Development of a retrieval-augmented generation framework for automated proposal generation through Azure cloud

Abstract

Artificial intelligence (AI) has been increasingly adopted to improve knowledge retrieval and decision-making in enterprise environments; however, proposal preparation in the pre-sales stage still relies heavily on manual searches across fragmented document repositories and heterogeneous file formats. This study addresses this gap by proposing a retrieval-augmented generation (RAG)-based system that automates the retrieval, summarization, and generation of proposal-related information from both internal repositories, such as OneDrive and SharePoint, and external web sources. The proposed system integrates Azure Blob Storage, Azure AI Search, Azure OpenAI, and Bing Search within a RAG framework, supported by a web-based interface developed using React/Next.js. Unlike conventional keyword-based search tools, the system interprets user intent and delivers consolidated, relevant information to support proposal drafting. Experimental evaluation in a pre-sales use case demonstrates a reduction in manual information retrieval effort and improved content relevance, while achieving an average response generation time of 102 milliseconds, enabling real-time interaction. Overall, the findings demonstrate how secure, enterprise-grade cloud integration and RAG-based conversational systems can transform pre-sales workflows by allowing professionals to shift their focus from manual information gathering to higher-value strategic content development.