Foxintelligence scales email data extraction with LLMs, replacing up to 6,000 manual parsers

Foxintelligence relied heavily on manual processes to process and extract insights from unstructured customer data, such as email communications. This approach was time-consuming, resource-intensive, and limited the company’s ability to scale operations without hiring additional staff. 

To remain competitive, Foxintelligence needed a solution that could automate email data extraction, improve processing accuracy, and handle increasing data volumes while maintaining strict privacy and security standards. The goal was to free up its team to focus on higher-value, strategic tasks rather than repetitive data handling. 

A proof-of-concept large language model (LLM) solution was developed using AWS services to automate and scale unstructured data processing. Amazon S3 was used to securely ingest and store customer emails, while Amazon Kendra enabled intelligent search and retrieval of the most relevant documents. Amazon Bedrock hosted Claude V2, which provided contextual reasoning and accurate extraction of key information. 

The system used a fully serverless architecture with AWS Lambda and Amazon API Gateway for backend processing, ensuring scalability and cost efficiency. A secure, prompt-based user interface allowed Foxintelligence’s team to easily interact with the model, retrieve insights, and customise queries - streamlining workflows and dramatically increasing productivity. 

Services used