AI Customer Support Agent

Role: Customer support architect and AI implementation specialist.

Task: Build a complete AI-powered customer support system using {{AI_TOOL}}.

INTAKE CLASSIFICATION:
Design 6 category intent classifier:
1. Product question (route to: knowledge base + human if unresolved)
2. Billing/ refund (route to: billing team + auto-ticket creation)
3. Technical issue (route to: tech support team + bug report template)
4. Feature request (route to: product team + upvote system)
5. Complaint (route to: human escalation + priority flag)
6. Sales inquiry (route to: sales team + meeting scheduler)

RESPONSE TEMPLATES:
For each intent, create:
- Opening acknowledgment (1 sentence, empathetic)
- Resolution steps (numbered, specific)
- Escalation trigger (specific phrases that mean "hand off to human")
- Closing (set expectation for next step)

HUMAN ESCALATION:
- When to escalate (5 specific triggers)
- Handoff format (what the human sees when they take over)
- Customer notification (how you tell them a human is taking over)

QUALITY CONTROL:
- Weekly review: Random 10 tickets, score them, adjust responses
- Customer satisfaction tracking (simple yes/no thumbs up)
- Monthly false positive report (intents that triggered incorrectly)

Rules:
- Every response template is 50-150 words
- Escalation triggers are specific phrases, not general rules
- The AI should NEVER guess at billing amounts or timeline commitments
- Human handoff must happen in under 2 minutes of escalation signal