How do you implement automated customer support ticket routing and escalation?

Automated ticket routing uses AI classification and business rules to direct support requests. Architecture: ticket ingestion (email, chat, web form, phone transcription) feeds into an NLP pipeline that extracts intent, urgency, product area, and customer tier. Classification: fine-tune a text classifier on historical ticket data (categories, priorities, assigned teams), or use zero-shot with an LLM for initial deployment. Routing logic: combine AI classification with business rules: route by product area to specialized teams, escalate VIP customers to senior agents, route technical issues to engineering support, and route billing issues to finance. Skill-based routing: match ticket requirements to agent skills, consider agent workload and availability, and implement round-robin within skill groups for fair distribution. Escalation automation: time-based escalation (no response within SLA triggers manager notification), sentiment-based escalation (detect frustration or anger in customer messages), complexity-based escalation (ticket touches multiple product areas), and customer-tier escalation (enterprise customers get automatic priority boost). Implementation: integrate with helpdesk platforms (Zendesk, Freshdesk, Intercom) via API, implement auto-tagging for reporting, provide suggested responses from a knowledge base using RAG, and auto-resolve common issues (password resets, status inquiries, FAQ questions) with chatbot integration. Metrics: first-response time, resolution time, customer satisfaction, routing accuracy, and auto-resolution rate. Target: 40% auto-resolution, 90% routing accuracy, 50% reduction in first-response time.