CVS and Cigna use OpenAI agents across clinical systems
OpenAI said CVS Health agents handle the system steps for tens of thousands of consults a day. Cigna Healthcare uses them to pull member records from more than 10 systems.

Cigna Healthcare case managers had to search more than 10 separate clinical and administrative systems to prepare for a member call, OpenAI said on Nov. 16. OpenAI said its computer-use agents can now bring that information together in seconds. The records include benefits and coverage, providers, appointments and prior interactions.
Computer-use agents work inside existing software. They read the screen, click, type and move between applications, so they can finish steps where an API does not exist. OpenAI said it is working with CVS Health, Cigna Healthcare, Abridge and Ambience on prior authorization requests, referrals, discharge planning and outreach between visits. The company announced the work at HLTH 2026 in Las Vegas, Fierce Healthcare reported.
What CVS is running
CVS Health uses the agents to send documents, process referrals and support specialist consults, according to OpenAI. One production agent cleared a document backlog that had built up over months in a single weekend. Another handles the system steps for tens of thousands of consults each day.
CVS Health checks the agents’ work with validation checks and human review, using screenshots and action records. The data stays on CVS Health’s own infrastructure, and production data is not used to train foundation models, OpenAI said. CVS Health is releasing its Agent Kit, built with OpenAI’s Agents SDK, as open source.
“Less time on administrative work means more time for care. That’s a win for our teams and everyone we serve,” said Lagan Khare, executive director of engineering at CVS Health.
What Cigna built
Cigna Healthcare uses OpenAI models to gather member information for care managers before outreach. Madhu Nutakki, chief information officer of Cigna Healthcare, said the project moved “from concept to deployment in under eight weeks.”
The controls
Organizations can set which applications an agent can open, require approvals and keep audit logs, OpenAI said. Work that involves protected health information has to run on services covered by a Business Associate Agreement with OpenAI. OpenAI said organization data is not used to train its foundation models.
Ambience showed an agent preparing a prior authorization request on an insurer’s website from evidence in the patient’s chart. When the record is unclear, the agent stops, asks the clinician and waits for approval before it submits. Health systems choose which actions the agent may take, can pause it at any point and get a step by step log.
Abridge is building agents to help clinicians clear discharge barriers, such as medication access and home health coordination. OpenAI described that inpatient use as planned.
What OpenAI tested
OpenAI tested its GPT-6 Astra model on 325 reviewed tasks that clinicians handle in their EHR inboxes and published its own safety ratings. Reviewers checked whether errors, omissions or actions in the model’s work could plausibly harm a patient.
Karan Singhal, who leads health AI at OpenAI, told Fierce Healthcare that for the tasks not rated safe, researchers found no case where the model created a specific, plausible risk of patient harm. “In some cases, the model might overreach and might take an action without getting the proper sign-off or authorization,” he said. He added that the models “are not perfect today.”
OpenAI also showed a demo built on synthetic patient data. An agent reviews a patient’s visit notes, physical therapy records and prior imaging. It drafts an MRI order, an encounter note and a reply to the patient for the clinician to review. It then fills out a prior authorization request in the insurer’s portal for an authorization specialist to check.
The agents are available now through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians, OpenAI said.