ChatGPT Prompts for Customer Service

Write a customer service email response for this situation:

**Channel:** Email
**Customer sentiment:** {{sentiment}} (e.g., frustrated, angry, disappointed)
**Issue:** {{issue_description}}
**Customer history:** {{customer_context}} (e.g., "loyal customer for 3 years", "first-time buyer")
**Our policy:** {{relevant_policy}}
**Resolution options available:**
1. {{option_1}}
2. {{option_2}}

Tone: {{brand_tone}} (e.g., empathetic and professional, warm and casual)

Structure the response as:
1. Acknowledge their frustration specifically (reference their exact issue)
2. Take responsibility without blame-shifting
3. Present the resolution clearly
4. Set expectations for timeline
5. Close with a commitment to their satisfaction

Keep under 200 words. Do not use phrases like "I understand your frustration" — be more specific.

Generate a live chat response for a customer support agent.

**Customer message:** "{{customer_message}}"
**Issue category:** {{category}} (e.g., billing, technical, shipping, returns)
**Agent can:** {{available_actions}} (e.g., "issue refund up to $50", "escalate to tier 2", "offer discount code")

Requirements:
- Maximum 3 sentences
- Conversational but professional
- Include a specific next step or question
- If the issue can be resolved immediately, resolve it
- If not, set a clear expectation for when they'll hear back

Do NOT use: "Thank you for reaching out", "I apologize for the inconvenience", or any other filler.

Create an escalation script for this situation:

**Issue type:** {{issue_type}}
**Why it needs escalation:** {{escalation_reason}} (e.g., "exceeds agent refund authority", "requires engineering investigation")
**Information already collected:**
- {{info_1}}
- {{info_2}}

Generate:
1. **Internal escalation note** — summarize the issue, what's been tried, what the customer expects, and urgency level (1-5)
2. **Customer-facing message** — tell the customer their issue is being escalated, who will handle it, expected response time, and what (if anything) they need to do
3. **Handoff checklist** — bullet list of everything the next agent needs to know

Customer-facing tone: {{brand_tone}}

Generate an FAQ section based on these common support tickets:

{{paste_5_to_10_recent_tickets}}

For each FAQ entry:
- **Question:** Rephrase the customer's issue as a clear question
- **Answer:** Write a self-service answer that solves the problem without contacting support
- **Keywords:** 3-5 search terms customers might use to find this

Requirements:
- Write at a {{reading_level}} reading level
- Keep each answer under 100 words
- Include specific steps (not vague directions)
- If the answer involves navigation, use the format: Settings > Billing > Cancel Subscription
- Group related questions under category headers

Output as markdown.

Write a follow-up email for a customer who gave a {{rating}}/5 CSAT score.

**Original issue:** {{original_issue}}
**Resolution provided:** {{resolution}}
**Agent who handled it:** {{agent_name}}
**Days since resolution:** {{days}}

For ratings 1-2 (unhappy):
- Acknowledge we fell short
- Ask what we could have done differently
- Offer to reopen or connect with a senior agent
- Keep it personal, not automated-sounding

For ratings 3 (neutral):
- Thank them, ask one specific question about what would have made it a 5

For ratings 4-5 (happy):
- Thank them briefly
- Ask if they'd be willing to leave a review on {{review_platform}}
- Include direct link

Maximum 100 words. Sign from {{agent_name}}, not "The Support Team".

Classify this support ticket and extract structured data:

**Ticket:** "{{ticket_text}}"

Return a JSON object with:
{
  "category": "billing | technical | shipping | returns | account | feature_request | other",
  "urgency": "low | medium | high | critical",
  "sentiment": "positive | neutral | frustrated | angry",
  "key_issue": "one sentence summary",
  "product_mentioned": "product name or null",
  "action_needed": "what the customer wants",
  "suggested_response_type": "template | custom | escalation",
  "auto_resolve_possible": true/false,
  "reason": "why you classified it this way"
}

If the ticket contains multiple issues, return the primary issue and note secondary issues in a "secondary_issues" array.