Public sector agencies struggle to be agile due to legacy data hindering access to critical information. The Indiana Department of Transportation (INDOT) faced this when required to produce a government efficiency report within 30 days. This report needed to map statutory responsibilities to core purposes, but vital information was scattered across various internal documents. A manual review was estimated to take hundreds of hours, making the deadline seemingly impossible.INDOT chose to develop an AI-powered solution using their existing Google Cloud environment and Gemini's capabilities. Their team built an agile, week-long pilot program utilizing Retrieval-Augmented Generation (RAG). This process involved ingesting and transforming data using Python scripts and then indexing it with Vertex AI Search to create a searchable knowledge base.The RAG engine in Vertex AI efficiently retrieved relevant document snippets, which were then passed to Gemini for context-aware responses. This ensured that the AI's answers were based solely on INDOT's official documents, not external internet data. The pilot successfully generated draft reports for nine divisions with 98% fidelity, saving an estimated 360 hours of manual effort.This AI-powered approach enabled INDOT to be the largest state agency to submit its efficiency report on time. The AI-generated report framework became the official template for 60 other state agencies, demonstrating responsible AI use and building trust. INDOT's scalable and secure RAG system on Google Cloud created a reusable model for future innovation, accelerating their mission to better serve Indiana.