Escalation
What is escalation and when should a conversation be handed up?
Escalating is handing a conversation to someone who's allowed more or knows more. The rules for it are usually written nowhere, and you notice that only in the conversations that go wrong.
In short
Escalation is handing a customer conversation to an agent with more authority, more knowledge or more time, because the current handler can't complete the question. So escalation is about authority, not workload: passing a conversation because someone is busy is distribution, not escalation.
The three reasons to escalate
They require different things and often get lumped together, which leaves a single route to the one person who has to do everything.
- Authority
- The customer asks for something beyond the agent's mandate: a goodwill amount, an exception to policy, a credit above a limit. This needs a clear limit and a fast route, because the customer is waiting.
- Knowledge
- The question is technical or specialist. The right action is usually not handover but consultation: the specialist supplies the answer, the original agent keeps the conversation.
- Tone
- The conversation has stalled on the relationship rather than the content. A second person with a fresh view resolves that more often than the same person explaining once more.
Why escalation without a handover costs more
The most expensive form of escalation is the one where the customer retells their story. That's not only lost time, it's the moment the customer decides your organisation doesn't talk to itself. A conversation that reaches that point takes structurally more effort afterwards.
What makes a handover usable is three lines from the previous handler: what the customer wants, what's already been tried and what's been promised. That costs twenty seconds and saves the new handler a five-minute conversation.
Where do you draw the line on a mandate?
A mandate that's too tight turns every exception into an escalation and every escalation into a wait. A mandate that's too wide gives you a customer who gets a different answer from the third agent than from the first.
A usable rule of thumb is picking an amount where the handling time of the escalation outweighs the amount itself. If an escalation costs fifteen minutes of two people's time, escalating a ten-euro goodwill gesture costs more than simply granting it.
When does AI hand over?
With an AI agent escalation isn't an exception but a setting. What you define are the topics where it never answers alone: money, complaints, legal questions and anything asking for an exception to your policy.
There should also be a signal on uncertainty. An agent that can't find the answer in its sources should hand over rather than improvise. Measure how often that happens, because that list is also your list of missing knowledge base articles.
Where it goes wrong
Escalation as a bin for difficult conversations
If everything complex goes to the same person, within a month that person is the bottleneck and nobody else knows how those conversations work. Split escalation by reason and distribute the authority.
No timer on an escalated conversation
A conversation that's been handed up often disappears from the response time measurement because it changed owner. The customer keeps waiting all the same. Let the timer run across the whole chain.
Not telling the customer there's a handover
Silence during an escalation reads as no action. A message saying who's taking over and when an answer comes prevents the chaser email you'd otherwise get.
Never reviewing what got escalated
A month's escalation list is the best training agenda you'll get. If the same topic keeps returning, it isn't an escalation but a gap in the mandate or in knowledge.
What to do this week
Take last month's conversations that changed owner and note the reason for each: authority, knowledge or tone. If you mainly find authority on small amounts, your mandate is too tight and escalating costs more than the amount. If you mainly find knowledge on the same topic, an article is missing from your knowledge base.
What this looks like in a system
In Cuego the timer keeps running when a conversation changes owner, so the customer's waiting time stays visible across the whole chain. History travels with the handover and the previous handler can leave an internal note the customer doesn't see. For the AI agent you define per topic where it hands over, and conversations handed over on uncertainty land in a separate list showing you which knowledge is missing.
Cuego AIRelated terms
First Contact Resolution
First Contact Resolution is the percentage of questions finished in one go. Of all customer service numbers, this is the one that predicts the most work.
Read onTicket routing
Ticket routing determines which question lands with which agent. Set up poorly, it costs you the first twenty minutes of every morning.
Read onQueue
A queue is the list of conversations waiting on someone. How you sort that list determines which customers you systematically help too late.
Read on