Designing the handoff from agent to human
When an agent should hand off, what must travel with it, and why a zero handoff rate is a bad sign rather than a good one.
· 5 min read
What travels with the handoff
The worst handoff is the one where the human opens with "how can I help?" after the customer has already explained twice. What must arrive: the full transcript, the reason for the handoff, what the agent already attempted, and any data it extracted.
Waiting is worse than refusal
If no human is available, say so immediately and offer an alternative: a callback slot, an email. Leaving the customer waiting for a reply that never comes is the worst possible outcome, worse than saying "I cannot help with this".
Frequently asked
- Should I tell the customer they are talking to a machine?
- Yes. Disclosure is legally required in several jurisdictions, and more practically it raises satisfaction: a customer who knows they are talking to an agent phrases questions more clearly and resents its limits less.
Read next
Running an AI agent on WhatsApp Business
What running an agent on WhatsApp actually requires: the account, the 24-hour window, templates, and human handoff.
How to build an Arabic AI agent for customer service
A practical guide to an agent that understands Modern Standard Arabic and dialects, answers only from your knowledge, and hands off to a human when needed.
A security checklist before launching an AI agent
What to verify before your agent talks to a real customer: prompt injection, tenant isolation, data leakage, and audit.
