A retail company wants to deploy a Copilot Studio agent that answers employee questions about product return policies by grounding on documents in a SharePoint library. During requirements analysis, the architect finds the library contains policy PDFs from 2018 to 2024, several duplicate versions with conflicting return windows, marketing brochures mixed among policies, and many scanned image-only files without extracted text. Which data-readiness issue should the architect prioritize remediating FIRST to reduce the risk of the agent returning contradictory answers?