When Should an AI Agent Escalate to a Human?

AI agents should escalate when confidence, context, authority, emotion, or user value crosses a risk threshold. Here is a practical escalation framework.