The Reasoning Pipeline in Action

To demonstrate the power of Neurosymbolic AI, we will trace a single complex trauma claim—a Heart Attack benefit—through our a three-stage pipeline: from raw policy text to a deterministic payment verdict.

Truth Layer

Reasoning Layer

Decision Layer

STAGE 01

The Truth Layer: Verifiable Extraction

In trauma insurance, a 'Heart Attack' is not just a diagnosis; it is a strict clinical definition. We map this definition to a Knowledge Graph to ensure zero ambiguity.

By anchoring the clinical definition as a set of required nodes, we eliminate the risk of the AI 'assuming' a heart attack occurred based on the word alone.

Source Policy Text

'Heart Attack' means an acute myocardial infarction evidenced by the elevation of cardiac troponin in combination with acute ischemic ECG changes.

Knowledge Graph Triplet

Heart Attack

requires

Troponin Elevation

flowchart TD %% Node Definitions %% AND-group Causes (Light Green) E1(["CAUSE_1 Var: the diagnosis of a heart attack Cond: is present"]) E2(["CAUSE_2 Var: cardiac biomarkers Cond: Typical rise and/or fall of cardiac biomarkers with at least one value above the 99th percentile of the upper reference range"]) E6(["CAUSE_7 Var: the above tests Cond: are inconclusive"]) %% OR-group Causes (Mid Green) E3(["CAUSE_3 Var: signs and symptoms of ischemia Cond: consistent with a heart attack"]) E4(["CAUSE_5 Var: pathological Q waves in the ECG Cond: development of"]) E5(["CAUSE_6 Var: imaging evidence Cond: of new loss of viable myocardium or new regional wall motion abnormality"]) %% Logic Gates I0{"AND (Prec: False)"} I2{"OR (Prec: False)"} I4{"AND (Prec: False)"} %% Effects E0[("EFFECT_1 Var: we will pay a claim we Cond: will pay")] E7[("EFFECT_2 Var: inconclusive tests Cond: will consider other appropriate and medically recognised tests")] %% Layout and Relationships E1 --> I0 E2 --> I0 E3 --> I2 E4 --> I2 E5 --> I2 I0 --> I2 I2 ==> E0 I2 --> I4 I4 --> E6 E6 ==> E7 %% Styling style E0 fill:#43D585,stroke:#09110C,stroke-width:4px,color:#09110C style E7 fill:#43D585,stroke:#09110C,stroke-width:4px,color:#09110C style I0 fill:#F5F6F2,stroke:#1B2A22,stroke-width:2px,color:#09110C style I2 fill:#F5F6F2,stroke:#1B2A22,stroke-width:2px,color:#09110C style I4 fill:#F5F6F2,stroke:#1B2A22,stroke-width:2px,color:#09110C %% Cause Colors: Light for AND-group, Mid Green for OR-group style E1 fill:#EAF3EC,stroke:#1B2A22 style E2 fill:#EAF3EC,stroke:#1B2A22 style E6 fill:#EAF3EC,stroke:#1B2A22 style E3 fill:#B8D8B8,stroke:#43D585 style E4 fill:#B8D8B8,stroke:#43D585 style E5 fill:#B8D8B8,stroke:#43D585

STAGE 02

The Reasoning Layer: Causal Intelligence

We don't just search for 'Heart Attack' in the medical records. We build a causal chain that MUST be satisfied for a payout to occur.

Requirement: Cardiac Troponin Elevation

Requirement: Ischemic ECG Changes

Dependency: Both conditions must be concurrent

Outcome: Deterministic 'Yes' or 'No'

Our neurosymbolic engine traverses this graph. If the troponin is elevated but the ECG is normal, the path is broken. The result is not a 'maybe'—it is a deterministic 'No'.

STAGE 03

The Decision Layer: Comparative Certainty

The final output is a verifiable verdict. For a broker, this means the ability to spot a gap in coverage instantly between two different policies.

Policy A

Requires Troponin elevation

Requires ECG changes

Requires Hospital Admission

Policy B

Requires Troponin elevation

Requires ECG changes

No admission required

By comparing the causal architecture, the broker can tell the client: 'Policy B is superior because it removes the Hospital Admission requirement for heart attack claims.'

Verdict: Eligible

The causal path is complete.

la Troponin: 0.85 ng/mL ( > 0.04)

la ECG: ST-segment elevation found

Stop guessing. Start knowing.

Experience the power of deterministic AI for your regulated documents.

Book a demo