Policy to plan to scenario to action.
CDI turns a policy requirement set, a municipal crisis plan, local context, and scenario pressure into a ranked improvement brief. This proof uses Skärvik kommun — a fictional Swedish municipality — to demonstrate the full disaster-cycle decision flow from baseline preparedness to governance.
The Scenario
A spring flood warning coincides with a 72-hour power and telecom disruption. On 26 April 2026, the east valley bridge closes at 07:00, transformer failure takes out power at 07:45, and the mobile network drops at 08:30. Skärvik must activate protected spaces for 800 residents, support seven elderly-care homes, coordinate public information without mobile network, and prepare for recovery without a named recovery lead.
Proof Flow
Policy → Municipal Plan → Gap Engine → Scenario Stress → Corrective Actions → Governance
Policy Requirement Gap Analysis — 6 of 9 requirements open
POLICY REQUIREMENT GAP ANALYSIS
| Requirement | Phase | Status | Severity | Pressure |
|---|---|---|---|---|
| Backup communication works during telecom and power failure | Early Warning | MISSING | CRITICAL | 15 |
| Elderly-care transport and home-care continuity are stress-tested | Emergency Response | MISSING | CRITICAL | 9 |
| Protected-space activation includes accessibility, staffing, and activation thresholds | Resident Assistance | PARTIAL | CRITICAL | 10 |
| Recovery ownership and service restoration are named before the event | Recovery & Transition | MISSING | CRITICAL | 8 |
| After-action corrections are logged and versioned | Method & Governance | MISSING | CRITICAL | 9 |
| Mandate-period plan for extraordinary events is current | Baseline & Preparedness | OUTDATED | CRITICAL | — |
| Risk and vulnerability analysis covers extraordinary events and own operations | Baseline & Preparedness | COVERED | LOW | — |
| Geographic-area coordination and public information are explicit | Emergency Response | COVERED | LOW | 9 |
| Annual reporting and state/county follow-up can be produced from evidence | Method & Governance | COVERED | LOW | 9 |
Engine: cdi-municipal-proof-2026-04. All findings require human review before action.
Disaster-Cycle Impact — All Six Phases
DISASTER-CYCLE IMPACT — ALL SIX PHASES
Protected-Space Readiness — 4 of 5 checks fail
PROTECTED-SPACE READINESS
4 protected-space readiness gaps found for the demo stress assumptions.
| Check | Observed | Required | Status |
|---|---|---|---|
| Total protected-space capacity | 2200 | 800 | ✓ |
| Accessible protected-space capacity | 80 | 120 | ✗ |
| Protected spaces with backup power | 1 | 3 | ✗ |
| Protected spaces with redundant communications | 0 | 3 | ✗ |
| Protected spaces with named staffing model | 2 | 3 | ✗ |
Activation population: 800. Demo assumptions only. Numbers require municipal confirmation.
Compound Shock Impact — What Changed by Hour 4
COMPOUND SHOCK IMPACT — 4 BINDING DOMAINS AT HOUR 4
Three shock events applied sequentially. Baseline (grey) vs post-shock (coloured).
| Domain | Baseline | Hour 4 | Change | Status |
|---|---|---|---|---|
| Communications | 0.30 | 0.03 | -0.27 | BINDING |
| Operational | 0.65 | 0.10 | -0.55 | BINDING |
| Infrastructure | 0.55 | 0.11 | -0.44 | BINDING |
| Care continuity | 0.65 | 0.20 | -0.45 | BINDING |
| Access | 0.72 | 0.35 | -0.37 | DEGRADED |
| Market | 0.85 | 0.47 | -0.39 | DEGRADED |
| Financial | 0.88 | 0.48 | -0.40 | DEGRADED |
| Protection | 0.85 | 0.85 | +0.00 | VIABLE |
Engine: cdi-shock-delta-2026-04. Multiplicative compounding. Uncertainty: generator fuel level unknown, home-care vehicle inventory unknown.
INKANYAMBA — Modality Feasibility at Hour 4
INKANYAMBA — MODALITY FEASIBILITY AT HOUR 4
Five modalities scored against the post-shock constraint profile. Ordered by feasibility score.
Engine: inkanyamba v1-deterministic. All verdicts require human review before operational use.
CDI — CONNECTED INTELLIGENCE
The backup-communications gap is both the top corrective action and the primary constraint on service delivery.
