Skip to content

HealthShare AI Assistant

This page tracks HealthShare AI Assistant and its boundary with Clinical Viewer and Unified Care Record.

Supported Positioning

InterSystems announced HealthShare AI Assistant in November 2025 as a generative-AI capability for users of HealthShare. The release positions it as a way for clinicians, case managers, and administrators to ask questions, request patient summaries, and navigate longitudinal records using conversational interaction.

Clinical Viewer and Navigation Application Boundary

The launch release says HealthShare AI Assistant operates within HealthShare Clinical Viewer and the Navigation Application. It also describes pre-built and custom prompts, source traceability, configurable data profiles, role-based access controls, and audit tracking.

That makes HealthShare AI Assistant relevant to Clinical Viewer analysis, but it remains distinct from the viewer itself. Clinical Viewer is the clinical presentation component for HealthShare/UCR information; AI Assistant is a newer assistant capability embedded into that workflow surface.

InterSystems also publishes a HealthShare AI Assistant sample repository with deployment configuration examples. The repository confirms that detailed deployment and use documentation exists, but access requires an InterSystems account, so public verification of configuration, prompt, model, and governance detail remains limited.

AI Model Services and Customer Boundary

InterSystems product terms add an important commercial and operational boundary: HealthShare AI Assistant is listed in the UCR 2025.2 product terms and uses HealthShare AI Model Services, described as a Runtime Service powered by external AI vendors. The public terms and AI guidance do not identify every customer-specific model, profile, region, data flow, or contractual term, so those details remain deployment-specific.

The strongest public governance example found in this pass is Healthix in the United States. InterSystems says that deployment uses private-cloud processing, consent and access controls, source references, discarded working data, no model training from patient data, and no autonomous writes, actions, or advice. This is useful as a governance pattern, but it is vendor-published and US-specific; it is not UK availability or UK clinical-safety evidence.

Evidence Status

Confidence is moderate-high for vendor launch positioning, stated workflow/governance claims, the existence of sample deployment configuration artefacts, and the public product-term boundary for AI Model Services. Confidence remains limited for UK deployment, clinical-safety case, model/profile specifics, prompt-safety material, and customer-specific runtime terms.

Remaining Follow-up Evidence

  • Clinical-safety, governance, model, prompt, runtime-service, and audit documentation for a UK deployment.
  • Public documentation for deployment and use beyond the account-gated material referenced by the sample repository.
  • Customer or supported-region evidence for UK availability and deployment.