Part of our AI models explained: ChatGPT, Claude, Gemini and the rest guidePart of our Coding agents explained: Claude Code, Cursor, Codex and how to choose guide
Google Cloud on November 10, 2026 unveiled Gemini for healthcare, a preview agentic system that can pull patient records from Epic and Oracle Health and bring in billing and Medicare data for coverage reviews and related workflows, with what the company calls appropriate human oversight. A study from The Harris Poll and Google Cloud, cited in the same release, says clinicians lose nearly 28 hours each week to documentation and administrative work.

What the preview can connect to
Launch customers named in the release are HCA Healthcare, Northwell Health, Optum, Highmark Health, Hartford HealthCare, Blue Cross Blue Shield of Massachusetts and Stanford School of Medicine. The American Society of Clinical Oncology is listed as a contributor, not as a launch customer.
The release describes Gemini for healthcare as being built on Google Cloud's "universal Gemini agent," with specialized skills and connectors. For electronic health records, it offers direct, secure read integrations with Epic, via open.epic APIs, and Oracle Health, via Ignite APIs for FHIR, so care teams can pull active conditions, medications and recent labs into a workflow. Google Cloud says the system does not write back to the EHR, "to safeguard clinical data integrity." Connectivity for NextGen Healthcare is described as in development and slated for the coming months.
On the administrative side, the company says the product can look up federal and clinical standards including CMS Coverage Data, the National Provider Identifier registry, RxNorm and ICD-10. Prior authorization is mentioned as a workflow supported by a partner tool, Infinitus Relay, which the release also says automates benefit verification. Other named payment and eligibility connectors, including some in limited release, are listed as well. Those are Google Cloud's descriptions of the partner stack, not a claim that the preview itself already completes a prior authorization.
Google Cloud says the platform supports open model choice across Google, Anthropic and OpenAI, with "smart routing" that can send high-volume administrative tasks to faster, lower-cost models and reserve higher-reasoning models for more complex clinical and operational work. It will be available to Gemini Enterprise Standard and Gemini Enterprise Plus customers. The release calls this a preview; it does not say the preview is limited to the United States.
What the 28-hour figure is, and isn't
The nearly 28 hours a week figure is from the Harris Poll and Google Cloud study cited in the launch release. It is a statement about clinicians' current documentation and administrative load, not a measured result of Gemini for healthcare. The same release separately cites a 2025 American Medical Association report that 41.9% of physicians experienced burnout, and an American Hospital Association estimate that administrative costs account for 25% to 35% of total U.S. healthcare spending. Those are also source-attributed context, not outcomes of this product.
Google Cloud says the system "supports HIPAA compliance," that patient records, prompts and outputs remain in an organization's dedicated enterprise environment, and that it does not use customer data to train its foundation models. It also says AI responses are grounded in verified patient charts and an organization's approved guidelines, with traceable citations, and that "consequential decisions remain in qualified hands." Those are company statements about design and governance.
What to watch
The immediate test is whether a preview with read-only EHR access and a long partner list can actually shorten chart prep and coverage reviews without writing back into the record. Codema's Agents hub tracks how these packaged agents are being sold into specific industries, and our explainer on agent harnesses covers the tooling that sits around a model when the work is a multi-step workflow rather than a single chat. Watch whether NextGen connectivity arrives as described, how much of prior authorization is handled by Gemini itself versus partner tools, and whether Google Cloud later states a general-availability date or any geographic limit the November 10 release does not include.
FAQ
What is Gemini for healthcare?
It is a preview agentic system that Google Cloud describes as built on its "universal Gemini agent," with specialized skills and connectors. It will be available to Gemini Enterprise Standard and Gemini Enterprise Plus customers.
Which health record systems does it connect to, and can it write to them?
It offers direct, secure read integrations with Epic, via open.epic APIs, and Oracle Health, via Ignite APIs for FHIR. Google Cloud says it does not write back to the EHR "to safeguard clinical data integrity," and NextGen Healthcare connectivity is in development.
Who are the launch customers?
The release names HCA Healthcare, Northwell Health, Optum, Highmark Health, Hartford HealthCare, Blue Cross Blue Shield of Massachusetts and Stanford School of Medicine. The American Society of Clinical Oncology is listed as a contributor, not a launch customer.
Is the 28-hour figure a result of Gemini for healthcare?
No. It comes from a Harris Poll and Google Cloud study cited in the release and describes clinicians' current documentation and administrative load. It is not a measured result of the product.
How does Google Cloud say it handles patient data?
Google Cloud says the system "supports HIPAA compliance," that patient records, prompts and outputs remain in an organization's dedicated enterprise environment, and that it does not use customer data to train its foundation models. Those are company statements about design and governance.
Sources
- Google Cloud, Google Cloud Launches Gemini for Healthcare (Nov. 10, 2026)googlecloudpresscorner.com
Topics
John F.AI desk
John F. is a Codema.ai house byline: a pen name used by our editorial desk, not one individual journalist. It covers the AI beat: new models and how they're priced, open models, AI coding tools and agents, and security incidents that affect AI developers. Pieces under this name are written with AI assistance and reviewed by a human editor before they're published.
All stories by John F.





