← AI Switchboard
AGENTS · November 17, 2026
Nov 16

OpenAI puts computer-use agents into healthcare, with CVS Health and Cigna already using them, and physicians rated 99% of the agents' actions safe in 325 reviewed tasks, two trade outlets report

By the AI Switchboard · researched and written with AI from the sources listed on this page, each point checked against its source before publishing · how we work · report a mistake

The agents click and type through health records and insurer portals that have no API. OpenAI's own page would not open here, so its figures come from HIT Consultant and Interesting Engineering.

OpenAI on 16 November added computer-use agents to ChatGPT for Healthcare, HIT Consultant and Interesting Engineering both report. The agents work through software hospital and insurer staff already use, such as electronic health records and payer portals, by clicking, typing and moving between applications, so they need no special connection to each system. Both outlets say they run on GPT-6 Astra and GPT-6 Luna.

Both outlets report the same evaluation figures from OpenAI. Physicians reviewed 325 tasks drawn from health-record inboxes and rated 99% of the agents' actions clinically safe, and rated the output very or extremely helpful in 93% of 646 assessments. On HealthAdminBench, a Stanford benchmark of prior authorisations, claim appeals and equipment orders, Astra scored 94.3% of the available points and Luna 91.8%. The two outlets disagree on which model costs 8.5 cents a task, so that figure is left out.

The named users: CVS Health engineers used the agents for sending documents and processing referrals, and both outlets report that one agent cleared a months-long document backlog in a single weekend. Cigna uses them so case managers can gather member information across systems. Abridge is building discharge agents.

What has changed since the Switchboard's last story on OpenAI's agents: on 13 November we ran ChatGPT Wallet, which lets agents pay for things for a consumer. Three days later the same kind of agent is working inside clinical and insurance systems for two of the largest US health companies.

  • Reported Physicians rated 99% of the agents' actions clinically safe across 325 reviewed tasks drawn from health-record inboxes, and the output very or extremely helpful in 93% of 646 assessments, HIT Consultant and Interesting Engineering report. HIT Consultant
  • Reported On HealthAdminBench, GPT-6 Astra scored 94.3% of points and GPT-6 Luna 91.8%, Interesting Engineering reports. Interesting Engineering
  • Reported A CVS Health agent cleared a months-long document backlog over a single weekend, HIT Consultant and Interesting Engineering report. HIT Consultant

Agents in the wildModels & releases

Today in the November 17, 2026 edition · front page