Google Cloud launches Gemini for healthcare in preview

Google Cloud has announced an agentic AI offering for healthcare that reads from Epic and Oracle Health records without writing back. What was announced, what is still in preview, and the questions buyers should ask.

A doctor in a white coat with a stethoscope works at a desk computer, with a tablet beside the keyboard.
A clinician working at a computer in a medical office.

On 10 November 2026, Google Cloud announced Gemini for healthcare, which it describes as a purpose-built agentic AI solution for clinical, operational and administrative work. It is one of several industry packages built on Google's Gemini Enterprise platform.

What was announced

According to Google Cloud's announcement:

  • Status. The product is initially available in preview, for Gemini Enterprise Standard and Plus customers. No general availability date was given.
  • Use cases. Medical records synthesis, pre-appointment preparation, coverage and policy review, payer benefit determination, registration through to claims processing, and clinical guideline research.
  • EHR connections. Read-only connections to Epic and to Oracle Health (through its FHIR APIs). Google says the agents do not write back to the record, to protect clinical data integrity. A NextGen Healthcare connection is said to be in development.
  • Other data sources. Connectors to coding and drug vocabularies such as ICD-10 and RxNorm, US coverage data, clinical guideline content and a range of partner systems.
  • Models. Google says the platform can use models from Google, Anthropic and OpenAI, routing between faster and more capable models.
  • Launch customers. Named organisations include HCA Healthcare, Northwell Health, Hartford HealthCare, Highmark Health, Optum, Blue Cross Blue Shield of Massachusetts and Stanford School of Medicine.

On privacy and oversight, Google says patient data, prompts and outputs stay within the customer's own environment, that customer data is not used to train its foundation models, that answers are grounded in the patient chart with traceable citations, and that each agent inherits the user's existing role-based permissions. These are the vendor's own statements; the announcement does not cite independent clinical validation.

Why it matters

Two design choices stand out. First, the agents read from the record but do not write to it, which keeps clinicians responsible for what goes into the chart. Second, the product is built on connectors and standards such as FHIR, which shows how much the value of AI in healthcare depends on the quality and accessibility of the underlying data. As WHO's representative in Nigeria put it this month, AI depends on quality data and interoperability.

The launch is aimed at large, mostly US, health systems, and many of its connectors are US-specific. But the pattern applies anywhere, including African hospitals evaluating AI tools.

Questions to ask any AI vendor

  • Where is patient data processed and stored, and is it used to train the vendor's models?
  • Can the tool write to the record, or only read? Who approves anything it suggests?
  • Does each answer show its sources, so a clinician can check it?
  • Does the tool respect the same roles and permissions as the EHR?
  • Is every use logged for audit, and can the facility switch it off?
  • What evidence exists that it works in a setting like yours?

Syneom EHR's clinical decision support follows the same rules: it is advisory only, never writes to a consultation or order, sends only de-identified context to an AI provider the facility has authorised, and records every recommendation and action for audit. Read more in our article on AI clinical decision support.

Photo: A clinician working at a computer in a medical office. Shixart1985, CC BY 2.0, via Wikimedia Commons.

Get started

Plan your deployment with Syneom Health.

Fully developed healthcare technology solutions, ready for implementation across clinics, diagnostic centres, hospitals and healthcare institutions. Each deployment is configured to your facility’s departments and workflows.

The lit entrance of a medical clinic at night.