Updated October 2026
AI prompts for customer service are reusable instructions that tell a generative AI tool what to write for a given situation, such as an order status reply or an apology for a delayed shipment. The ten prompts below give you the structure, the placeholders to fill in and the instructions that make each reply sound human.
Agents handle everything from damaged items to billing disputes, and the wording of a prompt decides whether the AI’s reply is generic or usable. A good prompt names the situation, the facts to include and the tone to use.
This guide gives 10 purpose-built AI prompts for customer service, each with the instructions that make the reply sound human, followed by the three categories of tools that run them. They are designed to support your team, not replace it.
Table of contents
- Why do AI prompts matter for customer service?
- 10 AI prompts for customer service you can copy
- What makes a good customer service prompt?
- Which prompt patterns help agents beyond replying to customers?
- What should agents never paste into an AI prompt?
- Which tools run AI prompts for customer support?
- How do you combine prompts with verified knowledge?
- Frequently asked questions about AI prompts for customer service
Why do AI prompts matter for customer service?
Generative AI is like a brilliant new hire: it has access to infinite knowledge, but it doesn’t know your brand’s specific rules or tone until you teach it. That “teaching” happens through prompts.
A generic request like “Write a reply to this angry customer” will result in a robotic, generic apology. However, a well-crafted prompt turns that same AI into your best support agent. Here is why the specific wording matters:
- Consistency at Scale: Prompts act as guardrails. They ensure that whether an agent has been with you for two years or two days, the output remains professional, accurate, and on-brand.
- Balancing Speed and Empathy: In CX, speed usually kills empathy. Good prompts fix this by instructing the AI to acknowledge the customer’s frustration first, before diving into the technical solution.
- Reducing Cognitive Load: Support agents face “decision fatigue” from constantly crafting unique replies. A library of verified prompts allows agents to focus on solving the problem, letting the AI handle the typing.
10 AI prompts for customer service you can copy
To get you started, here are 10 essential AI prompts for customer service covering common high-stress scenarios. Each one is built to help your AI tools generate empathetic, accurate and brand-aligned replies.
The table lists what each reply has to include. The prompts follow in the same order.
| # | Scenario | What the reply must include |
|---|---|---|
| 1 | Order status | Thanks, the current status, and an invitation to ask more |
| 2 | Product availability | Link if in stock; if not, timing, an alternative and how to be notified |
| 3 | Delayed shipment | Acknowledgment, current status, and a goodwill gesture if appropriate |
| 4 | Broken or faulty item | Apology, a photo upload route and the next step under policy |
| 5 | Missing item | Order details, and a reship, refund or carrier follow-up |
| 6 | Wrong item | Apology, a photo for investigation and the correct item or pickup |
| 7 | Cancellation | Order status, whether cancellation is possible, and what to do if not |
| 8 | Return request | Eligibility, step-by-step instructions and what happens after receipt |
| 9 | Exchange | Approval status, how to return the original and when the replacement ships |
| 10 | Refund | Refund status, the amount and when it will arrive |
1. For a customer checking on their order status
Prompt
Create a warm and respectful message for [Customer Name], who is asking about their order. Thank them for reaching out and express appreciation for their patience. Tell them that the order is [being packed, dispatched, reaching them in X hours/days]. Invite them to contact us if they need anything else. Use a tone that is [friendly, professional, reassuring].
Humanize it by instructing AI to:
- Show genuine appreciation for their time and patience
- Make them feel seen and valued, not just another support ticket
- Use positive, courteous language that affirms their importance
2. For a customer asking if a product is in stock or when it will return
Prompt
Create a friendly and helpful response to [Customer Name] asking about a product’s availability.
If the item is in stock, thank them for their interest and provide the link: [product URL].
If the item is out of stock: explain that the product is [temporarily unavailable, expected back in X days/weeks]. Recommend a suitable alternative: [similar product name or link]. Let them know we’ll [notify them when it’s back, suggest checking back on X date]. Match that is [warm, helpful, informative].