CDI found this with two separate engines pointing at the same root cause. The municipal proof engine identified the gap in the preparedness review. The INKANYAMBA engine identified the same gap as the binding constraint on coordinated service delivery. Fix the comms gap — the operational score rises — service delivery becomes viable.
Top Three Corrective Actions
TOP THREE CORRECTIVE ACTIONS — RANKED BY SEVERITY
EVIDENCE & BOUNDARY
Policy requirements are sourced from: MCF municipal preparedness agreement 2024-2026 · Vinnova civil-defence prototype call 2026 · SOU 2024:65 · MSB/WIS documentation.
The Skärvik municipal plan is a fictional document with deliberate gaps. No real municipality, individual, or personal data is used in this proof.
This proof is deterministic decision support. It does not allocate public benefits, classify individuals, or make autonomous emergency decisions. Every finding must be reviewed by a responsible human owner before it is acted on.
→ How the shock delta engine computed these constraint scores
→ Full INKANYAMBA engine documentation
→ Automation and AI Policy
Vinnova Pilot Funding Ask
CDI is seeking pilot funding under the Vinnova civil-defence prototype call 2026 (Morgondagens skyddsrum, call 2026-00702). The pilot needs: one municipal or regional needs owner; civil-defence and protected-space co-design; policy QA and Swedish municipal-plan evidence model; a scenario library covering power, telecom, flood, protected-space, evacuation, reception, and recovery; student or research-assistant validation; and a clear AI governance and automation boundary.
Sources: Vinnova call 2026-00702 · MCF municipal preparedness agreement 2024-2026 · SOU 2024:65 · MSB/WIS.
The Simulation — What Changes When CDI Is in the Room
WHAT SKÄRVIK HAD GOING IN — 26 APRIL 2026, 06:00
Three documents. This is what the crisis coordinator had available when the SMHI warning arrived.
These documents existed regardless of whether CDI was used. CDI’s value is what happens next.
WHAT CDI CHANGED — THREE CORRECTIONS IMPLEMENTED BEFORE THE EVENT
Corrections sourced from Tier 1 municipal proof engine output. Accept/reject decision by: municipal preparedness coordinator. CDI does not implement corrections — it identifies, ranks, and tracks them.
WHAT CHANGES WHEN CDI IS IN THE ROOM
Same disaster. Same municipality. Same timeline. Different preparation.
| WITHOUT CDI Skärvik as it was — gaps unfixed |
WITH CDI Top 3 corrective actions implemented |
|---|---|
|
06:20 (T+00:00)
SMHI issues Class 2 flood warning for Skärvik valley. Expected peak: 14:00.
ELEVATED
→ Warning received and relayed. Krisledningsnämnd chair alerted. No further action plan activated — crisis plan does not define trigger thresholds.
|
06:20 (T+00:00)
SMHI issues Class 2 flood warning for Skärvik valley.
ELEVATED
✓ Full communications posture established within 15 minutes. All channels active: Rakel, runners on standby, WIS live. No dependency on mobile network.
|
|
07:00 (T+00:40)
Älvdalsbron (east valley bridge) closes due to flood risk.
CRITICAL
✗ Solhemmet (68 residents, 34 mobility-impaired) cut off by road. Status: unknown. No action possible without transport protocol or alternative route.
Gap: care_transport
|
07:00 (T+00:40)
Älvdalsbron closes. East valley cut off.
MANAGED
✓ Response to bridge closure: 8 minutes. Solhemmet coordinator reached by Rakel within 25 minutes. Vehicle en route via coastal track. No residents isolated.
|
|
07:45 (T+01:25)
East valley transformer fails. Power outage: Solhemmet, Norrskolan, ~1 800 residents.
CRITICAL
✗ Solhemmet goes dark. 68 residents, no staff communication. Coordinator cannot verify resident welfare or activate care protocol. Generator at Medborgarhuset operational but activation threshold not defined.
Gap: backup_comms
|
07:45 (T+01:25)
East valley transformer fails. Power outage.
HIGH
✓ Medborgarhuset protected space activated within minutes of power failure — threshold was pre-defined. Solhemmet: runner present, generator failure noted, insulin refrigeration flagged to duty nurse.
|
|
08:30 (T+02:08)
East valley mobile base station battery depleted. Mobile network down.
