What Is an Escalation Rule and When Does It Trigger for MeUpdated 5 days ago

An escalation rule is an automated condition that routes a specific type of incoming message directly to human review — bypassing the AI auto-response system to ensure sensitive, complex or high-stakes messages are handled by a person rather than automation.

Escalation rules are not a fallback for when AI fails. They are a deliberate design decision — specific categories where automation is actively inappropriate regardless of how well the AI could technically respond.

What I configure escalation rules for during the support setup:

High-value refund requests — any message requesting a refund above a defined threshold is flagged for human review. These require discretion and context that standard policy answers cannot provide.

Emotionally charged messages — the system detects language indicating frustration, distress or anger. Messages matching these signals are routed to human review rather than receiving an automated response that might escalate the situation.

Messages about ongoing issues:

• A buyer who has already submitted a complaint and is following up
• A buyer whose message references a previous unresolved interaction
• Messages requiring reading a full conversation history before responding accurately

Unrecognized question categories — any message the AI cannot confidently categorize is automatically escalated rather than attempting a response that might be inaccurate.

I configure the escalation threshold for each rule based on the store's specific policies.