Ticket Classification Skill

This skill teaches Claude how to classify support tickets by urgency and category for intelligent routing.

When to Use This Skill

Use this skill when the user requests:

• "Classify this ticket"
• "Analyze this support request"
• "What's the urgency of this ticket?"
• "Route this customer request"
• "Triage these tickets"

Classification Matrix

Urgency Levels

Urgency Keywords

URGENT indicators:

• "system down", "outage", "all users affected", "can't access"
• "security breach", "unauthorized access", "data loss"
• "production broken", "critical failure", "complete outage"
• "can't login" (affecting all/most users)

HIGH indicators:

• "error", "500", "broken", "not working", "failed"
• "doesn't work", "stopped working", "suddenly failing"
• "enterprise", "business critical", "blocking"
• "multiple users affected", "production issue"

MEDIUM indicators:

• "issue", "problem", "slow", "delayed"
• "not working as expected", "performance degraded"
• "workaround available", "intermittent"
• "standard customer", "non-critical"

LOW indicators:

• "how do I", "question", "wondering", "can you"
• "feature request", "enhancement", "suggestion"
• "documentation", "guide", "tutorial"
• "minor bug", "typo", "cosmetic issue"

Category Classification

Technical:

• Error messages (500, 404, timeout, exception)
• Stack traces and error codes
• API issues, webhook failures, integration problems
• Login, authentication, authorization errors
• Performance issues (slow, timeout, connection)
• Database, server, infrastructure problems
• Code-related issues

Billing:

• Payment, invoice, charge, receipt
• Subscription, plan, tier, upgrade, downgrade
• Refund, credit, dispute, cancellation
• Credit card, payment method, billing address
• Pricing, cost, overage, limit
• Trial, free tier, premium features

General:

• How-to questions ("How do I...")
• Account settings, profile, preferences
• Feature questions ("Does this support...")
• Documentation requests
• General inquiries
• Onboarding help
• Non-technical, non-billing questions

Classification Process

Follow these steps:

1. Read the ticket subject and body
2. Scan for urgency keywords (URGENT > HIGH > MEDIUM > LOW)
3. Identify customer tier (Enterprise vs Standard) if mentioned
4. Count affected users if mentioned (more users = higher urgency)
5. Categorize by indicators (error codes = technical, invoice = billing)
6. Apply classification matrix
7. Determine routing based on category and urgency

Examples

Example 1: Technical - HIGH

Subject: 500 Errors on API Endpoint
Customer: Acme Corp (Enterprise)
Body: We're getting 500 errors on /api/users endpoint. Started 2 hours ago.
Affects our production app with 5,000 users.
Error: "Internal Server Error: Database connection timeout"

Classification:

• Urgency: HIGH
• Category: technical
• Routing: engineering
• Reasoning: Production errors affecting enterprise customer, specific error code (500), database issue, multiple users affected, but not complete outage (URGENT would be if entire system down)

Example 2: Billing - MEDIUM

Subject: Question about invoice charges
Customer: Small Business Co (Standard)
Body: My invoice shows $500 but I expected $400. Can you explain
the additional $100 charge? I'd like this clarified before payment.

Classification:

• Urgency: MEDIUM
• Category: billing
• Routing: finance
• Reasoning: Billing inquiry from standard customer, not blocking business, wants clarification, payment not processed yet, standard 4-hour SLA applies

Example 3: General - LOW

Subject: How to export data to CSV
Customer: TechStartup Inc (Standard)
Body: I'm trying to export our user data to CSV but can't find
the export option. Not urgent, just planning for quarterly review.

Classification:

• Urgency: LOW
• Category: general
• Routing: support
• Reasoning: How-to question, user explicitly states "not urgent", feature question, can check KB first, 24-hour SLA acceptable

Special Cases

Escalation Triggers

Immediately escalate (regardless of category) if:

• URGENT urgency level
• Enterprise customer + HIGH urgency
• SLA at risk (<20% time remaining)
• Security breach mentioned
• Data loss or corruption
• Regulatory compliance issue

Auto-Resolve Candidates

Consider KB auto-resolve if ALL true:

• LOW urgency only
• General category
• High-confidence KB match (>90%)
• Common, documented issue
• No customer data involved

Ambiguous Cases

Multiple issues in one ticket:

• Classify by highest urgency issue
• Note multiple categories if present
• Route to team that handles most critical issue

Missing information:

• Assume Standard tier if not specified
• Classify conservatively (higher urgency if unclear)
• Note information gaps in analysis

Output Format

Always provide structured classification:

{
  "urgency": "URGENT|HIGH|MEDIUM|LOW",
  "category": "technical|billing|general",
  "routing": "engineering|finance|support|escalation",
  "customer_tier": "enterprise|standard|unknown",
  "sla_deadline": "Time when response is due",
  "key_indicators": ["keyword1", "keyword2", "keyword3"],
  "confidence": 0.95,
  "reasoning": "Brief explanation of classification",
  "escalation_needed": true|false,
  "auto_resolve_possible": true|false
}

Quality Checklist

Before finalizing classification, verify:

• Urgency matches keyword indicators
• Category aligns with issue type
• Customer tier considered in SLA calculation
• Routing appropriate for category + urgency
• Special cases checked (escalation, auto-resolve)
• Confidence level appropriate

Tips

Finding Urgency Indicators

Look for impact scope:

• "all users" → URGENT
• "5,000 users" → HIGH
• "some users" → MEDIUM
• "just me" → LOW (unless enterprise)

Look for business impact:

• "blocking production" → URGENT/HIGH
• "affecting sales" → HIGH
• "inconvenient" → MEDIUM/LOW

Categorizing Accurately

Technical indicators:

• Error codes, stack traces
• API, endpoint, integration
• Login, auth, database

Billing indicators:

• $, price, cost, charge
• Invoice, receipt, subscription
• Payment, refund, credit card

General indicators:

• Questions starting with "How"
• Feature names without errors
• Account, settings, profile

Routing Decisions

Route to engineering if:

• Technical category + code/API/database
• Requires code changes or debugging

Route to finance if:

• Billing category
• Payment processing needed

Route to support if:

• General category
• KB might have answer
• Can be resolved without engineering/finance

Route to escalation if:

• URGENT urgency
• Enterprise + HIGH + at risk
• Security or data loss

Common Pitfalls

❌ Don't:

• Classify everything as URGENT (crying wolf reduces trust)
• Ignore customer tier (enterprise = faster SLA)
• Miss keywords in subject line (often contains urgency)
• Categorize based on team availability (classify by issue type)
• Skip KB search for common questions (wastes human time)

✅ Do:

• Read full ticket body (key details often buried)
• Note affected user count (scale matters)
• Check for enterprise customer mentions
• Look for error codes and stack traces
• Consider business impact beyond technical severity
• Suggest KB search for LOW priority how-to questions

Success Criteria

A good classification should:

1. ✅ Match urgency to SLA requirements
2. ✅ Route to correct team for resolution
3. ✅ Consider customer tier and business impact
4. ✅ Identify escalation needs proactively
5. ✅ Enable auto-resolution when safe
6. ✅ Provide clear reasoning for decisions