Escalation

Escalation is handing a conversation to somebody with more knowledge or more permissions to resolve it. It can go from one support tier to another, from an automated agent to a person, or from support to engineering — and in every case whoever receives it has not seen the case until that moment.

What is escalation?

Escalation is handing a conversation to somebody with more knowledge or more permissions to resolve it. It can go from one support tier to another, from an automated agent to a person, or from support to engineering — and in every case whoever receives it has not seen the case until that moment.

Also: handoff · hand-off · transfer

Why it matters

What changes in a SaaS

It is the most expensive operation in support and the one done worst. Expensive because it doubles the time spent and usually interrupts somebody more expensive; done badly because the context almost never travels with the conversation, so the customer explains the problem again and whoever picks it up repeats the diagnosis that had already been made.

Una mano alcanzando un teléfono de disco sobre una mesa

Escalation in detail

  • What has to travel with the conversation

    What the customer asked, what has already been checked, what has been ruled out, and why it stopped. Without those four things the handoff is not an escalation: it is starting from zero with a customer who is already waiting.

  • Escalating early is not failing

    An agent wrestling for half an hour with something outside their remit costs more than one who escalates in two minutes. The mark of a bad escalation is not speed, it is arriving with no context or arriving at the wrong person.

  • Escalation from an AI agent

    It is the least flashy part of an automated agent and the one that shows most. A system that always answers, even when it does not know, leaves the person with an angry customer and a conversation that has to be undone. One that recognises its limit and hands over with the case assembled saves time instead of creating it.

  • What to measure

    The percentage of conversations escalated, where they go, and how many come back down a tier. A high return rate means the wrong things are being escalated; a high rate of escalations to engineering means there is a product problem.

Questions about escalation

When should a ticket be escalated?

When resolving it needs knowledge or permissions the person handling it does not have, or when the customer asks. Setting a maximum time before escalating stops a case circling at the wrong tier.

What information should an escalation include?

The original question, what has already been checked, what has been ruled out, and the reason for the handoff. Without that, whoever receives it repeats work already done.

How does an AI agent escalate a conversation?

It should hand over with the case assembled: the history, the data it looked up and why it stopped. A handoff that only says "let me get a person" is worse than not having intervened.

Related terms

A term on its own is only half understood. These come up in the same conversation.

Try it without dropping what you use