris B. Mamula Pittsburgh Post-Gazette Jul 21, 2026 In medicine, the uses for artificial intelligence have pretty much been limited to back office stuff - notetaking and appeals to health insurers to cover special medicines and treatments, those sorts of things. That could be ready to change in Western Pennsylvania as the role of AI accelerates further into direct patient care in a pilot planned at Allegheny Health Network. The 16-hospital Highmark affiliated health care system has inked a deal with Palo Alto, Calif.-based UpDoc Inc. to aid in caring for people with chronic diseases by using AI to monitor and adjust medication, as needed, between office visits. If the results of a small clinical trial of the technology are any indication, remote patient intervention may be poised to transform diabetes care. "It's adding a cook to the care team, it's opening the door and allowing the AI agent to join the team," primary care physician and UpDoc CEO Sharif Vakili said. "It's adding something important between doctor's visits; it's doing things in a new way." Here's how it'll be rolled out by primary care physicians at AHN. Doctors will select patients with type 2 diabetes for regular telephone or text contact with UpDoc's conversational AI platform. An AI agent will periodically check on the patient's symptoms, insulin and glucose, and recommend medication adjustments to optimize care. For patients with unruly bloodstream glucose - and they are legion - this often means multiple physician appointments, burdening medical providers and adding to physician burnout at a time when there's not enough primary care doctors, including in parts of Western Pennsylvania, experts say. For patients in the trial, their doctor will come up with a preset range of acceptable glucose levels. Failure to stay within those limits, despite medication tweaks, will prompt a recommendation for the patient to reach out to the treating doctor. All interactions between the AI agent and patient will become part of the patient's electronic medical record for physician review. Clinical applications for AI have been slow to get traction, said Holly Wiberg, assistant professor of Operations Research and Public Policy at Carnegie Mellon University's Heinz College. But that may be changing. "The clinical AI promise has really yet to be realized," she said. "A patient-facing interactive tool is really an exciting kind of frontier. You can reach more patients at a lower cost." Nearly one-quarter of the 33 million U.S. adults with type 2 diabetes in the U.S. have poor glycemic control because they often receive suboptimal doses of insulin, according to a 2023 JAMA Open Network study that was conducted by Vakili, UpDoc chief technology officer and internist Ashwin Nayak and others. "Therapeutic inertia," physician time constraints and competing demands due to a crush of medical practice appointments were among the reasons cited for inadequate insulin dosing. The current standard of care for diabetics with poor glucose control is simply to call the doctor for an appointment, which can delay care. But in the small clinical pilot conducted at Stanford Medicine, people with diabetes using an AI platform achieved effective insulin dosing nearly four times faster than those receiving standard care - median 15 days versus more than 56 days. Medical adherence - a notoriously difficult challenge in caring for people with diabetes - was also better with AI intervention: 83% compliance versus 50% with standard treatment, according to the study, which had 32 participants. Results of the trial were published in the American Medical Association's JAMA Network Open in 2023, the year UpDoc was founded. Getting glucose levels right reduces long-term complications from diabetes, including kidney disease, eye and nerve damage and heart disease, which ultimately holds down treatment costs. There are between 170,000 and 220,000 people with diabetes in the eight-county Pittsburgh metropolitan area, according to extrapolated projections of prevalence by the Centers for Disease Control and Prevention, and direct medical costs of treating the disease range up to $2.6 billion annually. In a prepared statement, AHN said it was anticipating use of UpDoc's digital platform. "Allegheny Health Network is looking forward to introducing this innovative AI platform with a particular emphasis on empowering primary care clinicians," said Amy Crawford-Faucher, chair of AHN Primary Care Institute. "The future integration of UpDoc within our electronic health records will further strengthen care coordination, reduce administrative burden and enable our physicians to devote more of their time to high-value, clinical decision-making - ultimately driving improved long-term health outcomes for our patients and improved population health." Crawford-Faucher was unavailable for further comment. Home monitoring of chronic diseases has been something of a holy grail in medicine, said Derek C. Angus, former chair of the Department of Critical Care Medicine at the University of Pittsburgh and UPMC, who is not affiliated with UpDoc. With infinite patience and no social or cultural barriers to undermine patient care, AI could revolutionize disease management. "The AI revolution in health care is both wonderful and terrifying," he said. "These AI tools are potentially fantastic and they may finally crack the knot of home-based remote care. "On so many levels it is so attractive to think about these tools, but there still can be side effects. They could end up making bad recommendations, but we don't know," he said. "The upside is amazing and the downside is real as well," Angus said. UpDoc has secured $16 million in oversubscribed seed financing from leading health care investors and strategic partners, including the American Diabetes Association, Cathay Innovation, Eli Lilly and Co., Mayo Clinic and others. UpDoc's Vakili said the company planned to roll out other chronic disease management uses for the platform in the coming months. First Published: July 21, 2026, 1:00 a.m. Kris B. Mamula has been a health care business reporter for the Post-Gazette since 2015. A Carnegie Mellon University graduate, he was part of the team that won a Pulitzer Prize for coverage of the synagogue shooting in 2018. Join the Conversation If you would like your comment to be considered for a published letter to the editor, please send it to [email protected]. Letters must be under 250 words and may be edited for length and clarity. Allegheny Health Network LATEST business Create a free PG account. Already have an account?

UpDoc, a clinical AI startup, has secured FDA clearance for the first Software as a Medical Device using patient-facing large language models. The company has raised $18 million in oversubscribed seed funding from investors including Mayo Clinic, Eli Lilly and Polaris Partners. The platform enables clinicians to deploy AI agents that automate clinical tasks such as adjusting medications, facilitating lab orders and coordinating care between scheduled visits. Unlike administrative AI tools, UpDoc integrates directly with electronic health records to deliver care requiring licensed clinician involvement. Founded by Stanford Medicine researchers, UpDoc is being deployed at Cleveland Clinic, Allegheny Health Network and UCSF Health. The platform uses proprietary orchestration architecture combining clinical guidelines, patient data and physician governance to manage chronic disease at scale whilst maintaining clinician oversight.