Google Cloud launched Gemini for Healthcare, an AI agent designed to connect hospital and insurer systems and handle administrative and information tasks such as summarizing records, checking coverage, preparing cost estimates and prior authorizations, and searching medical guidance.

The product is initially available in preview in the United States, with launch customers including Northwell Health, Highmark Health, HCA Healthcare, Optum, Blue Cross Blue Shield of Massachusetts, Hartford HealthCare and Stanford School of Medicine.

Unlike a chatbot, the agent can turn a staff member’s goal into a multistep plan, retrieve information under organizational rules and provide citations for people to review; organizations can choose models suited to clinical reasoning or faster paperwork tasks.

It draws on health records, billing platforms, clinical guidelines and U.S. Medicare datasets, while working within defined guardrails and human oversight.

A key part of the integration is matching records across organizations and systems to the right person, giving AI workflows a more complete and trusted identity foundation.

Healthcare organizations can configure workflows around their own clinical protocols and existing permissions, with governance tools that support multiple models and link generated answers to records and payer policies.

The broader ecosystem includes integrations or connections from GoodRx, UiPath, Verato and InterSystems, while Altimetrik, AuxoAI and Egen will help customers build and deploy the agent.

The launch is aimed at reducing the administrative load on clinicians, who the article says lose nearly 28 hours a week to paperwork and related tasks.

Northwell CEO John D’Angelo said fragmented records can force clinicians to repeat tests or make decisions without seeing relevant history, contributing to unnecessary care and higher costs.

Highmark Health President Karen Hanlon said prior authorization is clinicians’ biggest administrative frustration; Highmark reports that 78% of approved outpatient authorizations submitted digitally are processed automatically, with clinicians reviewing cases before a denial.

Whether the system delivers value will depend on cutting manual handoffs and delays without weakening review, since organizations must decide which tasks AI can handle and which require clinician or administrator approval.

Accuracy is another central concern: inaccurate summaries or missed information can create more work, so organizations should measure benefits in specific workflows and be ready to withdraw systems that do not deliver.