About
Faced with 16 TB of unstructured emails, PDFs and policy documents, a leading broker turned to Firemind’s AI in insurance operations workflow with AWS Lambda for file classification, Amazon Kendra for natural‑language search and Amazon Bedrock for retrieval‑augmented generation, which delivered 9/10 accurate answers and cut search time from hours to seconds.
Challenge
The client had accumulated over 16 terabytes of data in its OpenText system. The information - a mix of emails, PDFs, spreadsheets, and policy documents - was largely unstructured and carried little metadata. Employees struggled to search, analyse, or interpret the content effectively. As a result, important information was overlooked, and decision-making was slower and less consistent than required.
Firemind developed a Retrieval Augmented Generation (RAG) solution to make the client’s data practical and usable. A custom AWS Lambda process identified and categorised files, while Amazon Kendra indexed the content so staff could query it. Using large language models through Amazon Bedrock, employees were able to ask questions in plain English and receive concise, accurate answers drawn directly from the documents.
Services used
-
Amazon Kendra – indexing and querying of unstructured data
-
Amazon Bedrock – Anthropic Claude models for summarisation and responses
-
AWS Lambda – file identification and preprocessing
-
Amazon S3 – scalable storage and retrieval
Results
- Accuracy of 9/10 in responses to claim file queries
- Manual search time reduced from hours to seconds, improving productivity across underwriting and claims teams
- Consistent answers across departments, lowering the risk of missed information
- Clear path to scale with further integration into OpenText and additional use cases identified
