Cloud-Scale Business Intelligence Using Generative AI and Data Mesh Architecture

Abstract

Cloud-scale Business Intelligence (BI) increasingly requires organizations to integrate heterogeneous data sources, decentralized data ownership, intelligent analytics, and secure self-service access. This research proposes a novel Generative AI–Data Mesh Business Intelligence (GAI-DM-BI) framework that integrates domain-oriented data products, a self-service Data Mesh platform, semantic knowledge representation, Retrieval-Augmented Generation (RAG), Large Language Models (LLMs), Text-to-SQL generation, automated query validation, and federated governance into a unified cloud-scale architecture. The proposed framework transforms natural-language business questions into governed analytical queries through a multi-stage pipeline involving intent classification, metadata retrieval, semantic grounding, LLM-based SQL generation, validation, secure execution, and insight generation. Domain-specific data products provide structured and governed access to Customer, Sales, Finance, Marketing, Operations, and Supply Chain information, while the semantic layer connects business concepts with technical schemas and KPI definitions. Experimental evaluation is performed using 5,000 benchmark analytical queries across heterogeneous enterprise data products and compares the proposed framework with centralized BI, Cloud Data Lake BI, Data Mesh BI, and conventional LLM+RAG BI approaches.