Humanize it by instructing AI to:
- Acknowledge their interest sincerely
- Guide rather than sell
- Offer a thoughtful alternative, not just a generic option
3. For a customer who is frustrated by a delayed shipment
Prompt
Generate a calm and empathetic message to [Customer Name] whose order is delayed. Acknowledge their frustration and thank them for their patience. Let them know that the shipment is currently [in transit, delayed due to X, expected to arrive on X date]. If appropriate, offer a goodwill gesture such as [discount code, loyalty points, voucher]. Close with an open offer to assist further. Use a tone that is [understanding, sincere, professional].
Humanize it by instructing AI to:
- Own the inconvenience without making excuses
- Reassure with action, not just an apology
- Personalize the message with their name and order details
4. For a customer who received a broken or faulty item
Prompt
Write a compassionate message to [Customer Name] who received a damaged item. Apologize sincerely and thank them for letting us know. Let them know we’ll resolve the issue quickly. Ask them to upload a photo using [upload method or link]. Then outline the next step: [refund, replacement, or return process based on policy]. Use a tone that is [reassuring, respectful, brand-aligned].
Humanize it by instructing AI to:
- Lead with empathy and gratitude
- Don’t make them work too hard to get a resolution
- Be clear and proactive about what happens next
5. For a customer reporting a missing item from their delivery
Prompt
Create a warm and helpful message for [Customer Name] who reported a missing item. Thank them for flagging it and acknowledge how frustrating this must be. Confirm the order details using [CRM, order history, or tracking link]. Let them know we will [reship the item, issue a refund, or follow up with the carrier] based on [fulfilment policy]. Close with a kind note inviting them to contact us for anything else. Use a tone that is [understanding, supportive, brand-appropriate].
Humanize it by instructing AI to:
- Focus on the emotional impact, not just fixing the problem
- Be transparent about next steps
- Reinforce that their experience matters to the brand
6. For a customer who received the wrong item
Prompt
Write a kind and understanding message to [Customer Name] who received an incorrect item. Apologize for the mistake and thank them for bringing it to our attention. Ask for a photo of the received item via [upload method or link] to help us investigate. Let them know we will [ship the correct item, arrange a pickup, or provide next steps] based on [brand policy]. Keep the tone [respectful, human, apologetic, and solution-focused].
Humanize it by instructing AI to:
- Take full responsibility without blame-shifting
- Make the customer feel like a priority, not a problem
- Reassure them that the fix will be easy and quick
7. For a customer requesting to cancel their order
Prompt
Create a respectful and efficient message for [Customer Name] who wants to cancel their order. Confirm the order status is currently [unshipped, shipped, processing] and inform them that cancellation is [possible/impossible] based on that status. If it can’t be cancelled, explain what they can do next: [refuse delivery, initiate return]. Use a tone that is [courteous, helpful, and non-defensive].
Humanize it by instructing AI to:
- Show understanding for their change of mind
- Avoid defensiveness and instead keep it easy and open
- Offer clear next-step instructions
8. For a customer initiating a return request
Prompt
Write a helpful and friendly response to [Customer Name], who wants to return an item. Thank them for reaching out and confirm the item is eligible for return under [return policy terms]. Provide step-by-step instructions: [return form, label link, deadline]. Let them know what to expect after it’s received: [refund timeline, confirmation email]. Use a tone that is [courteous, confident, and supportive].
Humanize it by instructing AI to:
- Make the return process feel simple and hassle-free
- Respect their right to return without friction
- Use clear formatting for step-by-step instructions
9. For a customer requesting an exchange
Prompt
Create a friendly and accommodating response to [Customer Name] requesting an exchange. Thank them for their message and confirm that the exchange is [approved, eligible, or conditional] per [exchange policy]. Ask them to return the original item using [return method or link], and confirm that the replacement item will be sent once we [receive the original item, process request]. Tone should be [understanding, helpful, and flexible].
Humanize it by instructing AI to:
- Show appreciation for their interest in staying with the brand
- Clarify timelines and requirements to avoid confusion
- Reassure that the exchange will be smooth and easy
10. For a customer requesting a refund
Prompt
Write a considerate and reassuring response to [Customer Name], who has asked for a refund. Thank them for reaching out and confirm their refund is [in process, approved, or pending return verification] according to [refund policy]. Let them know the amount being refunded and when they can expect to see it: [X days, original payment method]. Use a tone that is [transparent, warm, and appreciative].
