You are building your first customer-facing AI Agent and now you are running into CX specific issues? Deploying an AI Agent that talks directly to your end customer is fundamentally different from building one that supports your live agents behind the scenes. It requires different prompting, different evaluation, and a different relationship with your assistant configuration than fulfiller use cases ever demanded. In this CRM AI Academy session, you will learn the six step framework for building requestor assistants that hold up under real, unpredictable customer conversations. In this video, Fernando Castro, Sr Product Manager for CRM AI, talks about: Why requestor AI Agents behave nothing like fulfiller ones, and the four consequences that follow from facing an unmoderated customer How to scope your assistant and split responsibilities so the orchestrator routes conversations in an accurate and scalable manner Why step-by-step prompting breaks in live customer conversations How to build a two layer evaluation plan for this use case