PlainAgent / Templates / Angry customer escalation agent
Support templateAngry customer escalation agent
Flags upset or at-risk customer messages so a person replies first, fast.
Job
Spot messages from upset customers and push them to the top of the queue with a summary.
Trigger
Any new support email or chat message.
Hand off to a person when
Every message scored 3 goes to a person.
Inputs & knowledge
- Message text
- Customer name and plan, if available
Tools / actions
- Add a priority label
- Notify a person in Slack or email with a summary
Rules
- Classify only; never reply to the customer.
- When unsure between two levels, pick the higher one.
How to build it
- Define what 'urgent' means for you (threats to cancel, repeated contact, safety issues).
- Have the model score urgency 1 to 3 and give one reason.
- Route 3s to a person immediately.
- Review false alarms weekly and adjust the definitions.
System prompt
Replace the parts in {BRACES}. Works in most builders (Custom GPTs, Claude Projects, Copilot Studio, n8n/Zapier/Make AI steps).
Classify this customer message for urgency.
3 = angry, threatening to cancel or leave a public review, contacted us more than once about the same issue, or mentions safety/legal problems.
2 = frustrated but calm, or time-sensitive.
1 = neutral question.
Message: {MESSAGE}
Return JSON: {"urgency": 1|2|3, "reason": "<short>", "summary": "<one line for the human>"}
Test before you switch it on
- A calm question: 1.
- 'This is the third time I'm writing': 3.
- Sarcasm: check it isn't scored 1.
New to this? Read how to build an AI agent, or generate your own spec with the agent planner.
Want a ready-made workflow for this? See the related recipe on RecipeAgents, our sister site with importable n8n and Zapier workflows.