Insurance Claim Processing Agent

Agent Description:

The Insurance Claim Processing Agent is an AI-powered automation system built to streamline car insurance claim workflows. It ingests user-uploaded PDFs, validates claim data against policy records, and calculates the final payout amount using business logic and API validation.

Purpose and Components
  • Purpose: To automate the extraction, validation, and settlement of car accident claims using a PDF ingestion pipeline and real-time policy verification.
  • Components:
    • A document extractor for parsing user-submitted claim forms.
    • A validator that confirms policy activity and coverage for the accident.
    • A payout calculator that approves or rejects claims based on the input and estimates final reimbursement amounts.
Supported Capabilities
  • Parsing multi-page car insurance claim PDFs and converting them to structured JSON.
  • Validating policies and accident coverage against backend insurance databases using API calls.
  • Calculating payouts based on policy status and estimated damage values.
  • Generating a clear, user-facing summary of the claim decision and payout.
LLM Used
  • Google Vertex
    Note: To learn more about the LLM and to modify its behavior, refer to the Configuring LLM settings section.

Sub-Agents

1. Claim Extractor Agent

  • Role: Data Extractor.
  • Scope: Read multi-page insurance claim PDFs and extract essential claim details.
  • Description: This subagent accepts car insurance claim forms submitted in PDF format. It uses a PDF reader to extract structured data from multiple pages, including:
    • Name
    • Policy Number
    • Vehicle Number
    • Accident Date
    • Location
    • Email
    • Contact Number
    • Vehicle Model
    • Claim Type

    • Estimated Damage Amount

      The output is returned as a clean JSON object for validation by downstream agents.

  • Tool Used:PDF Reader Tool

2. Claim Validator Agent

  • Role: Claim Validator.
  • Scope: Use API data to confirm policy coverage and accident eligibility.
  • Description: This subagent receives structured claim data and validates it by querying the backend insurance system using a GET API tool. It checks:
    • Whether required fields are present (name, policy_number, accident_date, estimated_damage_amount)
    • Whether the policy is active
    • If the accident date falls within the policy coverage period
    • If the claim type is covered under the policy

    Based on these validations, it either approves or rejects the claim for payout processing.

  • Tool Used: GET Request Tool

3. Payout Calculator Agent

  • Role: Payout Calculator.
  • Scope: Compute and generate a response based on claim eligibility and damage estimate.
  • Description: This subagent takes the validated output from the Claim Validator Agent. If the claim is eligible, it calculates the payout amount using predefined business rules and returns:
    • The final status of the claim (Approved or Rejected).
    • The calculated payout amount.
    • A justification summary in plain text, formatted for direct user communication

      If the claim is denied, it includes an appropriate explanation for rejection.

Tools Used
  • PDF Reader Tool – For extracting structured data from user-uploaded documents.
  • GET Request Tool – For validating claim details against internal policy databases.
Note: For details on modifying the Tools, refer Tools Library section.
Agent Workflow Behavior Summary
  • The Claim Extractor Agent reads the uploaded car insurance PDF and extracts structured claim fields into JSON.
  • The Claim Validator Agent validates the claim using API data, checking policy activity, accident eligibility, and required fields.
  • The Payout Calculator Agent reviews validated inputs, computes the payout amount if applicable, and generates a plain-language summary of the decision for the claimant.
Sample Questions:
  • Read the content of this path<pdf link> and tell me if I am eligible for the claim insurance