OpenAI Brings Computer-Use Agents to ChatGPT for Healthcare, With 20+ New Plugins

AI agents in healthcare illustration

OpenAI Brings Computer-Use Agents to ChatGPT for Healthcare, With 20+ New Plugins

OpenAI announced on November 16, 2026 that ChatGPT for Healthcare is gaining computer-use agents that can operate the software health teams already use, along with improved clinical search and more than 20 third-party plugins. The update targets the administrative work between a clinical decision and the care a patient receives: referrals, prior authorizations, document transfers, and discharge planning.

Background: the admin work between systems

Healthcare is fragmented, and teams often need to move information between systems by hand before a patient can get care. When a clinician recommends an MRI, for example, the information an insurer needs may already be in the patient’s medical record, but a team member still has to retrieve it and type it into a separate portal to complete an authorization request. OpenAI’s computer-use agents are built for that gap: they carry out multi-step workflows by reading screens, clicking, typing, and moving between applications, including software that has no API for the task at hand.

Key details: what was announced

Computer-use agents for clinical and admin workflows. OpenAI published new evaluation results for its GPT-6 Astra and GPT-6 Luna models on healthcare tasks. On physician-shaped EHR inbox tasks, reviewers rated 99 percent of 325 GPT-6 Astra tasks safe, judging whether errors or actions could plausibly harm a patient. Physicians rated the work “very helpful” or “extremely helpful” in 93 percent of 646 assessments.

On administrative work such as prior authorizations, appeals, and medical equipment orders, OpenAI tested the models on HealthAdminBench, an independent benchmark from Stanford researchers. GPT-6 Astra earned 94.3 percent of available rubric points, while GPT-6 Luna earned 91.8 percent despite costing 98 percent less, at 8.5 cents per task. Both models were evaluated at maximum reasoning effort.

Real deployments at large health organizations. OpenAI named four organizations already working with the technology. CVS Health is using computer-use agents to automate repetitive work across clinical systems, including sending documents and processing referrals. In one production workflow, an agent cleared a months-long document-sending backlog over a weekend, and another handles the system steps for tens of thousands of consults each day. CVS Health is open sourcing its Agent Kit, built with the OpenAI Agents SDK.

Cigna Healthcare is using the models to assemble member context that case managers previously had to gather from more than 10 separate clinical and administrative systems. Abridge is developing agents that help clinicians carry out care plans across applications that lack APIs, including help resolving discharge barriers. Ambience has built controls around the capability: in its demonstration, an agent preparing a prior-authorization request stops when the record is ambiguous and waits for approval before submitting, while health systems can pause the agent and get a step-by-step audit log.

More than 20 new healthcare plugins. The plugin ecosystem now includes Aiva, Amion, Assort Health, Availity Extend, Elation Health, Ferry, IMO Health MCP Server, Infinitus Relay, K Health, Luma Health, Metriport, Notable, Oracle HDI, PenguinAI Medical Coding, Photon, Redox, Stedi, Subsalt, TriZetto, Turquoise, and UpToDate Expert AI. They are available today in ChatGPT for Healthcare, including through ChartGPT, which brings ChatGPT into the EHR workflow. Workspace administrators must enable the new plugins. OpenAI gives an example of chaining them: Metriport can retrieve an outside discharge summary while UpToDate Expert AI supplies follow-up monitoring.

Improved clinical search. OpenAI also upgraded trusted clinical search in ChatGPT for Healthcare and ChatGPT for Clinicians, improving how the system finds, evaluates, and cites medical evidence with inspectable citations. In a blind evaluation of more than 1,300 physician reviews of answers to real-world clinical questions across multiple specialties, physicians rated the updated experience more highly across all eight dimensions OpenAI measured, including factual correctness, safety, actionability, and source authoritativeness.

Privacy and availability. ChatGPT for Healthcare provides role-based access controls and audit logs. Organizations must use eligible services covered by a Business Associate Agreement with OpenAI, with the required security and retention settings, to support HIPAA-compliant workflows involving protected health information. OpenAI says content shared with ChatGPT for Healthcare is not used to train its models. ChatGPT for Clinicians is free for eligible, verified clinicians in the U.S., and select plugins are available there as well.

Why it matters

This is a workflow release rather than a new model, but it is a significant one. The most expensive part of healthcare AI is rarely the model; it is the last mile of getting work done inside software that was never built to be automated. By pairing screen-level agents with published safety and cost numbers, OpenAI is making a case that agents can take on high-volume administrative work, with the cheaper Luna model nearly matching Astra on admin benchmarks at a fraction of the cost.

The guardrails are the other half of the story. Every example OpenAI published keeps a human in the loop for clinical decisions, and the controls from partners like Ambience, pausing, approval steps, and audit logs, show what deployment actually requires. For health teams, the practical next step is narrow: pick one handoff, such as an authorization packet or a referral close-out, define which applications the agent may open, and decide which steps stay with a person.

Sources

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *