How human handoff works
Handoff has three parts. A trigger decides when it happens: an explicit request, low confidence, repeated failure, negative sentiment, or a listed sensitive topic. A notification tells the right person: an inbox flag, an email, or a push alert. And a takeover lets that person continue with the history, a summary, and any details the assistant collected.
The most common mistake is a handoff to nobody: the bot says “connecting you to an agent” when no one is available. Outside staffed hours, the honest version is collecting details and promising a follow-up by a specific route.
Example: a law firm intake call
A caller to a law firm describes a car accident, then asks whether she should accept the insurer’s offer. That is legal advice, which the firm listed as always needing a person. The assistant says an attorney will review it, collects the accident date and her contact details, and the firm sees the conversation flagged with a summary. This scenario is illustrative.
Human handoff in InsertChat
InsertChat hands a conversation to your team when a customer asks for a person, the assistant is unsure of an answer, a customer says twice that an answer didn’t help, a customer is upset, or a request needs an action outside its permissions. Add topics that always need a person and the message customers see while a person is asked to help. The conversation is marked needs-you and the teammates you choose are notified.
On website chat, a teammate can take over in the same window and reply in real time while the AI steps back. On calls, InsertChat does not transfer live; the assistant collects details and your team follows up. Set it up with how to set up human handoff, or see the law firm setup.
