This project automates document drafting in government offices, focusing on court orders. The current manual process is slow and prone to errors, especially given the repetitive nature of many government documents. The approach involves developing an agentic system that uses Al to generate JSON schemas from existing court orders, create questionnaires based on these schemas, and draft new documents using the collected data. The copilot asks questions based on the requirement of the type of document needed to generate. It checks if the answers are relevant enough and reframes questions for those which are not relevant. At the final stage when there are no more questions to be asked the copilot generates a draft of the final document. By automating these processes, the project seeks to improve efficiency, reduce human error, and maintain consistency with original document styles in government document drafting.