Customer Support Mentor

An autonomous agent that builds and optimizes customer support operations — designing ticket routing, creating knowledge bases, implementing SLA management, and training support teams on escalation protocols and communication best practices.

When to Use This Agent

Choose Customer Support Mentor when:

• You need to set up or improve customer support workflows and SLAs
• You want to create a knowledge base or help center with self-service content
• Support ticket volume is growing and you need better routing and automation
• You need to design escalation paths and train support team members

Consider alternatives when:

• You need a customer success program for enterprise accounts (use a CS consultant agent)
• You need chatbot development (use a conversational AI agent)
• You need help desk software selection (evaluate tools directly)

Quick Start

Copy
# .claude/agents/customer-support-mentor.yml
name: customer-support-mentor
description: Build and optimize customer support operations
agent_prompt: |
  You are a Customer Support Operations Expert. Help with:

  1. Design ticket routing and prioritization rules
  2. Create SLA tiers with response and resolution targets
  3. Build knowledge base articles for common issues
  4. Develop escalation procedures and decision trees
  5. Create macros and templates for common responses
  6. Set up support metrics and quality assurance

  Support principles:
  - First response time matters more than resolution time
  - Self-service should handle 70% of questions
  - Escalation paths must be clear and documented
  - Every resolved ticket is a potential KB article

Example invocation:

Copy
claude "Design a tiered support system for our SaaS with free, pro, and enterprise customers"

Sample support design:

Tiered Support System Design
━━━━━━━━━━━━━━━━━━━━━━━━━━━

Tier Structure:
           Free         Pro           Enterprise
Channel:   Email only   Email + Chat  Email + Chat + Phone
SLA (1st): 24 hours     4 hours       1 hour
SLA (res): 72 hours     24 hours      8 hours
Priority:  Normal       High          Urgent
Dedicated: No           No            Named CSM
Coverage:  Business hrs Business hrs  24/7

Routing Rules:
  1. Enterprise customers → dedicated queue (auto-assign to CSM)
  2. Pro customers → priority queue (round-robin assignment)
  3. Free customers → general queue (FIFO assignment)
  4. Keywords "down", "outage", "broken" → escalate to urgent
  5. Billing issues → billing specialist queue

Automation:
  - Auto-reply with estimated response time
  - Suggest KB articles based on ticket content
  - Auto-close after 7 days of no customer response
  - CSAT survey sent after resolution

Core Concepts

Support Operations Framework

Component	Purpose	Metrics
Ticket Routing	Direct to right agent/team	Routing accuracy, reassignment rate
SLA Management	Ensure timely responses	SLA compliance %, breach rate
Knowledge Base	Enable self-service	Deflection rate, article helpfulness
Escalation	Handle complex issues	Escalation rate, MTTR
Quality Assurance	Maintain response quality	CSAT, QA score, tone consistency

Knowledge Base Article Template

Copy
## [Issue Title]

**Applies to:** [Product tier / version]
**Last updated:** [Date]

### Problem
[1-2 sentence description of what the user is experiencing]

### Solution

**Quick Fix:**
[Step-by-step instructions for the most common resolution]

1. Go to Settings → Account
2. Click "Reset API Key"
3. Copy the new key and update your integration

**If that didn't work:**
[Alternative solution for edge cases]

### Related Articles
- [Link to related KB article 1]
- [Link to related KB article 2]

### Still need help?
Contact support at [email protected] with:
- Your account email
- Screenshot of the error
- Steps you already tried

Escalation Decision Tree

Customer reports issue
  │
  ├─ Is it a known issue? → Link KB article, close
  │
  ├─ Can Tier 1 resolve? → Resolve, document, close
  │
  ├─ Is it a bug? → Create bug ticket, notify engineering
  │   └─ Is it P1 (service down)? → Page on-call engineer
  │
  ├─ Is it a feature request? → Log in feature tracker
  │
  └─ Complex/unclear? → Escalate to Tier 2
      └─ Still unresolved? → Escalate to engineering

Configuration

Option	Type	Default	Description
tiers	string[]	["free", "pro", "enterprise"]	Customer support tiers
channels	string[]	["email", "chat"]	Support channels
slaTargets	object	{ firstResponse: "4h", resolution: "24h" }	SLA targets by tier
kbPlatform	string	"zendesk"	KB platform: zendesk, intercom, helpscout
autoRouting	boolean	true	Enable automated ticket routing
csatSurvey	boolean	true	Send CSAT surveys after resolution

Best Practices

1. Respond fast, even if you cannot resolve fast — A first response within 1 hour that says "I see your issue, I'm looking into it" is better than a resolution response in 4 hours. Customers primarily want acknowledgment that their issue has been received and someone is working on it.

2. Turn every resolved ticket into a KB article candidate — If a support agent resolves an issue that is not covered in the knowledge base, they should flag it for KB creation. Over time, this builds a comprehensive self-service library that reduces ticket volume. Target: 70% of common questions answerable via KB.

3. Implement keyword-based auto-routing for accuracy — Tickets containing "billing," "invoice," or "charge" should route to the billing team. Tickets with "API," "integration," or "webhook" should route to technical support. Simple keyword rules correctly route 80% of tickets and reduce reassignment time.

4. Use macros for consistency, not for laziness — Create templated responses for common issues, but require agents to personalize them. A macro that starts with the customer's name and references their specific situation feels human. A macro pasted without modification feels robotic and damages trust.

5. Measure CSAT at the ticket level, not the agent level — CSAT measures whether the customer's issue was resolved satisfactorily, which depends on the product, the process, and the agent. Using CSAT as an individual performance metric incentivizes agents to avoid difficult tickets and cherry-pick easy ones.

Common Issues

Support ticket volume grows faster than the team — Hiring cannot keep up with ticket volume. Focus on deflection: improve the knowledge base, add in-app tooltips for common confusion points, and implement chatbot triage for simple questions. Every 1% improvement in self-service deflection reduces ticket volume proportionally.

Escalated tickets get lost between teams — A ticket escalated from support to engineering disappears into a development backlog with no customer communication. Assign an owner to every escalated ticket who is responsible for updating the customer, even if the resolution is in another team's hands. The customer should never have to ask "what's the status?"

SLA compliance drops during peak periods — Ticket spikes (product launches, outages, billing cycles) cause SLA breaches. Build capacity for 1.5x normal volume, create pre-written responses for anticipated spikes (launch day FAQ), and implement priority triage during peaks that focuses on high-value customers first.