HLTH 2026: CVS Health, Cigna and Abridge collaborating with OpenAI to test computer-use agents

AI agent concept
CVS Health said in one production workflow, an agent cleared a months-long document-sending backlog over a weekend. (BangonPitipong/GettyImages)

LAS VEGAS—OpenAI is expanding computer-use agents for healthcare, enabling AI to navigate electronic health record systems, payer portals and other software to tackle repetitive and manual tasks.

Computer-use agents, also called computer-using agents, are AI systems designed to interact with software interfaces the same way a human does—by viewing a screen, moving a cursor, clicking buttons and typing text.

Research teams at OpenAI have been exploring how computer-use agents can assist healthcare teams with fragmented clinical and administrative workflows while keeping people in control, Karan Singhal, head of Health AI and Artificial General Intelligence at OpenAI, told Fierce Healthcare.

"Around six months ago we set a research bet that we thought that the abilities for agents to use computers would end up being helpful and reliable enough in healthcare to help people with clinical and administrative workflows," Singhal said in an exclusive interview sharing evaluation results on OpenAI's early work with computer-use agents in healthcare.

Singhal also shed light on how OpenAI is working with CVS Health, Cigna Healthcare, Abridge and Ambience to apply computer-use agents to specific workflows, including referrals, prior authorizations, discharge planning, specialist consults and care manager preparation. 

Healthcare systems are often disconnected, forcing administrative staff and care teams to manually move information from one system to another before a patient can get care. If a clinician orders an MRI, for example, the insurer may need information from the patient’s medical record, but a staff member still has to pull that information and enter it into a separate portal to request authorization.

OpenAI executives and researchers framed computer use as the next step in healthcare AI, moving beyond copilots that draft or summarize information to agents that can actually navigate EHRs, payer portals and other applications on a user's behalf. 

Computer-use agents can bridge healthcare's fragmented software environment, especially where application programming interfaces (APIs) and other integrations don't exist or don't connect an entire workflow, Singhal noted.

"We have been working with health systems on things like ChatGPT for Healthcare and our integration with Epic, but there are still limitations in how useful AI agents can be today. Computer use is really a way of resolving a lot of these barriers that exist between AI capabilities and healthcare software so that care teams can deliver better care, more effectively and more efficiently," he said.

OpenAI continues to invest in APIs, plugins and MCP connectors to integrate AI with healthcare software, but the industry’s fragmented and legacy technology can make those integrations slow and incomplete, Singhal said. Computer use can help bridge those gaps by allowing AI to work across existing software interfaces and take on manual, repetitive workflows.

"What we hope to do with this release is move us closer to a world where, because AI is connected to more services, AI can help more with a lot of these workflows," he noted.

Care teams, developers and individual clinicians can access computer use for healthcare through ChatGPT for Healthcare, ChatGPT for Clinicians and the OpenAI API.

In a research demonstration using synthetic patient data, OpenAI showed how a computer-use agent could support a clinician and an authorization specialist working on the same MRI request. In the demonstration video, a patient messages their clinician about lower back pain. The clinician then instructs the computer-use agent to review the patient’s visit notes, physical therapy records and prior imaging in the EHR. The computer-use agent drafts an MRI order, an encounter note and a reply to the patient for the clinician to review. 

In the same demo video, the agent then uses the same records to prepare a prior authorization request in the insurer’s portal for an authorization specialist to review.

"The clinician and authorization specialist were both completely in control of their workflow, but the computer-use agent was able to help across different parts of the workflow and translate that into action," Singhal noted about the demo video.

OpenAI evaluated GPT‑6 Astra on tasks clinicians handle in their EHR inboxes. These tasks, shaped by physician input, require Astra to carry out multi-step workflows designed to test both clinical reasoning over longitudinal patient context and navigation of complex interfaces. Reviewers examined the model’s actions alongside the notes, orders, and patient messages it produced, assessing whether the work would be helpful and clinically safe in the context of the task, according to an OpenAI blog post.

OpenAI says its GPT-6 Astra model can perform multi-step EHR tasks such as reviewing patient records, drafting messages and notes, preparing medication refills and proposing orders, while leaving key actions for clinician review. In OpenAI's evaluation, 99% of 325 reviewed tasks were rated clinically safe, while physicians rated the model "very helpful" or "extremely helpful" in 93% of 646 assessments.

Clinical safety was assessed by evaluating whether inaccuracies, omissions or actions in the model’s work could plausibly lead to patient harm, OpenAI said.

Regarding the 1% of tasks rated not clinically safe, Singal said researchers did not identify any cases 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.

"The models are not perfect today, and what we're sharing is an update on research results where we're seeing this critical point where the models seem to be useful in practice," he added.

OpenAI researchers also evaluated computer-use agents for administrative tasks, such as prior authorizations, appeals and medical equipment orders.

OpenAI designed its computer-use agents with robust safeguards and data security in place, according to the company. The agents can be configured with limits on system access and actions, and audit logs provide a clear record of agent actions and approvals, giving healthcare teams visibility into how each workflow was completed.

Healthcare organizations putting computer-use agents to work

Organizations are developing, piloting and deploying computer-use applications across a range of workflows and administrative tasks.

CVS Health is using OpenAI’s computer-use capabilities to automate repetitive work across clinical systems, including sending documents, processing referrals and supporting specialist consults. In one production workflow, an agent cleared a months-long document-sending backlog over a weekend, OpenAI detailed in the blog post. Another agent handles the system steps for tens of thousands of consults each day.

Generative AI company Abridge is building computer-use agents to help clinicians carry out care plans across use cases and applications they use daily, including those without APIs. The company also plans to develop an inpatient use case, adding EHR information and nurse and clinician context to identify what needs attention and to coordinate next steps. As an example, agents can work concurrently on medication access and home health coordination, using computer interaction alongside calls and messages. 

Cigna Healthcare is using OpenAI’s models to pull together member information from more than 10 clinical and administrative systems, including benefits, providers, appointments and previous interactions. The goal is to reduce the time case managers spend searching across systems and give them more time to support members, OpenAI wrote in the blog post.

Companies including Assort, Bunkerhill, Forus, Gravity and Tennr are testing agents that interact directly with EHRs and payer portals for referral management, authorization checks, medication approvals and other operational tasks.

“We're continuing to invest in the models, making the models better at computer use. An important way that we think that they could continue to improve is around improved latency and costs,” Singhal noted.

OpenAI sees its role in healthcare as building foundational AI capabilities and letting healthcare companies handle the “last mile” of implementation, Singhal noted. Computer use is one example: OpenAI provides the technology and safeguards, while partners apply it to specialized workflows and deploy it within healthcare organizations.

The company also announced today improvements to trusted clinical search in ChatGPT for Healthcare and ChatGPT for Clinicians, including how ChatGPT finds, evaluates and cites medical evidence. OpenAI also expanded its healthcare plugin ecosystem with more than 20 new healthcare plugins from companies across clinical care, care coordination and business operations.