Automated Underwriting Pipeline

AI ↔ AI

Application flows through risk scoring, actuarial modeling, and pricing with human review for edge cases.

5 nodes · 5 edgesinsurance
agenthumansystem
Visual
Application Intake Agentagent

Parse application form, validate fields, enrich with third-party data.

sequentialRisk Scoring Agent
Risk Scoring Agentagent

Compute risk score from applicant profile, claims history, and credit data.

sequentialActuarial Model Agent
conditionalUnderwriter Review
Actuarial Model Agentagent

Apply loss ratio models and mortality/morbidity tables to compute expected loss.

sequentialPolicy Pricing Agent
Policy Pricing Agentagent

Set premium based on risk score, expected loss, and competitive positioning.

conditionalUnderwriter Review
Underwriter Reviewhuman

Senior underwriter reviews edge cases where risk score is borderline.

uc-underwriting-pipeline.osop.yaml
osop_version: "1.0"
id: "underwriting-pipeline"
name: "Automated Underwriting Pipeline"
description: "Application flows through risk scoring, actuarial modeling, and pricing with human review for edge cases."

nodes:
  - id: "intake"
    type: "agent"
    subtype: "worker"
    name: "Application Intake Agent"
    description: "Parse application form, validate fields, enrich with third-party data."

  - id: "risk_score"
    type: "agent"
    subtype: "llm"
    name: "Risk Scoring Agent"
    description: "Compute risk score from applicant profile, claims history, and credit data."

  - id: "actuarial_model"
    type: "agent"
    subtype: "llm"
    name: "Actuarial Model Agent"
    description: "Apply loss ratio models and mortality/morbidity tables to compute expected loss."

  - id: "policy_pricing"
    type: "agent"
    subtype: "llm"
    name: "Policy Pricing Agent"
    description: "Set premium based on risk score, expected loss, and competitive positioning."

  - id: "human_review"
    type: "human"
    subtype: "review"
    name: "Underwriter Review"
    description: "Senior underwriter reviews edge cases where risk score is borderline."
    security:
      approval_gate: true

edges:
  - from: "intake"
    to: "risk_score"
    mode: "sequential"
  - from: "risk_score"
    to: "actuarial_model"
    mode: "sequential"
  - from: "actuarial_model"
    to: "policy_pricing"
    mode: "sequential"
  - from: "policy_pricing"
    to: "human_review"
    mode: "conditional"
    when: "risk_score.category == 'borderline'"
  - from: "risk_score"
    to: "human_review"
    mode: "conditional"
    when: "risk_score.flags.length > 0"
    label: "Flagged applications skip to human"