RAG aplicado a business intelligence para la gestión eficiente de insights dentro de una institución financiera ecuatoriana

Abstract

The project "RAG applied to business intelligence for the efficient management of insights in an Ecuadorian financial institution" aims to develop an AI system integrating Retrieval-Augmented Generation (RAG) with Business Intelligence tools. The goal is to enable commercial users to perform natural language queries through a self-service portal, reducing response times and workload for the analytics team. The solution addresses the need to improve information accessibility and decision-making processes. The system uses a DuckDB backend and OpenAI models to translate queries into SQL statements and present the results through a Streamlit interface. Empirical testing was conducted to fine-tune system performance. The results show a 98% reduction in response times compared to the traditional process and a 78% satisfaction rate among users, validating the solution's effectiveness. The project concludes that combining RAG with BI significantly enhances data accessibility and operational efficiency. Keywords: Retrieval-Augmented Generation, Business Intelligence, Large Language Models, Data Management.