Prepared by Chris for Dave and Andres, November 18, 2026
New medical assistants (MAs) are answering patient messages, handling refills and sending orders while still in training. That work is the bottleneck.
| Item | Fact |
|---|---|
| Release | Preview, launched November 10, 2026. Not generally available. |
| Access | Gemini Enterprise Standard and Plus customers. No published price. Those plans list from about $30 per user per month. |
| HIPAA | Google says it is "built to support HIPAA compliance." Google's HIPAA rules say not to use preview services with patient data unless Google states otherwise. Google has published no such statement for this product. |
| Covered today | Gemini Enterprise and Gemini Enterprise Agent Platform are on Google's HIPAA-covered list. Gemini for Healthcare is not. |
| EHR links | Read-only connections to Epic and Oracle Health electronic health records (EHRs). The Epic connector is also a preview. NextGen is announced. Other systems connect through partners such as Keragon. |
| Orders | It cannot place them. It reads from the EHR and does not write back. |
| Built-in tasks | Record summaries, pre-visit prep, coverage and policy review, and payer benefit checks. Practices can add their own. |
| Results | Google has published no accuracy or error rates. |
The first three steps do not depend on Gemini for Healthcare, and all of them carry over to whichever AI you use.
Shadow the MAs. Count messages and orders by type, and time each one.
Your policies, FAQs and refill rules, plus the messages that always go to a provider.
Run staged messages with no real patient information through the AI. Compare its drafts with what an MA would send.
Only on a service with a signed BAA: Claude under Anthropic's BAA, or Gemini or Claude on Google Cloud under Google's BAA. An MA approves every draft before it goes out.
Once Google covers it, run it against the same playbook and staged test set.
Timing is an estimate. It depends on API access to your EMR (electronic medical record) and messaging tool.
Each phase starts only after the one before it shows MA hours saved.
The plan above. One AI agent drafts replies to routine messages, and an MA approves each one.
Refill and order prep for a provider to sign. Plain-language notes for results the provider marks normal.
Membership, billing and scheduling questions. Intake follow-ups and records requests for new members.
Monthly accuracy and hours-saved report. Andres owns the playbook. Re-test Gemini for Healthcare once Google covers it.
Timing is an estimate. Team: Chris leads. Jordan and his team build.
First step: send the name of your EMR and messaging tool. That decides how the pilot connects.
Sources, checked November 18, 2026:
Not legal advice. Confirm BAA coverage with each vendor before using patient data.