Handoff design
Human handoff rules for AI customer support
The handoff rules that keep AI support helpful, honest, and easy for teams to operate.
Route high judgement work immediately
The handoff list should be explicit before the agent goes live. Refund approvals, cancellations, legal or medical questions, account changes, threats, complaints, and policy exceptions should move to a person.
This keeps the customer experience honest. The agent can explain that the team will review the request instead of inventing an outcome.
Make the handoff useful to the team
A handoff is not just a notification. It should include who the customer is, what they asked, what the agent already answered, what is still unresolved, and what action the team should take next.
That context is what turns automation from another inbox into a support workflow.
Use handoffs to improve the agent
Every handoff is a signal. Some show missing knowledge. Some show unclear policy. Some show a workflow that may eventually be safe to connect.
Reviewing those patterns is how the agent becomes more useful without giving it more authority than the business is ready to support.
FAQ
Frequently asked questions
When should an AI chatbot hand off to a person?
Handoff should happen when the customer asks for it, the answer is uncertain, identity or account access matters, money or policy judgement is involved, or the conversation shows risk or strong dissatisfaction.
What information should an AI handoff include?
Include the customer contact, a concise issue summary, the transcript, sources already used, what remains unresolved, urgency, and the recommended next action.