Project: TMSR — AI-Enhanced Incident Reporting Assistant

Tokio Marine — EY GDS Consulting (Mar 2025 – Present)

A Generative AI PoC that automates the analysis of insurance incident reports by comparing police reports against incident descriptions, extracting key information, and surfacing discrepancies — improving efficiency and auditability for claims analysts.

Business Problem Statement

Incident reports at Tokio Marine were processed manually, requiring significant analyst effort to extract relevant details, identify patterns, and assess risk. A particularly time-consuming step involved manually comparing the incident description provided by the insured against the police report to find inconsistencies and flag potential fraud indicators. The PoC aimed to automate these steps using Generative AI, improving efficiency and auditability at scale.

Solution & Key Capabilities

• Automatically reads and interprets police reports and incident descriptions using Gen AI.
• Extracts relevant information from incidents according to a structured requirements framework.
• Identifies inconsistencies, discrepancies, and gaps between the two documents.
• Flags individual alert points separately — both in the incident description and in the police report.
• Produces bilingual output reports in English and Portuguese for the Brazilian market.

Front-End Features

• Upload interface for Police Report and Incident Description documents.
• One-click automatic comparative analysis execution.
• Structured report output with categorized alerts — e.g., conflicting dates, inconsistent locations, missing information.
• Bilingual report delivery (English & Portuguese).

Scope

The PoC covers automated document comparison and report generation. Additional chatbot interaction for ad-hoc comparison queries between the two documents is out of scope for the Pilot phase, as it would require further customization.

Tools & Frameworks Used

• Generative AI — document parsing, information extraction, and comparison
• Large Language Models (LLMs) for bilingual report generation
• Python — backend processing and AI orchestration
• Document upload and processing pipeline
• AWS cloud infrastructure
• Front-end interface for document upload and report display