The whole dayMon 16 Nov 2026
AI Compass
WeeklyArchiveAbout

OpenAI adds computer-use agents and 20+ ChatGPT Healthcare plugins

Available on 16 November 2026 across ChatGPT for Healthcare, free US ChatGPT for Clinicians, ChatGPT Work, Codex and the OpenAI API under HIPAA Business Associate Agreements.

Mint-and-teal gradient OpenAI launch card reading More ways to move care forward with ChatGPT for Healthcare
Image: OpenAI

OpenAI rolled out computer-use agents, upgraded clinical evidence search and more than 20 healthcare plugins on 16 November 2026 across ChatGPT for Healthcare and ChatGPT for Clinicians (free for verified US clinicians), with computer-use workflows also live in ChatGPT Work, Codex and the OpenAI API. Building on enterprise governance tools introduced alongside Businesses in ChatGPT, workspaces covered by an OpenAI Business Associate Agreement (BAA) can restrict accessible apps, enforce human approval gates, capture step-by-step audit logs and exclude clinical data from model training.

Stanford HealthAdminBench scores, CVS Health Agent Kit and EHR plugins

Evaluating multi-step screen navigation across electronic health records (EHRs) and insurer portals without full APIs, physicians rated GPT-6 Astra clinically safe on 99% of 325 EHR inbox tasks and very or extremely helpful on 93% of 646 reviews. On Stanford’s independent HealthAdminBench covering prior authorisations, appeals and equipment orders at maximum reasoning effort, GPT-6 Astra scored 94.3% of rubric points while GPT-6 Luna scored 91.8% at 8.5 cents per task, 98% cheaper than Astra.

In live deployments, CVS Health cleared a months-long document-sending backlog over a weekend, processes tens of thousands of daily consult steps and is open-sourcing its Agents SDK Agent Kit, alongside production workflows at Abridge, Ambience, Assort Health and Cigna Healthcare. More than 20 plugins now run inside ChatGPT and ChartGPT, including UpToDate Expert AI, Availity Extend, Elation Health, IMO Health MCP Server, Infinitus Relay, Metriport, Oracle HDI, PenguinAI Medical Coding and Turquoise Health, paired with clinical search improvements preferred across all eight dimensions in over 1,300 blind physician evaluations.

Key details

HealthAdminBench results (max effort)
GPT-6 Astra 94.3% rubric score; GPT-6 Luna 91.8% at 8.5 cents per task (98% lower cost than Astra)
Physician clinical evaluations
99% clinically safe (325 EHR inbox tasks); 93% very/extremely helpful (646 reviews); 1,300+ blind search reviews
Supported surfaces and procurement
ChatGPT for Healthcare (direct or AWS Marketplace), free US ChatGPT for Clinicians, ChatGPT Work, Codex and API
Compliance and open-source kit
HIPAA BAA eligibility, RBAC, action/screenshot audit logs, zero model training; CVS Health Agents SDK Agent Kit

Release details & context

Multi-plugin clinical chains and Athena EHR referral close-outs in Codex

OpenAI documented multi-connector workflows inside a single ChatGPT conversation, such as pairing Luma Health unconfirmed appointment lists with TriZetto eligibility checks, combining Metriport outside hospital discharge summaries with UpToDate medication monitoring guidance, or chaining IMO Health procedure codes, Stedi benefits checks and Turquoise Health negotiated hospital rates to estimate out-of-pocket knee MRI costs.

Where partner APIs cover only partial workflows, teams pair plugins with GUI computer use: Assort Health attaches referrals via Athena’s API and then uses computer use inside Codex to select the referring provider and mark the referral visit scheduled in the Athena web interface, while Cigna Healthcare consolidated context previously spread across more than 10 separate clinical and administrative portals.

Common questions

Does OpenAI train its foundation models on patient records processed in ChatGPT for Healthcare or ChatGPT for Clinicians?

No, content shared in ChatGPT for Healthcare and ChatGPT for Clinicians is excluded from model training, and both offerings support HIPAA Business Associate Agreements when configured with eligible services and retention controls.

How do Ambience and CVS Health prevent autonomous computer-use agents from submitting unverified clinical actions?

CVS Health validates state via screenshots and action logs inside enterprise infrastructure, while Ambience pauses whenever chart evidence is ambiguous and requires explicit clinician sign-off before submitting prior-authorisation forms or patient replies.

OpenAI release timeline (44 launches tracked)

  1. Earlier10 Nov 2026
    OpenAI launches Businesses in ChatGPT beta and Business login for third-party tools

    OpenAI opens Businesses in ChatGPT in beta with verified profiles, live product feeds and Business Agents, plus Business third-party login.

  2. Earlier11 Nov 2026
    OpenAI rolls out Finances in the ChatGPT desktop app for US users across all tiers

    OpenAI rolls out Finances in the ChatGPT desktop app for US Free, Go, Plus and Pro users to analyse spending and build custom tools.

  3. Earlier12 Nov 2026
    OpenAI unveils Daybreak Defense Factory and Red Team cyber agents

    OpenAI unveils early-access Daybreak Defense Factory and Daybreak Red Team agents for continuous web-app penetration testing and Codex patch drafting.

  4. Earlier13 Nov 2026
    OpenAI launches ChatGPT Wallet for tokenised agent payments in dots

    OpenAI launches ChatGPT Wallet in dots for US Pro subscribers, using Amex, Mastercard and Visa payment tokens so agents never see card numbers.

Explore the full OpenAI dossier or browse all product releases.

Also in the Monday 16 November edition

Read all 15 launches on one page
  1. 01
  2. 02
  3. 03
  4. 04
  5. 05
  6. 06
  7. 07
  8. 09
  9. 10
  10. 11
  11. 12
  12. 13
  13. 14
  14. 15

Spotted an error? Email aiforsuccess.in+compass@gmail.com or tag @ai_for_success on X