Sovereign AI for regulated Canada.

Three models on one vertically integrated stack. Each makes the next more defensible.

• Phase 0, Sovereign Compute Pilot (reference build): grant-eligible proof of concept. Portable compute deployed on client sites to validate data sovereignty, a NIST 800-53 baseline, and the procurement path before production scale.
• Model 1, Sovereign AI Deployment (target operating model): bespoke software plus managed on-premise or co-located infrastructure for regulated clients. Clients own the data. Compliance is built in, not bolted on.
• Model 2, Proprietary Data Assets (scaling): anonymized, permissioned operational data from Model 1 becomes licensable training material and enrichment APIs.
• Model 3, Sovereign Compute Hubs (government-targeted): inference and HPC clusters for government co-location, on multi-year contracts. Operational control and data residency create the switching cost.

One line: we build the only AI stack regulated buyers are legally allowed to use.

Recurring by design. At our minimum base retainer, adding one client a month compounds into real run-rate inside a year, and most deployments bill well above base.

Illustrative model at the minimum base retainer, held flat for clarity. Not a forecast or a commitment; actual contract values vary and most run higher.

Municipal government and healthcare are our starting focus. The same sovereign stack serves both, and healthcare is the higher-value, higher-defensibility wedge.

• Municipalpredictable procurement, defined compliance, faster first contracts. The proving ground.
• Healthcarelarger contracts, deeper switching costs, deep institutional support (Infoway, Vector, CIHR), and public outcome proof (CHARTWatch cut unanticipated in-hospital deaths 26%, peer-reviewed in CMAJ, Sept 2024).
• The healthcare partnership and pilot motion runs on a separate deck. The point: one stack serves both verticals, so we are not dependent on any single segment, and it extends to the regulated industries on the next slide.

Sources: CMAJ / Unity Health Toronto (2024).

Four segments where regulated buyers need AI and have defined procurement paths to buy it.

The demand signal is public: Canada Health Infoway's AI Scribe Program is the largest federally funded health AI deployment in Canadian history (10,554 clinicians enrolled, 708,654 encounters by Aug-Sep 2025). Regulated sectors are moving on AI now, and they need a stack they are legally allowed to use.

Sources: Canada Health Infoway (2025).

Three forces open a 24 to 36 month window.

• Regulatory urgencyNIST 800-53 Rev. 5, provincial AI procurement rules, and EU AI Act spillover into Canadian standards are forcing buyers to act.
• Sovereignty pressurefederal strategy increasingly favors domestic AI scale-ups; public-sector buyers are actively moving off pure US cloud dependence.
• Talent arbitrageCalgary and Edmonton AI ecosystems (Amii, University of Alberta) are under-recruited relative to Toronto and Vancouver.

Miss the window and incumbents (Deloitte, CGI, IBM) fill it with overpriced, underperforming solutions.

Note: prior versions cited a specific power-draw multiplier for the compute hardware. We removed it pending a measured benchmark rather than ship an unverifiable claim.

Built with no outside capital. This is a deployed systems company, not a science project.

• Phase 0 reference buildA geotechnical knowledge assistant running on Oction's on-premise stack, validated in a regulated Canadian enterprise environment. This is a proof-of-concept, not a contracted service engagement.
• Autonomous automation layer operating on dedicated hardware around the clock.
• 30,000+ indexed knowledge chunks across a triple-layer RAG/CAG/KAG memory system with hybrid search.
• Live infrastructurelocal LLM proxy, memory API, vector store, and cache, all healthy.
• NIST 800-53 security framework implemented, documented, and audit-ready.
• Active buyer pipelinemultiple Alberta municipal conversations and two systems-integrator relationships at pilot stage.

Honest framing: Oction Labs has not entered into any service or revenue contracts under the company to date. Conversations and pilots are in progress; no production enterprise contract is signed yet. We do not claim revenue we do not have.