An AI agent shouldn't try to handle everything. Escalation rules define when it should step back and hand the conversation to a real person instead.
When escalation happens
No answer – the AI can't find a relevant answer for the customer's question.
Frustration – the AI detects the customer is getting frustrated with the conversation.
Explicit request – the customer directly asks to speak to a human.
Where escalated conversations go
You choose the target inbox escalated conversations land in, so they reach the right team instead of a general queue nobody's watching closely.
Controlling what the AI can see
Before it even gets to escalation, you can control what data the AI is allowed to access while it's handling a conversation, including conversation history, contact details, and the customer's location. This is configured in the Access tab, alongside Training and Escalation.