A Practical Control Model for AI Agents in a CMS

How to scope permissions, assign responsibilities, approve actions, and verify outcomes when an AI agent works with a CMS.