CCA-F start here · Trap gym · Who escalates?

Who escalates?

A support agent hands the wrong cases to people: angry customers with easy problems go up, calm customers with hard ones stay. The tempting answers swap in a signal that seems objective: a mood score, the model's own confidence, a trained classifier. Each is an unreliable stand-in for what matters. The exam's answer is to say plainly when to escalate, and show a few examples.

D5 Context & ReliabilityD1 Agentic ArchitectureExam guide 5.2 and 1.4About 15 minutes

At a glance: Escalate when the customer asks for a person, when there's a policy exception or gap, or when the agent can't make progress; never on sentiment or self-rated confidence.

Start here: the new shop assistant

A shop hires a new assistant and says: “call the manager when you need to.” A week later:

What went wrong. They call the manager whenever a customer raises their voice, even about a $5 return. A calm customer asks for a refund the shop never gives, and the assistant quietly says yes.
What fixes it. “Call me for anything over $150, anything the refund policy doesn't cover, or anyone who asks for the manager. Everything else, you handle.” Then you walk through four real cases together.

You could give the assistant a mood meter, or ask “how sure do you feel?” before each sale, or build a system that learns from last year's calls. Each one measures something near the real question. None of them tells the assistant where the line is. And when they do call the manager, “someone needs help” is not enough: the manager needs who, what went wrong, how much, and what to do.

Separate classifier trained on old tickets System prompt criteria + examples Customer message the ticket Claude the support agent Self-rated confidence “I'm 0.8 sure” Sentiment score how upset they sound Escalate or resolve? the decision escalate_to_human with a structured summary Resolve refund, reply, done A person acts on the summary rates
The one question to ask: is this a trigger, or a feeling? Write those triggers into the prompt and show a few example chats on each side. When it escalates, send a structured summary (customer ID, cause, amount, recommended action), not the transcript.

Words in the question that tell you which side you're on

Criteria + examples

  • “escalates simple cases, attempts complex ones”
  • “improve escalation calibration”
  • “the prompt says escalate when needed”
  • “asks for a real person” → escalate now

Unreliable proxies

  • sentiment, frustration, tone of the message
  • “confidence score”, “rate how sure it is”
  • “classifier trained on past tickets” (overbuilt)
  • “pick the most likely account” (heuristic)

Try it: replay a day of 100 tickets

Pick a ticket type to follow, then pick the escalation rule. The grid is one day of 100 tickets in arrival order; the type you picked is highlighted. The diagram above shows where your rule acts.

1. The ticket type to follow
2. The escalation rule

resolved, rightescalated, rightescalated, no needkept, needed a person

What the person receives

The rates are illustrative, chosen to show the shape. They are not measurements. The day's mix is 55 simple, 15 angry but simple, 12 policy exceptions, 8 calm but complex and 10 requests for a person, so the best possible first-contact resolution is 70%. The exam guide's point is that sentiment and self-reported confidence are unreliable proxies for complexity.

Things to try: Angry but simple + Sentiment score, then Calm but complex + Sentiment score: the same rule fails both ways. Then Policy exception + Self-rated confidence. Last, any type with Criteria + examples, and watch the day's resolution rate.

The three traps

TrapWhat the wrong option looks likeWhy it loses
Unreliable proxyEscalate on negative sentiment; escalate when Claude's self-rated confidence is low; pick the “most likely” account when two matchMood is not complexity, and the model is often most confident on the cases it gets wrong. A heuristic pick sometimes acts on the wrong customer.
Overbuilt fixA separate classifier trained on historical ticketsNeeds labelled data and retraining, learns past habits, and misses new kinds of case. Criteria and examples in the prompt fix an unclear boundary for far less.
Losing the handoffResolve first when the customer asked for a person; escalate with “needs help” and a link, or with the whole transcriptAn explicit request is a trigger on its own. The person needs a structured summary (customer ID, cause, amount, recommended action) to act without reading the chat.

Practice: three exam-style questions

Answer first, then open “Think it through”.

Rule to remember

Escalate on triggers, not feelings: write the criteria, show examples, and hand over a structured summary. A customer who asks for a person goes at once.

Exam guide 5.2: “Triggers: customer requests a human; policy exceptions/gaps; inability to progress.” “Escalate immediately on explicit demand; offer resolution when straightforward.” “Sentiment-based escalation and self-reported confidence are unreliable proxies for complexity.” Exam guide 1.4: “compile structured handoff summaries (customer ID, root cause, refund amount, recommended action) for humans without the transcript.”

← All traps