How escalation works
Support teams usually describe two types. Functional escalation moves an issue to someone with the right skills, like a technician. Hierarchical escalation moves it up to someone with more authority, like a manager who can approve a refund. Urgent escalation adds speed: someone is alerted now, not at the next shift.
Escalation rules work best written down: which situations escalate, to whom, by what route, and what the customer is told in the meantime. For an AI assistant, those rules decide when it stops trying to help and brings a person in, with the details already collected.
Example: an upset customer at an auto shop
A customer emails an auto repair shop that his bill is higher than the estimate and he wants it fixed today. The assistant recognizes an upset customer and a payment dispute, doesn’t argue or promise a refund, tells him the manager will review the invoice, and flags the conversation for the owner with the estimate and invoice details noted. This scenario is illustrative.
Escalation in InsertChat
InsertChat escalates on its own rules and yours. Urgent alerts fire for a caller in distress, a safety risk, or a serious service failure; the assistant reads the safety steps you wrote and collects details for whoever responds. Upset customers, with threats, severe frustration, or repeated complaints, are brought to a person by default.
Emails about complaints, payments, or legal matters are held for your team even when automatic replies are on. Escalated conversations are marked needs-you and chosen teammates are emailed. See after-hours answering for urgent calls and the auto repair setup.
