The emergence of coding agents such as Claude Code, Cursor, and others signifies a transformative shift in software development by allowing natural language descriptions to automate technical tasks, thereby facilitating "low-code IT" for business users. However, these agents face significant limitations in understanding documents, a critical component of enterprise applications, as most enterprise data is embedded within complex document formats like PDFs and DOCXs. Enhancing these agents with document understanding tools can improve their ability to build applications that meet business requirements more accurately and resiliently. The text discusses various strategies to equip agents with document intelligence, such as integrating with the Model Context Protocol (MCP), utilizing command-line interfaces (CLI) for advanced document operations, and developing AI-native workflows for better application adaptability and effectiveness. The ultimate goal is to enable coding agents to bridge the gap between business needs and technical execution by understanding and processing enterprise documents, offering a robust foundation for document-centric workflows.