Good ERDs re-motivate the problem at a high level, then point back to specific sections of the Product Requirements Doc - PRD to motivate portions of the work. In the traditional SDLC, pointing to the "why" provided critical context to the team doing the actual implementation. It allowed them to deviate from the ERD's plan more easily if a better approach to solving the underlying problem became clear while doing the implementation. I believe that context is equally important for agentic implementation.