Free guide · works with ChatGPT, Claude or Gemini
ChatGPT prompts for customer service replies
Gives you a complete, reusable system to answer enquiries, complaints, refunds, delays, and reviews quickly, kindly, and consistently in your brand voice. You finish with send-ready replies, a living reply library, and a maintenance loop that improves results each week.
Step 1, free.
Copy this, paste it into your AI, and swap the [BRACKETS] for your own details.
Setup: Capture brand voice and policies
Provide [BRAND SNAPSHOT], [CUSTOMER PROMISE], [WRITING SAMPLES (3–5 emails or reviews you’ve written)], [TONE YOU WANT (3 adjectives)], [FORBIDDEN WORDS/PHRASES], [APPROVED PHRASES], [NON‑NEGOTIABLE POLICIES], [COMMON SENSITIVE TOPICS], [ESCALATION CONTACTS], [SIGN‑OFFS YOU USE]. Act as a brand voice analyst and CX policy codifier to turn this into a reusable Voice & Policy Profile. Work through it like this: 1) Mine [WRITING SAMPLES] to extract Voice Pillars (e.g., warmth level, directness, formality), sentence patterns, and empathy moves; quote 3 short examples to anchor. 2) Translate [TONE YOU WANT] into concrete do/don’t style rules (greetings, contractions, emoji, apology structure, positive framing, plain-English constraints). 3) Build an Empathy & Courtesy Lexicon: approved apologies, appreciation, assurance, ownership, and recovery phrases; map to situations. 4) Compile a Policy Digest from [NON‑NEGOTIABLE POLICIES] and [COMMON SENSITIVE TOPICS]: what we can/can’t offer, proof required, time limits, red lines. 5) Add Operational Basics: [SIGN‑OFFS], [ESCALATION CONTACTS], [FORBIDDEN WORDS/PHRASES], [APPROVED PHRASES]. Give me the output as: Brand Voice & Policy Profile with sections: A) Voice Pillars (bullets), B) Style Rules (table: Rule, Do, Don’t, Example), C) Empathy & Courtesy Lexicon (table: Situation, Phrase, Purpose), D) Policy Digest (table: Topic, Allowed, Not Allowed, Conditions, Evidence), E) Operational Basics (contacts, sign‑offs, words list). Quality bar: Every rule has a clear Do and Don’t with one concrete example drawn from your samples or written to match them. If any field is missing, mark it MISSING and ask 1–3 precise questions to fill the gap; don’t invent content.
How to get a good answer.
- 01Fill every [BRACKET] with real detail. Vague in, vague out.
- 02Read the answer, then reply with what's wrong or missing. The AI fixes it in seconds.
- 03Check facts, prices and anything legal or safety-related yourself before you use it.
The rest of the workflow.
Step 1 gets you started. The full Customer Reply & Reputation Kit takes you all the way, 15 steps in order:
- 01Setup: Capture brand voice and policies (free above)
- 02Setup: Map customer touchpoints
- 03Setup: Build policy decision tree
- 04Production: Build reply component library
- 05Production: Create scenario playbooks
- 06Production: Complaint resolution workflow
- 07Production: Difficult situations de-escalation set
- 08Production: Review response bank
- 09Production: Case intake form (context-to-brief)
- 10Quality: Tone, risk, and policy check
- 11Quality: Human approval checklist
- 12Delivery: Channel formatting and send pack
- 13Delivery: CRM logging and tag map
- 14Maintenance: Weekly insights and improvement backlog
- 15Maintenance: A/B test and refresh schedule
Pay once, keep it. Delivered instantly with a copy button on every prompt.
Questions
Does this work with the free version of ChatGPT?
Yes. The prompt works in the free versions of ChatGPT, Claude and Gemini, on a phone or a computer.
What do I put in the [BRACKETS]?
Your real details: notes, numbers, names, whatever the bracket asks for. The more specific you are, the better the answer.
Is the AI's answer always right?
No. Treat it as a strong first draft: check facts, figures and anything legal or safety-related before you use it.
What's in the full Customer Reply & Reputation Kit?
15 steps run in order, a system prompt that sets the AI up for the job, worked examples, fix-it prompts and a quality checklist. It's NZ$19.50 once, no subscription.
Know someone who'd use this?
Send them the free prompt: