Amazon Finance’s AI Assistant: Revolutionizing Data Discovery

Amazon Finance tackled challenges in financial planning and analysis by developing an AI-powered assistant using Amazon Bedrock and Amazon Kendra. Analysts previously struggled with time-consuming manual data searches across disparate systems and a lack of institutional knowledge preservation. This new solution leverages Retrieval Augmented Generation (RAG) combining intelligent retrieval via vector stores in Amazon Kendra with augmented generation using Anthropic's Claude 3 on Amazon Bedrock. The system allows analysts to use natural language queries to access financial data and documentation, significantly improving search speed and accuracy. Key features include semantic search, contextual response generation, and robust enterprise security. The architecture uses Streamlit for the user interface, Lambda for authentication, and Fargate for containerized deployment. Evaluation results showed a 30% reduction in search time, an 80% improvement in accuracy, and 85% increased efficiency. Analysts now can quickly locate data sources, understand internal processes, and access relevant information, leading to improved decision-making and productivity. The solution's success highlights the potential of AWS generative AI services for enhancing enterprise data discovery and knowledge management.

(Source: https://aws.amazon.com/blogs/machine-learning/how-amazon-finance-built-an-ai-assistant-using-amazon-bedrock-and-amazon-kendra-to-support-analysts-for-data-discovery-and-business-insights/)