36A. Insurance, Logistics, Energy and Aviation

This chapter presents four additional sectors in condensed form, completing the ten-sector coverage. Each follows the same analytical structure: the cross-domain phenomenon that current architectures cannot see, the correlation pattern that would reveal it, and the specific constraint that makes it difficult.

36A.1 Insurance #

The Invisible Phenomenon #

Claims fraud, particularly organized fraud, is a network phenomenon presented as a set of individual claims. A staged collision involves a driver, a passenger, a medical practitioner, a vehicle repairer, and a legal representative — each of whom appears in multiple claims across multiple insurers, in combinations that are individually unremarkable and collectively definitive.

Separately and more benignly, claims experience is multi-factorial in ways underwriting does not capture: a property’s claims history correlates with weather exposure, local infrastructure age, maintenance behaviour, and occupancy pattern — signals that exist across policy, claims, external, and IoT data but are not related.

Correlation Pattern #

flowchart LR
    subgraph IN["Sources"]
        direction TB
        I1["Policy administration"]
        I2["Claims management"]
        I3["Third-party involvement<br/><i>repairers, medical, legal</i>"]
        I4["Telematics / IoT<br/><i>where consented</i>"]
        I5["External<br/><i>weather, hazard, registry</i>"]
        I6["Industry fraud registers"]
    end
    subgraph SC["Scopes"]
        direction TB
        C1["<b>fraud.claim_network.v3</b><br/>M2 (k≤4) · M5 · 24 mo · batch<br/>Shared-participant community<br/>detection across claims"]
        C2["<b>uw.exposure_reality.v2</b><br/>M1·M2·M3 · 36 mo · batch<br/>Which factors actually predict<br/>claims, vs. which are priced"]
        C3["<b>ops.claim_trajectory.v2</b><br/>M1·M5·negative · 90 d<br/>Which claims are drifting<br/>toward dispute or leakage"]
    end
    subgraph OUT["Outputs"]
        direction TB
        O1["L2 — investigation<br/>prioritization with<br/>network evidence"]
        O2["L2 — underwriting<br/>factor review"]
        O3["L3 — staged claim<br/>handler intervention"]
    end
    IN --> SC --> OUT
    classDef a fill:#1b2f3d,stroke:#5b8fb0,color:#e6f2fa
    classDef b fill:#0a5570,stroke:#3fd0f0,stroke-width:2px,color:#ffffff
    classDef c fill:#3d3313,stroke:#d4a636,color:#fff8e6
    class IN,I1,I2,I3,I4,I5,I6 a
    class SC,C1,C2,C3 b
    class OUT,O1,O2,O3 c

Figure 51. Insurance correlation pattern (illustrative). Claim network detection is a graph community-detection problem over shared participants — the archetypal M2 application at depth.

The Specific Constraint #

36A.2 Logistics #

The Invisible Phenomenon #

Supply chain disruption propagates through relationships that no single system models. A port congestion event affects specific vessels, carrying specific containers, containing specific SKUs, destined for specific distribution centres, feeding specific stores, supporting specific promotional campaigns — with a customs delay, a driver-hours constraint, and a warehouse capacity limit interacting along the way.

Each system tracks its own leg. The compound delay — and, critically, which downstream commitments it endangers — is computed by nobody until something fails.

Correlation Pattern #

Table 73. Logistics correlation scopes (illustrative). Cold chain reaches L4 because diverting a shipment for inspection is genuinely reversible and the consequence of not acting — spoiled pharmaceutical or food product reaching consumers — is severe.

The Specific Constraint #

Logistics correlation crosses organizational boundaries constantly — carriers, ports, customs, 3PLs, and suppliers are separate entities with separate systems and no obligation to share. Data availability, not correlation capability, is the binding constraint. The framework’s contribution here is to make the value of shared data measurable, which may support the commercial case for sharing agreements.

Deep graph traversal (k≤5) is required to trace propagation, which pushes against the depth limits in Chapter 26. This scope must run in the batch layer with careful pruning rather than as interactive traversal.

