The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard
Minor World

OpenAI brings computer-use agents to ChatGPT for Healthcare with 20-plus plugins

By aminah ·17 Nov 2026 ·5 min read
A clinician in a white coat using a tablet showing an electronic health record in a hospital
OpenAI has expanded ChatGPT for Healthcare with computer-use agents that can work inside the software health teams already use, alongside improved clinical search and more than 20 new plugins. The company announced the changes on 16 November 2026, presenting them as a move from answering questions to getting administrative work done. A computer-use agent works much as a person at a keyboard does: it reads the screen, clicks, types and moves between applications. In healthcare, that means an agent can draft an order in an electronic health record (EHR), gather notes and prior imaging, and then prepare a prior authorisation request in an insurer's portal, rather than leaving staff to retype the same facts from one system into another. OpenAI says the agents are available through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians, powered by its GPT-6 Astra model, with a cheaper GPT-6 Luna model for high-volume administrative work. ## What the agents will actually do The focus is paperwork, not diagnosis. OpenAI names referrals, prior authorisations, discharge planning and outreach between visits as the first workflows, and says it is working with Abridge, Ambience, Cigna Healthcare and CVS Health on them. In a demonstration built on synthetic patient data, a patient message about lower back pain led the agent to review visit notes, physiotherapy records and prior imaging, draft an MRI order, an encounter note and a reply for a clinician to review, and then prepare the authorisation request for a specialist to check. CVS Health is the furthest along of the named partners, according to coverage of the announcement. One of its production agents cleared a months-long document-sending backlog over a weekend, while another handles the system steps behind tens of thousands of specialist consultations a day. Cigna Healthcare is using the agents to assemble member context for care managers who previously gathered information across more than ten separate systems. Abridge is building discharge coordination, and Ambience has designed its prior-authorisation agent to stop and ask a clinician when the record is ambiguous. ## The safety figures — and what they do not prove OpenAI's headline safety figure comes from its own evaluation of GPT-6 Astra on EHR inbox work, shaped with physician input. Reviewers judged whether inaccuracies, omissions or actions in the model's work could plausibly lead to patient harm. They rated 99% of 325 reviewed tasks safe, and rated the work "very helpful" or "extremely helpful" in 93% of 646 assessments. On HealthAdminBench, an independent benchmark from Stanford researchers covering prior authorisations, appeals and medical-equipment orders, Astra earned 94.3% of available rubric points. Luna earned 91.8% while costing 8.5 cents per task. Those numbers should be read for what they are: physician assessments of the model's work, not measured patient outcomes. OpenAI's head of Health AI, Karan Singhal, told Fierce Healthcare that researchers found no case in the remaining 1% where the model created a specific, plausible risk of patient harm, as reported by AI Frontier Post. Independent replication in live hospital settings is what would turn a vendor result into a settled one. ## Controls, privacy and the new plugins OpenAI says organisations can add approval steps, restrict which applications and records an agent can touch, and keep an audit log of every action. Patient data stays under the organisation's control and is not used to train its models, the company says, while HIPAA-covered use requires a Business Associate Agreement and the required security settings. The companion update to ChatGPT for Healthcare improves trusted clinical search — how the product finds, evaluates and cites medical evidence — and adds more than 20 plugins. The named partners include Aiva, Amion, Assort Health, Availity Extend, Elation Health, Ferry Health, IMO Health, Infinitus Relay, KMED, Luma Health, Metriport, Notable, Oracle HDI, PenguinAI, Photon, Redox, Stedi, Subsalt, TriZetto, Turquoise and UpToDate Expert AI. Together they cover scheduling, referrals, records, insurance checks and clinical reference. Workspace administrators must enable the new plugins, and computer-use access in ChatGPT Work and Codex is subject to workspace settings. The separate ChatGPT for Clinicians service is free for eligible, verified clinicians in the United States. ## Why this matters Healthcare runs on software that often does not talk to itself. The record lives in one system, the insurer's portal in another, and someone in between retypes the same facts by hand. If these agents work as described, under human review and with a full audit trail, they could return hours to clinicians and administrators. The test will be what happens in production, under audit, in hospitals beyond the large, well-resourced launch partners. For now, the honest summary is narrower than the headline: a real product, available now, aimed at the least glamorous and most expensive part of healthcare — the paperwork between the decision and the care. ## Frequently asked questions **What did OpenAI announce for healthcare on 16 November 2026?** OpenAI made computer-use agents available for healthcare work through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians, and added improved clinical search and more than 20 healthcare plugins to ChatGPT for Healthcare. **How safe did OpenAI say GPT-6 Astra is on clinical tasks?** In OpenAI's own evaluation, physician reviewers rated 99% of 325 EHR inbox tasks clinically safe, and rated the work very or extremely helpful in 93% of 646 assessments. On Stanford's independent HealthAdminBench, Astra earned 94.3% of rubric points. These are assessments of the model's work, not measured patient outcomes. **Which health organisations are using the agents first?** CVS Health, Cigna Healthcare, Abridge and Ambience are the named partners. CVS says one production agent cleared a months-long document backlog over a weekend; Cigna is using agents to assemble member context its care managers previously gathered across more than ten systems. Source: OpenAI — OpenAI for Healthcare; second source: AI Frontier Post — "OpenAI brings computer-use agents to healthcare — GPT-6 Astra rated safe on 99% of 325 EHR tasks" (https://aifrontierpost.com/articles/openai-computer-use-agents-healthcare-gpt-6-astra/).