AI that runs inside the hospital, not across the border.

Canadian health AI is governance-constrained, not technology-constrained.

• A 2026 review cited by Vector Institute found 50% of medical AI studies tested models on exam-style questions and only 5% on real-world clinical data. The gap is trust and oversight, not raw accuracy.
• Health Canada regulates AI as a medical device, but national post-market surveillance and performance dashboards remain aspirational. Safety is a shared responsibility the regulator cannot close alone.
• Physician adoption of administrative AI is running ahead of liability and accountability clarity (CMA survey data, 2024).

The winning vendor is the one that makes AI provable, local, and auditable. That is the problem Oction is built to solve.

Sources: Vector Institute Toolkit v2.0; Health Canada AI-enabled medical device guidance (2024); CMA National Physician Health Survey (2024).

One platform, four capabilities that map directly to Canadian health privacy law.

Four forces make sovereign medical AI a near-term procurement reality, not a thesis.

• Fragmentation creates demand13 provincial and territorial systems, each with its own privacy law, EHR vendor, and consent model. A neutral, sovereign infrastructure layer is the structural fit.
• Regulatory pullPHIPA, HIA, FIPPA, Law 25, and PIPEDA all push toward data minimization and local processing. The US CLOUD Act makes cross-border cloud a documented risk.
• Funder momentumCanada Health Infoway, Vector Institute, CIHR, and IRAP are actively funding health AI governance, scribes, and interoperability.
• Proof the market is liveInfoway's AI Scribe Program enrolled 10,554 primary care clinicians and captured 708,654 clinical encounters by Aug-Sep 2025. This is the largest federally funded health AI deployment in Canadian history.

Sources: Canada Health Infoway AI Scribe Program (2025); CIHI / Infoway interoperability frameworks.

We enter through the buyers who own data risk and clinical workflow.

Anchor region for the first pilot: Alberta. Province-wide Epic (Connect Care), single health authority (AHS), and active Amii Health Innovation Lab support reduce integration and procurement friction.

Start with the safest, highest-trust use case. Earn the right to the clinical ones.

Phase 1: Clinical Knowledge Assistant (Months 1 to 4)

• Local RAG over hospital policies, protocols, and drug formularies.
• First usersnurses, residents, pharmacists. No diagnosis, no patient-facing risk.
• 90-day paid pilot scoped to one health authority, clinic, or research site.

Phase 2: Ambient Documentation / Scribe (Months 4 to 8)

• Local speech-to-text plus structured notes; EMR integration via HL7 FHIR and CA Core+.

Phase 3: Research Data Sharing (Months 8 to 12)

• Synthetic data and differential privacy for multi-site studies.

Phase 4: Embedded Clinical Decision Support (Year 2)

• Source-grounded CDS inside Epic, Cerner, or Meditech workflows.

We deliberately lead with documentation and retrieval, not diagnosis, to keep clinical liability low while trust is built.

We are looking for one reference health-system partner for a measurable 90-day pilot.

What the partner gets:

• A sovereign AI layer running on their hardware, their data never leaving the premise.
• A measured benchmark reportlatency, source-citation rate, and hallucination reduction on their corpus.
• Alignment with IRAP, Vector, Infoway, or CIHR to offset pilot cost with non-dilutive capital.
• No cloud lock-in and no per-token meter.

What we need from a partner:

• One clinical use case (policy and protocol retrieval is the recommended start).
• A data-governance and clinical sponsor.
• Access to a representative corpus to benchmark against.

The first partner becomes the reference that defines the category for Canadian health systems.

We label targets as targets. These convert to validated claims during Phase 1.

All figures above are targets until measured in pilot.