CRITICAL
✗ COMMUNICATIONS BLACKOUT. Coordinator cannot reach Solhemmet, Norrskolan, or field units. 68 isolated elderly residents. No welfare-check capacity.
Gap: backup_comms
|
08:30 (T+02:08)
Mobile network down in east valley.
MANAGED
✓ Mobile loss: no operational impact. Rakel active. All 7 care facilities checked in via Rakel or runner within protocol window. Coordinator has full situational awareness.
|
|
10:00 (T+04:00)
Hour 4. Krisledningsnämnd convened at Kommunhuset.
CRITICAL
✗ No field picture available. No modality analysis. Committee makes ad-hoc decision: do not activate protected spaces yet. Solhemmet still unreachable. No welfare check.
Gap: protected_space_activation
|
08:50 (T+02:30)
INKANYAMBA modality assessment triggered.
MANAGED
✓ INKANYAMBA verdict: Hybrid — FEASIBLE 0.71. Goods pre-positioned via coastal route. Service coordination via Rakel. First structured decision at hour 2, not last improvised one at hour 24.
|
|
02:00 next day (T+20:00)
Runner reaches Solhemmet on foot via coastal path.
CRITICAL
✗ First welfare check: 20 hours after bridge closure. 3 residents need medication management. Insulin refrigeration failed 18.5 hours ago — clinical risk established.
Gap: care_transport
|
10:00 (T+04:00)
Hour 4 status: all facilities accounted for.
MANAGED
✓ All 7 welfare checks complete. 34 mobility-impaired Solhemmet residents transported to coastal facility. Insulin cold-chain maintained. Zero hospitalisations.
|
|
06:00 next day (T+24:00)
Power restored. Mobile network restored.
HIGH
→ Medborgarhuset opened as protected space — 24 hours after initial warning. 3 care residents require hospital transfer from insulin failure. Recovery: no named lead. Handoff undefined.
Gap: recovery_lead
|
06:00 next day (T+24:00)
Power restored. Acute phase winding down.
ELEVATED
✓ Named recovery lead activated. Structured handoff from krisledningsnämnd. After-action report scheduled. Plan correction cycle triggered.
|
|
06:00 day 3 (T+72:00)
Acute phase ends. Krisledningsnämnd deactivates.
ELEVATED
✗ Recovery undefined. No named lead. No structured handoff. 3 residents hospitalised. After-action report: not scheduled. Corrections: not logged.
Gap: recovery_lead
|
06:00 day 3 (T+72:00)
Acute phase ends. Krisledningsnämnd deactivates.
LOW
✓ 0 residents hospitalised. Event fully documented. Corrections logged. Plan revision triggered. Recovery lead in post.
|
OUTCOME SCORES — WITHOUT CDI vs. WITH CDI
| Domain | Without CDI | With CDI | Δ |
|---|---|---|---|
| Communications | 0.03COLLAPSED | 0.18DEGRADED | +0.15 |
| Care continuity | 0.20FAILED | 0.38DEGRADED | +0.18 |
| Operational | 0.10BINDING | 0.22CONSTRAINED | +0.12 |
| Infrastructure | 0.11BINDING | 0.11BINDING | +0.00 |
| Access | 0.35DEGRADED | 0.50DEGRADED | +0.15 |
| Recovery | 0.00UNDEFINED | 0.70MANAGED | +0.70 |
Scores computed by CDI shock delta engine. Infrastructure is unchanged — the transformer was always going to fail. What CDI changes is whether the municipality had a decision at each step, or only a gap to manage.
CDI SIMULATION — WHAT CHANGES
Three corrective actions. Zero hospitalisations. A named recovery lead before the event ended.
CDI identified the backup-comms gap in the preparedness review. Without CDI, Skärvik hit a communications blackout at T+2h — Solhemmet was unreachable for 20 hours and three residents were hospitalised from equipment failure. With CDI, Rakel and runners were active before the mobile network failed. All seven care facilities were reached within the protocol window. INKANYAMBA returned a Hybrid modality verdict at T+2:30 — the first structured decision, not the last ad-hoc one.
The simulation does not show a perfect outcome. Infrastructure still failed — the transformer was always going to fail. What CDI changes is whether the municipality had a decision to make at each step, or only a gap to manage.