36A.3 Energy #

The Invisible Phenomenon #

Grid reliability, asset health, and demand behaviour are analysed separately and are causally entangled. A transformer’s degradation rate depends on loading history, ambient conditions, harmonic distortion from distributed generation, and maintenance history. A local demand anomaly may indicate an emerging fault, a metering error, unauthorized abstraction, or a genuine behavioural change — and distinguishing them requires correlating meter data, network telemetry, weather, and customer records.

Correlation Pattern #

flowchart TB
    subgraph SRC["Sources"]
        direction LR
        E1["SCADA / grid telemetry"]
        E2["Smart meters<br/><i>interval consumption</i>"]
        E3["Asset condition<br/><i>DGA, thermal, partial discharge</i>"]
        E4["Distributed generation<br/><i>solar, storage, EV</i>"]
        E5["Weather & irradiance"]
        E6["Maintenance history"]
        E7["Customer & market"]
    end
    subgraph SC["Scopes"]
        direction TB
        S1["<b>asset.degradation.v4</b> — L3<br/>M1·M3·M5 · 24 mo<br/>Which loading, thermal and<br/>harmonic conjunctions<br/>accelerate asset ageing?"]
        S2["<b>grid.anomaly_cause.v3</b> — L2<br/>M1·M2·M3 · 7 d<br/>Fault vs. metering error vs.<br/>abstraction vs. behaviour"]
        S3["<b>flex.response_reality.v2</b> — L3<br/>M1·M3 · 12 mo<br/>Which flexibility signals<br/>actually produce response?"]
        S4["<b>safety.veg_risk.v2</b> — L2<br/>M1·M2·M5 · seasonal<br/>Vegetation, weather and<br/>circuit conjunction risk"]
    end
    SRC --> SC
    classDef a fill:#1b2f3d,stroke:#5b8fb0,color:#e6f2fa
    classDef b fill:#0a5570,stroke:#3fd0f0,stroke-width:2px,color:#ffffff
    class SRC,E1,E2,E3,E4,E5,E6,E7 a
    class SC,S1,S2,S3,S4 b

Figure 52. Energy correlation pattern (illustrative). Asset degradation correlation is a strong M3 candidate because the physical mechanisms are well understood and causal hypotheses are experimentally testable — as in manufacturing.

The Specific Constraint #

36A.4 Aviation #

The Invisible Phenomenon #

Aviation operations are a tightly coupled network in which small perturbations propagate non-linearly. An aircraft’s delayed arrival affects its next rotation, the crew’s duty limits, the connecting passengers, the gate allocation, the ground handling schedule, and — through crew and aircraft rotation — flights hours later and continents away.

Separately, predictive maintenance in aviation is exceptionally multi-factorial: component health correlates with flight cycles, route profiles, environmental exposure, operating technique, and maintenance history, across data held by the operator, the manufacturer, and the maintenance provider.

Correlation Pattern #

Table 74. Aviation correlation scopes (illustrative). Safety precursor correlation is capped at L1 — informational only — because aviation safety depends on a non-punitive reporting culture, and any perception that correlated data is used for enforcement would suppress the reporting the system depends on.

The Specific Constraint #

Additionally, airworthiness determinations are subject to regulatory frameworks requiring qualified human judgement. Predictive maintenance correlation informs the engineer; it does not determine the outcome.

36A.5 Cross-Sector Pattern Synthesis #

Ten sectors have now been examined. Five patterns recur across all of them:

Table 75. Cross-sector correlation pattern synthesis. Every scenario in Part IV reduces to one or more of these five patterns, which suggests that a reference implementation supporting them well would serve most enterprise correlation requirements — a hypothesis worth testing and proposed as future work in Chapter 40.

Two constraints also recur universally:

Autonomy is capped by irreversibility, not by confidence. In every sector, the actions the organization most wants to automate are the ones least eligible for automation — credit decisions, clinical interventions, benefit determinations, airworthiness calls, grid switching. The framework’s consistent answer is that these remain human, and the value is delivered by making the human’s decision better-informed and faster rather than by removing the human.

The correlation that is most valuable is frequently the one that is most sensitive. Staffing to clinical outcomes, operator to defect, safety report to individual, citizen to service history. In every case the framework’s answer is the same: the correlation should exist, its purpose must be bound computationally to a legitimate use, and the prohibited uses must be enforced by architecture rather than asserted in policy.

Next: 37. Comparison with Existing Technologies →