Human handoff
Quick definition
Human handoff is the moment an AI assistant or chatbot passes a conversation, with its context, to a human agent. It happens when a request needs a person or the customer asks for one.
On this page
When an AI should hand over
Typical triggers are: the customer explicitly asks for a person; the request is outside what the assistant knows or is allowed to do; the stakes are high (a complaint, a refund dispute, a safety or legal question); the customer is upset; or the assistant has failed to help after a few exchanges. Many businesses also route specific topics — B2B quotes, large orders — straight to a person.
The decision can be made by rules, by the AI itself following instructions, or by both. What matters is that it happens before the customer gives up.
What a good handoff includes
The person who takes over should not have to ask the customer to repeat everything. A good handoff carries the full conversation, a short summary of what the customer wants and what has been tried, and whatever is known about the customer. The team needs to be notified, and the AI must stop answering in parallel so the customer is not talking to two voices at once.
The customer, in turn, should be told what happens next — that someone has been notified and roughly when to expect an answer — rather than left in silence. Outside opening hours, that means saying honestly that the reply will come later.
Common mistakes
The worst is no exit: a bot that keeps answering “I can help you with that!” when it cannot. Close behind are handoffs that lose context, handoffs that promise a human who never comes, and an AI that claims it has escalated when nothing was actually sent. Track how many handoffs are requested, how long they wait and how they are resolved — those numbers show both the limits of the assistant and the load on the team.
Human handoff in AI SmartTalk
In AI SmartTalk, handoff is configured per assistant with instructions on when to escalate. When it triggers, the conversation is flagged in the shared inbox under a “Human needed” filter, oldest first, with a summary that each team member reads in their own language, and the team is notified by email, push and in-app alerts; an unanswered request sends one reminder after 30 minutes. The AI can be paused while a person replies and given the conversation back afterwards. By default the assistant hands over when the visitor explicitly asks for a human.
Frequently asked questions
What is the difference between handoff and escalation?
They are often used interchangeably. Escalation can also mean moving a case to a more senior or specialised team; handoff usually refers to the switch from an automated assistant to a person.
Should customers always be able to reach a human?
For customer service, yes, even if the reply comes later. Customers accept automation far more readily when they know a person is reachable.
How many conversations should go to a human?
There is no universal figure. It depends on how complex your requests are and how complete the knowledge base is. Watch the reasons for handoff: recurring ones point to content or tools to add.
Can the AI keep helping after a handoff?
It can, but it is usually better to pause it while a person handles the conversation, then hand control back once the issue is resolved.