Automated Underwriting Pipeline
AI ↔ AIApplication 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.
↓sequential→ Risk Scoring Agent
Risk Scoring Agentagent
Compute risk score from applicant profile, claims history, and credit data.
↓sequential→ Actuarial Model Agent
↓conditional→ Underwriter Review
Actuarial Model Agentagent
Apply loss ratio models and mortality/morbidity tables to compute expected loss.
↓sequential→ Policy Pricing Agent
Policy Pricing Agentagent
Set premium based on risk score, expected loss, and competitive positioning.
↓conditional→ Underwriter 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"