Welldoc launches AI care loop that keeps clinicians in control between visits

Welldoc said Wednesday it launched a next-generation digital-first care loop that connects a patient’s daily health data directly to clinicians through clinically validated closed Small Language Models and an AI care agent named Simon.

Patients with diabetes, obesity, and other complex conditions drown in disconnected apps and unverified chatbots while clinicians get none of the signal. Welldoc is betting closed, clinically governed small language models, plus a named care agent, can keep the human care team in charge between visits instead of outsourcing advice to a general chatbot.

On Wednesday, 30 September 2026, Welldoc announced its next-generation platform. The company calls itself a maker of AI-powered health technology for complex conditions. The platform is a digital-first care loop: a person’s day-to-day health data goes to that person’s care team, instead of sitting in a separate app. At the center is Simon, Welldoc’s AI care agent. An agent, here, is software that arranges support between office visits. The loop runs on Welldoc’s clinically validated closed small language models, which the release shortens to SLMs. A language model writes and answers in ordinary words. Closed means the model is Welldoc’s, not a general consumer chatbot. The company says that is how the guidance stays personal, measurable, and tied to a clinician. A clinician is the licensed person already on the care team. Welldoc’s own posting of the release datelines it Columbia, Maryland. Business Wire carries the release. Financial Content stamps the same item at 10:00 a.m. Eastern. Those lines are Welldoc’s.

Welldoc says health apps and sensors keep multiplying, while older programs still split care into pieces, lose users, and leave the return unclear. Return, here, means whether the program is worth what it costs. The company says people living with complex conditions, including obesity, diabetes, and early-stage dementia, are often swamped by apps that do not connect, chatbots nobody has checked, and data that never reaches a clinician. Dementia is a loss of memory and thinking that gets worse over time. Early-stage means the start of that course. Welldoc points that picture at a 15 November 2024 Guidea article on what Guidea calls point-solution fatigue. A point solution is one tool for one job. Those lines are Welldoc’s, including the citation.

Welldoc says the loop is meant to turn those daily touches into people who keep using the product, and into insights a care team can act on. The aims it prints are more room in a clinician’s day, and care matched to risk. Because the models are built for clinical use, Welldoc says an insight sits on the person’s real clinical picture, not a generic chat reply. Those lines are Welldoc’s.

Kevin McRaith, president and chief executive of Welldoc, said people need simple, connected solutions that bridge daily life and clinical care. He said the care loop “translates complex data into trusted, real-time guidance straight to clinicians,” so they can spot risk early, use resources wisely, and scale programs efficiently. Scale, here, means run the program for more people. That quotation is his, in the release.

Welldoc says real-world data and published studies show that putting AI into daily health management amplifies patient engagement and clinical outcomes. The figures it actually prints are about use, not a disease result. People using Welldoc’s AI features engage 20 percent or more often, and stay active up to four times as long as people who do not use those AI features. Twenty percent more is at least one extra active use for every five. Four times as long is the ceiling in the phrase “up to.” The release does not print a typical result for every person. Both figures are marked as Welldoc data on file for 2026. The same paragraph says Welldoc is still studying how that AI use changes health results, including weight. Those lines are Welldoc’s. The release does not print a weight number.

R “Ray” Wang, chief executive of Constellation Research, said AI is often used to speed a clinician’s paperwork, while patients want to use that power themselves. He said a continuous, closed care loop lets a person act between visits while the clinician stays in control, and that this is how AI can follow the rules and still personalize care for many people at once. That quotation is his, in the Welldoc release. Anand Iyer, Ph.D., chief AI officer at Welldoc, said healthcare built the pipes to share data and has not made enough of the insight in those pipes. He said Welldoc is moving from sharing data to producing clinical intelligence. That quotation is his, in the release.

