Human-in-the-Loop

A setup where an AI system stops at defined points so a person can approve, correct or reject its output before it takes effect. In outbound, a human typically reviews the target list, the first batch of AI-written emails and every reply that is not a plain no. Put checkpoints where a mistake is expensive or public.