Introducing Oswell: a health AI agent for Oscar members | hioscar Blog | Oscar
Introducing Oswell: a health AI agent for Oscar members
With Oswell, we’re building a personalized AI frontdoor to healthcare.
Behind the ScenesNews from Oscar
4 min read
Share article
With Oswell, we’re building a personalized AI frontdoor to healthcare.
Disclaimer: Oswell provides general health and benefits information only. It is not a doctor and does not provide medical advice, diagnosis, or treatment. The chatbot’s answers may be inaccurate or incomplete, and are not reviewed by a clinician. In an emergency, call 911. See our Terms of Service for more details.
Healthcare decisions come with questions at every turn: Do I need a doctor? Who should I see? What will it cost? What happens next? Each unanswered question adds friction to an experience that should instead feel simple and connected.
At Oscar, we believe technology should make healthcare simple, not complicated. This is why we set out to design something that could guide members across the full care experience, before, during, and after, while making Oscar the trusted starting point for every healthcare interaction.
The result is Oswell, an OpenAI-powered health and wellness agent built to make healthcare easier, more connected, and more personal.
A connected guide to care
Oswell gives members a single, reliable place to ask health and wellness questions. It answers common concerns, analyzes symptoms, prepares members for visits, and explains follow-up instructions afterward – on demand, 24 hours a day, seven days a week.
Unlike general-purpose AI tools, Oswell connects directly with Oscar’s systems and data. It draws from claims and medical records, Care Guide interactions, and virtual care data to personalize every response. When needed, it hands off to human teams or routes members to other Oscar services while carrying chat history forward so the experience never feels fragmented.
“Can you help me refill my diabetes med? I forgot the name of it.”
“My doctor said something about high cholesterol. Explain the lab results she gave me and what I should do.”
“My hip hurts. Is it just my arthritis? Or which doctor can I see asap and how much would it cost?”
Oswell is not a “robodoc.” We know that providers play a critical role, so we’ve designed Oswell to assist, not replace clinicians while empowering members to take even more ownership of their healthcare journey.
Oswell is here, and we’re just getting started. Today, a select group of Oscar members can try it out, and soon every member 18+ will get access in the months ahead.
Meeting member needs, powered by OpenAI
When we recently asked members about their AI tool usage, more than half told us they use AI tools like ChatGPT, but most do not turn to them for healthcare. The top reason, according to members, is that the information does not feel personal enough. Members want responses that reflect their specific needs, benefits, and data, all delivered within a secure environment they trust.
That is why Oswell is built directly within the Oscar platform, allowing it to offer personalization alongside privacy and accuracy. It is powered by new advancements in conversational AI, AgentSDK integrations, and significant improvements in model capability. These breakthroughs enable Oswell to give useful information in response to complex healthcare questions, know when to escalate care, and provide clear, human-like guidance in real time.
Building Oswell safely
Because Oswell operates in the deeply personal space of health, safety and trust are at the center of its development. Oswell is designed to ensure that members receive accurate, responsible, and secure guidance. The system has been tested, audited, and validated to meet rigorous clinical and ethical standards before launch.
Rigorous evaluation and testing: Oswell’s safety starts with a physician-audited test dataset built to expose common and high-stakes LLM healthcare errors, such as emergency escalation, suicidality, and red-flag symptom detection. Dozens of Oscar employees have tested the system internally, and a panel of physicians rigorously examined and challenged Oswell to ensure accuracy, reliability, and appropriate escalation.
Built-in guardrails and monitoring: Oswell is explicitly configured not to provide definitive diagnoses, promote self-harm, or offer unsafe or illegal advice. It can escalate to virtual, in-person, urgent, or emergency care when necessary, and an automated “LLM-as-a-judge” system continuously reviews a sample of conversations to flag and escalate concerning cases. A physician panel also audits random samples to maintain and enhance quality and safety over time.
Data protection and privacy: Oswell is available to members when they are logged into the Oscar member app. Oswell can only access information linked to the member who is logged in. It has no access to other members’ data.
Clear consent and disclaimers: By ensuring members understand both Oswell’s capabilities and its limits, we’re ensuring safety and building trust.
Just getting started
Oswell’s launch is just the beginning. We’re already exploring and developing further functionality such as allowing members to chat directly about medical charts and lab results, book appointments, and receive guided question suggestions before visits. Future updates will expand into insurance operations such as ID card requests and bill payments.
We are also building toward richer, more natural interactions with voice capabilities and media uploads, allowing members to describe or show symptoms directly. Beyond that, Oswell will support nutrition, social, and chronic condition management, while enabling real-time collaboration between care guides and virtual providers within the same chat thread.
Each of these milestones brings us closer to Oscar’s vision of a more seamless, connected, personalized healthcare experience for our members.