The release says the loop sits on more than 500 million health data points, across more than 80 health domains, with more than 21 million real-time sensor records, including continuous glucose monitors, and 9.2 million AI-driven engagements. A continuous glucose monitor, which the release shortens to CGM, is a sensor that reads blood sugar through the day. A health domain, here, is one kind of health information, such as glucose or weight. An engagement is one time the AI side of the product interacted with a person, on Welldoc’s count. A key-facts line on the same release says the models were trained on 500 million health data points across 80 health domains. The body says “more than 500 million” and “80+.” Both wordings are Welldoc’s. The release marks the body figures as data on file for 2026. It does not say an outside auditor checked them.

Welldoc says a global AI team in the United States and in Bangalore, India, keeps changing those models, including through AI lab partnerships with Johns Hopkins University and the Vellore Institute of Technology. It says custom AI for customers in life sciences, health plans, and health systems can be in place in two to three months. It says the platform is built so an auditor can check it, and that it meets federal and industry standards including HIPAA and NIST, with continuous safety checks, automated quality audits, and protections for patient privacy. HIPAA is the U.S. health-privacy law. NIST is the National Institute of Standards and Technology, the federal lab whose security standards companies use as a checklist. Those lines are Welldoc’s. The release does not attach an audit report, and it does not name a customer for the two-to-three-month line.

The key facts say the work is backed by 11 FDA clearances, 69 patents, and more than 100 peer-reviewed clinical publications, including 44 focused on healthcare AI. Peer-reviewed means other scientists checked the paper before a journal ran it. The about box is more specific on the clearances: 11 clearances on the FDA’s 510(k) path, for diabetes functions inside the digital health platform. On that path, the FDA clears a device when the maker shows it is close enough to a device already allowed on the market. It is a device clearance. It is not a drug approval. The about box says the 69 patents cover Welldoc’s AI and technology. The same box says many of the papers look at pairing a continuous glucose monitor with AI tools for digital health. The key facts also say Welldoc’s research sets benchmarks for reading a continuous sensor, predicting where a person’s health is headed, and what the company calls safe, FDA-compliant architectures. The release does not name that benchmark or print a score. It does not announce a new FDA clearance on Wednesday for the care loop. The 11 is the company’s count of diabetes clearances on the platform it already sells.

The about box describes the company. It says the platform coaches people on prediabetes, diabetes, high blood pressure, heart failure, and weight, with support for mental well-being and sleep. It says Welldoc has moved past heart-and-metabolism care into brain health, cancer, and immunology. Immunology is the study of the immune system. It says the company works with health plans, health systems, and employers. It calls Welldoc an FDA-cleared tool that helps a person with diabetes manage day-to-day care and stay linked to the care team. The box also lists press mentions and awards, including Time and MedTech Breakthrough. Those lines describe the company. They are not a list of customers for Wednesday’s launch. The Welldoc homepage, the same day, shows the launch headline. A separate line on that page says the company analyzes 500 million-plus unique metrics across 80-plus health domains each year, per user. That per-user line is the homepage’s. It is not the release’s “more than 500 million health data points.”

In plain terms, Welldoc said on Wednesday that daily health data, including glucose-sensor readings, can feed Simon, an AI care agent on Welldoc’s own clinical language models, and that the care team still gets the signal between visits. The use figures, at least 20 percent more active engagement and up to four times longer retention, are Welldoc’s, from its 2026 files. The 11 FDA clearances are the company’s count of diabetes clearances on the existing platform. The release does not name a health-system customer, does not print a price, and does not print a weight result or any other clinical score for the new loop.

The picture is a cream desk graphic of that loop. Three cards run left to right, with an arrow between each. Patient day lists sensors and the app, a continuous glucose monitor and other vitals, and daily prompts. Simon lists closed small language models, risk signals, and the time between visits. The care team lists a clinician’s view, capacity alerts, and a guided next step. A line under the title says closed SLMs, the Simon care agent, and a clinician in the loop. A line under the cards reads 11 FDA clearances, 500 million-plus health data points, and HIPAA / NIST. A credit line names Welldoc via Business Wire and September 30, 2026. It is a diagram of the loop in the release. It is not a photograph of a clinic, and it is not a screenshot of Simon inside the app.