Skip to main content

OpenAI details computer-use agents for healthcare workflows

16 NOVEMBER 2026·2 MIN READ·1 SOURCE

OpenAI says its computer-use agents can carry out multi-step workflows across applications, and that healthcare teams, developers and individual clinicians can access the capabilities through its API and healthcare products. The company also described an evaluation of EHR inbox tasks and work with healthcare organizations on applications including referrals and authorization requests.

OpenAI details computer-use agents for healthcare workflows

Key takeaways · 4

  • 01

    Teams can access the capabilities through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians.

  • 02

    OpenAI’s MRI demonstration used synthetic patient data, and the generated order, note and patient reply were drafts for clinician review.

  • 03

    Reviewers rated 99% of 325 assessed EHR inbox tasks safe; physicians rated the work very or extremely helpful in 93% of 646 assessments.

  • 04

    CVS Health says it uses validation checks and human oversight, with screenshots and action records to verify agent actions and application state.

Access and deployment controls

OpenAI says healthcare teams, developers and individual clinicians can use the capabilities through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians.[1] The agents can click, type and move between applications to complete multi-step workflows, according to OpenAI.[1] For secure deployments, organizations can configure approvals and access, and record audit logs.[1] These controls make approval design and auditability part of deployment planning: teams considering a workflow should decide which actions need human approval and what activity they need to record.

What the MRI demonstration showed

OpenAI’s example used synthetic patient data.[1] The agent reviewed visit notes, physical therapy records and prior imaging in an electronic health record before drafting an MRI order, an encounter note and a reply to the patient for clinician review.[1] It also prepared a prior authorization request in the insurer’s portal for an authorization specialist to review.[1] The demonstration illustrates a workflow that crosses clinical records and an insurer portal, but the outputs described were prepared for people to review; it does not establish that the agent independently made clinical decisions or submitted final documents.

Evaluation results and cost comparison

OpenAI evaluated GPT-6 Astra on tasks clinicians handle in their EHR inboxes.[1] Reviewers rated 99% of 325 assessed tasks safe, and physicians rated the work very or extremely helpful in 93% of 646 assessments.[1] GPT-6 Astra earned 94.3% of available rubric points.[1] GPT-6 Luna earned 91.8% and cost 8.5 cents per task, which OpenAI says was 98% less.[1] OpenAI says both models ran at maximum reasoning effort in this administrative evaluation.[1] The reported figures describe this evaluation; they do not establish performance across every clinical workflow.

Partners and production workflows

OpenAI says it is working with Abridge, Ambience, Cigna Healthcare and CVS Health on applications spanning authorization requests, referrals, discharge planning and outreach between visits.[1] CVS Health says its computer-use capabilities automate repetitive work across clinical systems, including sending documents, processing referrals and supporting specialist consults.[1] CVS reports that one agent cleared a months-long document-sending backlog over a weekend, while another handles system steps for tens of thousands of consults each day.[1] The company says it verifies agent actions and application state with validation checks, human oversight, screenshots and action records.[1] CVS is open-sourcing its Agent Kit, built with OpenAI’s Agents SDK.[1]

Computer-use agents are being positioned to handle administrative steps across clinical systems, not just work inside a single interface. For professionals evaluating them, the practical questions are which tasks are suitable for delegation, where human review belongs, and how actions will be verified.

Why it matters
For teams

AI training for your team

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

How this developed

  1. 16 November 2026

    OpenAI details computer-use agents for healthcare workflows

Sources

AI fluency, one session a day, built for your work.