Glossary
Human Handoff
The transfer of a call from an AI voice agent to a human agent — ideally with full context, so the human starts the conversation already informed.
Human handoff (also called escalation, warm transfer, or human-in-the-loop) is the critical design pattern that separates "AI you’d let your customers talk to" from "AI you wouldn’t". A well-designed handoff includes: (1) trigger conditions in plain English (customer asks for human, emotional cue detected, intent unknown, business rule met), (2) routing logic that selects the right human (skill, language, account tier), (3) a 5-second AI-generated briefing read to the human as the call connects, and (4) full transcript and structured data already in the agent’s desktop. Cold transfers ("please hold while I transfer you to someone who can help") are the failure mode — they degrade CSAT and waste human time.
Why it matters
- Customers who ask for a human get one in <5 seconds — better than queue-and-hold.
- Human agents start the conversation already briefed — average handle time drops 30–50%.
- Maintains trust in AI deployments — "ask for a human and you get one" is the trust contract.
- Routes the right human (skill, language, tier) on the first try — eliminates transfer chains.
- Allows AI to handle ambiguity gracefully ("let me get a specialist on the line") vs forcing wrong answers.