Google and InterSystems Bridge the Gap Between AI Agents and Health Data

BOSTON – November 10, 2026 – For the past three years, the healthcare technology sector has been captivated by the promise of generative artificial intelligence. Yet, behind the polished keynote presentations and ambient scribes, a fundamental plumbing problem has persisted: large language models are exceptionally smart, but they lack secure, standardized hands to interact safely with living patient data. Today, a new alliance aims to solve that infrastructure bottleneck.

InterSystems, a privately held data technology giant that powers more than one billion health records globally, has officially joined Google Cloud as a launch partner for Gemini for healthcare. The collaboration connects Google’s Gemini models directly to the InterSystems IRIS FHIR Model Context Protocol (MCP) server. In practical terms, this integration grants AI agents governed, standardized access to clinical data, signaling a critical transition from passive, conversational chatbots to active, agentic workflows capable of operating at the bedside.

For health system chief information officers and enterprise architects, the announcement represents a structural shift in how artificial intelligence consumes and interacts with electronic health records. By exposing the open Model Context Protocol directly from a standard Fast Healthcare Interoperability Resources (FHIR) endpoint, the partnership addresses the friction of linking high-parameter models to heavily governed data streams without bypassing the compliance protocols that keep hospitals running safely.

The Infrastructure Breakthrough: MCP Over FHIR

To understand the gravity of this integration, one must look at the underlying mechanics of how AI currently interfaces with clinical systems. Historically, connecting an algorithm to an electronic health record required brittle, custom-built API connectors for every specific tool or workflow. The Model Context Protocol, introduced as an open standard to bridge AI hosts and external systems, acts as a universal adapter.

By serving MCP directly from a FHIR endpoint, InterSystems IRIS for Health acts as a centralized, normalized clinical memory for Gemini. MCP-enabled clients can now dynamically discover and execute FHIR tools to search, read, create, update, and manage patient data.

“Healthcare organizations are looking for practical ways to apply AI to their data without compromising security, governance, or interoperability,” said Don Woodlock, president of InterSystems. “By connecting Gemini for healthcare with InterSystems IRIS FHIR MCP, we are establishing a trusted bridge between advanced AI and the health information these organizations rely on.”

The bidirectional potential of this architecture is what separates an agent from a simple query tool. An AI agent can not only consume a patient's complex medical history to provide a triage summary, but the protocol technically allows it to enrich the clinical record by writing structured findings back as standard FHIR resources.

“The leap from passive chatbots to agentic workflows depends entirely on interoperability. An AI agent is only as safe and competent as the pipeline feeding it,” noted Aashima Gupta, global managing director of Healthcare for Google Cloud. “Partnering with InterSystems grounds Gemini for healthcare in the bedrock of global healthcare data infrastructure, turning static medical records into active, governed intelligence that works safely at the bedside and beyond.”

The Cloud Giants’ Healthcare Turf War

Beyond the technical achievements, the InterSystems and Google Cloud partnership is a strategic maneuver in a fiercely contested enterprise healthcare IT market. The major cloud providers are currently locked in a turf war, aligning themselves with leading electronic health record vendors to capture the lucrative clinical AI ecosystem.

Microsoft has established a formidable lead through its deep, strategic collaboration with Epic Systems, which currently holds a commanding 42 percent share of the United States hospital market. Microsoft’s Azure OpenAI Service is tightly woven into Epic’s software, powering tools like AI Charting and clinician-facing agents that ambiently draft notes and summarize charts. Meanwhile, Amazon Web Services has deepened its ties with Oracle Health, focusing on automating professional fee coding and chart reviews within a scalable cloud architecture.

Google Cloud needed a powerful wedge to challenge Microsoft’s grip on Epic-dominated hospital networks. InterSystems provides exactly that. With a footprint spanning 80 countries and a platform that solves interoperability and scalability for massive, disparate datasets, InterSystems is a vendor-agnostic powerhouse. By anchoring Gemini to InterSystems’ vast data lake, Google Cloud is bypassing the application-layer monopoly of the dominant electronic health record vendors and integrating directly at the foundational data interoperability layer. It is a calculated play to offer health systems an alternative, highly scalable AI ecosystem that does not strictly rely on a single software vendor’s proprietary roadmap.

Write Access for Algorithms: The Governance Challenge

However, the transition from AI that summarizes data to AI that acts upon it introduces a minefield of regulatory, compliance, and ethical risks. Granting algorithms write access to a patient’s medical chart is the ultimate red flag for healthcare risk managers. If an AI agent hallucinates a diagnosis or executes an errant medication update into a production environment, the consequences could be fatal.

This is why the concept of governed access is the linchpin of the Google and InterSystems deployment. While the MCP protocol technically enables full create, read, update, and delete capabilities, the current implementation of Gemini for healthcare emphasizes strict guardrails. In its initial preview phases, Gemini primarily offers read-only access to major health records to maintain data integrity.

Healthcare compliance officers are closely monitoring how these agentic workflows align with evolving federal oversight. The Food and Drug Administration's guidelines for Software as a Medical Device heavily scrutinize AI modules that influence clinical decisions. The agency's Predetermined Change Control Plan requires developers to manage updates to adaptive algorithms meticulously, ensuring that an AI agent does not silently drift into unsafe behavior over time.

Furthermore, the American Medical Association has drawn a hard line on AI liability. The association's policy dictates that liability for AI-related harm should rest with the parties best positioned to mitigate those risks—namely, the developers designing the tools and the health systems deploying them. Physicians, the AMA argues, should not be held liable for algorithmic errors they cannot reasonably detect or control.

To navigate this liability framework, Google Cloud has engineered Gemini for healthcare to operate with mandatory human oversight. Clinical data, prompts, and generated outputs remain strictly within the customer's governed environment and are never used to train Google’s foundation models. Every AI-generated output is grounded in clinical records with traceable citations and exhaustive audit logging, providing IT leaders with the visibility required to intervene before an error reaches the bedside.

The integration of Google Cloud’s Gemini with the InterSystems IRIS FHIR MCP server is undeniably a watershed moment for clinical infrastructure. It proves that the technical pipelines required for autonomous healthcare AI are no longer theoretical. Yet, as this technology moves from the server room to the clinic, the ultimate success of agentic AI will not be determined by its ability to write code or query databases, but by the industry's ability to build liability frameworks that clinicians can actually trust.