Business Problem:
- The client is a $4B IT Services provider & had an insurance product powering the Insurance lifecycle, including Policy Querying, FNoL & Claims Settlement.
- Using traditional ML models for document text extraction & workflow-based automation, made the process slow & less accurate, with Claims STP less than 12%, quite below the industry standards.
- Application built on legacy technology hosted in their DCs, with nonnative components, resulted in high hosting & application maintenance costs.
Solution:
- Team GoML designed a new application, leveraging LangChain & Claude-v2 for intelligent document extraction, built on native AWS services, including Bedrock & Textract
- The new application gave users capability to query policy documents with higher accuracy & have NLP powered interaction with Insurance documents, including intelligent analytics on insurance data, such as the highest no of premium filed within a region etc.
- The application also automated downstream workflows, like Claims Settlement & FNoL, post data extraction, increasing the STP Claims significantly
Architecture:
Technology Stack:
