SOLUTIONS/DEFENCE

Decisions for the operators on the timeline.

Multi-domain fusion across imagery, RF SIGINT, radar, and acoustic feeds, resolved through a shared operational ontology. Persistent multi-modal object tracking, classification-bounded outputs, full lineage to a tamper-evident audit chain — all replayable end to end by your IG, ATO, and treasury board.

VERTICAL NOTELive demo at demo.maiaintelligence.io/fusion — Joint ISR Fusion for Arctic Operations, sanitized and public. Methodology paper at /research/multi-domain-fusion-2026. Available through Canadian SACC + IDEaS, U.S. SAM.gov + DoD prime/sub vehicles, and NATO interoperable procurement. Sovereign-cloud deployment for Protected B and equivalent workloads. Aligned to NIST AI RMF + the GoC NRC + DRDC research-security trend cards (Deep Learning for Health Decision Systems, Cognitive Computing).
01 · WHY DEFENCE OPERATIONS BREAK

The signal exists. The decision doesn’t.

6+
modalities (imagery · radar · RF · audio · video · text · telemetry) operators must mentally fuse on the timeline
62 days
median time from operational signal to documented response in mid-tier programs
73%
of mission data captured but never reaches a decision in the same calendar quarter
$3.1T
annual cost of poor decisions across global enterprise and public sector
MAIA, State of Operational Decision-Making 2026
02 · DECISIONS MAIA RESOLVES

Three decisions your operators stop waiting on.

USE CASE 01 · DEF-ARCTIC-FUSION

Persistent Arctic domain awareness — fuse satellite, radar, RF, acoustic into one Track

Signal
RADARSAT-class satellite pass returns an unidentified surface contact at 73.2°N, 095°W. Within 4 minutes, NORAD long-range radar at Inuvik reports a candidate plot in the same window, and an RF SIGINT receiver flags a low-frequency pulse from the same bearing.
Decision
Spatiotemporal correlator joins the three reports into a Detection at p=0.86, σ=±2.1 km. Track correlator associates with persistent Track TRK-ARC-217. Investigation auto-opens; the Track enters EngagementZone NWP-04 (Northwest Passage chokepoint), triggering Authorization lens. Output classified PROTECTED-B per max-input rule, capped at tenant policy.
Writeback
Track update written to the C2 system with multi-modal lineage (sat + radar + RF). Investigation case anchored to audit chain hash. Operator sees confidence, σ, contributing observations, and full reasoning chain in the explainability drawer.
USE CASE 02 · DEF-MISSION-TASKING

Task the right asset to the right mission inside the OODA loop

Signal
Pop-up tasking inside a 90-minute window; three platforms within range, two with prior-week sustainment overlap.
Decision
Route to the platform with continuity, attach the prior-week sustainment summary, hold for human override per ROE, escalate if no acknowledge.
Writeback
Tasking record updated · operations officer notified · audit packet pre-assembled for after-action and IG readiness.
USE CASE 03 · DEF-SUSTAINMENT

Reserve sustainment capacity before the next surge

Signal
Three platforms flagged for scheduled maintenance this week; two depots projecting 92% utilization next 72h.
Decision
Pre-stage two pool platforms to depot 2, route a service swap overnight, notify the depot officer with the projection.
Writeback
Sustainment system updated · notification logged · variance posted to the readiness record.
USE CASE 04 · DEF-INTER-AGENCY

Route a signal across services without losing it

Signal
Field operator in service A flags a development that materially affects service B's planning case open this week.
Decision
Open a routing record, attach the field signal as evidence, notify both service leads, hold for human review per cross-domain policy then escalate.
Writeback
Cross-service record created · notification audit chain D-... → S-... → R-... preserved end to end.
03 · COMPLIANCE POSTURE

Built so your auditor doesn’t have to take our word for it.

Audit-grade decision lineage
Every fused Track, every Detection, every action is replayable end to end via the FNV-1a hash chain. Built so an inspector general can trace cause to effect, modality by modality, without phoning the vendor.
Cross-classification fusion enforced
Classification is a runtime property on every entity. Read-down enforced (a scorer at level N can only read entities at ≤ N). Fused outputs labelled at MAX(input classifications) and capped at tenant policy. Bell-LaPadula in software, not in documentation.
Spatiotemporal alignment + uncertainty
Multi-modal contact reports join through a Mahalanobis-style space-time gate. Sensor reliability priors learned online from the audit chain. Confidence + uncertainty bounds propagate end to end and gate the action tier (auto / review / critical).
Protected B sovereign tenancy
Per-program tenancy. Canadian, U.S., and Five Eyes residency available with REL TO markings enforced at every read. Sovereign-cloud deployment for Protected B and equivalent workloads.
Aligned to GoC research-security
Built against the NRC + DRDC trend cards (Deep Learning for Health Decision Systems, Cognitive Computing). Severity-banded approval gates, hash-chained audit, reproducibility contract — every named concern mapped to a runtime safeguard.
SWaP-aware edge roadmap
Phase 1: cloud + tactical command post (production today). Phase 2: compiled Rust + sqlite-rs runtime for tactical wearable, vehicle-mounted, forward command post under degraded connectivity. Edge scorecard published at /edge on the live demo.
04 · WHAT TO EXPECT FROM A PILOT

Forward-looking, not retrofitted. How a defence pilot is scoped.

Scope
A typical defence pilot covers one program or one mission corridor, two flows from {tasking, sustainment, cross-service routing}. We integrate with your existing personnel, asset, and notification systems via supported APIs and respect the classification, residency, and procurement constraints set by your authorising authority.
Success criteria
Defined against your current baseline. Time from operational signal to documented action measured before and after. Audit packet pre-assembly per record. After-action and IG-ready logs out of the box, with redaction tooling exposed in the pilot tenant.
Timeline
Week 1, procurement onboarding, connectors, residency + classification confirmation. Weeks 2–4, sandbox decisions, first sandbox writeback. Weeks 5–10, expand scope and run an IG dry-run review. Weeks 11–14, production graduation review with your authorising authority, ATO, and operations leads.
First decision in production
Target, the first MAIA-authored decision lands in your operational system within 30 days of kickoff. Defence pilots run a dedicated audit + ATO dry-run before production cut-over.

MAIA is in pilot with a small number of design partners, by design. We do not publish anonymised partner outcomes here, the way many vendors do, and we will not invent ones. When we have an outcome we can name on a logo and attest to in audit, we will publish it. Until then, what you see above is how we structure the engagement, end to end.

Inside the loop, every loop.

Request a Defence briefing →See the sandbox