Humanize it by instructing AI to:
- Acknowledge their experience and decision without judgment
- Provide a clear and confident explanation of the process
- End on a positive, relationship-saving note
Don’t Stop at Ten. Get the Full 60+ Prompt Playbook
What makes a good customer service prompt?
The ten prompts above share a structure, and you can reuse it for any situation not on the list. Five parts do most of the work.
| Part | What to include | Example |
|---|---|---|
| Role | Who the AI is writing as | You are a support agent for [company] |
| Context | The facts of the case: order details, ticket history, plan, policy | The order shipped on [date] and is delayed at the carrier |
| Task | The single job to do | Write a reply that explains the delay and the next step |
| Constraints | What to include and avoid, taken from policy | Offer [goodwill gesture] only if the delay is over [days]; do not promise a delivery date |
| Tone and format | How it should sound and how long it should be | Calm and direct, under 120 words, no jargon |
Here is the difference in practice. A weak prompt says: “Reply to this angry customer about a late order.” A stronger one says: “You are a support agent for [company]. The customer’s order [number] shipped on [date] and is delayed at the carrier. Write a reply under 120 words that acknowledges the delay in the first sentence, gives the current status, offers [goodwill gesture] as policy allows, and does not promise a delivery date. Tone: calm and direct.” The second version gives the AI the facts and the limits, so the agent edits instead of rewriting.
Which prompt patterns help agents beyond replying to customers?
Agents also use AI for internal work. These four patterns are separate from the ten reply prompts above, and they follow the same five-part structure.
| Pattern | What it does | Starter prompt |
|---|---|---|
| Ticket summary | Condenses a long history so the next agent does not reread it | Summarize this ticket history in five lines: the issue, what was tried, what the customer was told, what is open, and the promised next step. History: [paste] |
| Tone rewrite | Softens or tightens a draft the agent already wrote | Rewrite this reply to be warmer and shorter, keeping every fact and number unchanged: [paste] |
| Troubleshooting steps | Turns a known fix into a customer-ready sequence | Turn these internal steps into a numbered list for a non-technical customer, one action per step: [paste] |
| Translation check | Helps with a reply in another language | Translate this reply into [language], keep the tone polite and formal, and list any phrases that may not translate directly: [paste] |
In every case the agent still checks the output. The AI drafts, and the agent decides what is sent.
What should agents never paste into an AI prompt?
Prompts often include customer details, which creates a data handling question that the prompt text itself cannot solve. Set a rule before agents start, and put it where they work.
- Do not paste payment card numbers, passwords, security answers or full identity numbers into any AI tool.
- Use only tools your organization has approved, with a data processing agreement that covers customer data, and check whether the vendor uses inputs for model training.
- Replace names, addresses and account numbers with placeholders when you use a general-purpose tool that is not approved for customer data.
- Review every draft against the customer record and the policy before sending. Check numbers, dates and promises first.
- Keep health, financial and other regulated information inside systems that are approved for it.
Which tools run AI prompts for customer support?
Great Gen AI prompts for customer service need the right execution. To stop agents from constantly switching tabs, you need tools that integrate AI directly into their workflow.
Here are the three key categories to consider, starting with the most critical: Knowledge Management.
1. Knowledge Management & Helpdesk Tools
The General Concept: A prompt is only as good as the information it can access. If you ask an AI to “write a refund policy email” but don’t give it your actual policy, it will hallucinate an answer (see why RAG fixes hallucinations in customer service). Knowledge Management (KM) tools solve this by acting as the “brain” of your support operation. They store your SOPs, policies, and troubleshooting guides, much like an AI knowledge base does. When you run a prompt within a KM tool, the AI pulls from your verified data, ensuring the output is not just polite but factually correct.
Top Tool:
Knowmax takes this a step further by combining Knowledge Management with Generative AI. Instead of just storing static documents (if the basics are new to you, start with what a knowledge base is), Knowmax uses your trusted knowledge base to ground its AI.
- How it works with prompts: You can set up prompts within Knowmax that automatically pull data from your decision trees or troubleshooting guides.
- The Benefit: Your agents get an instant, AI-generated response that is strictly based on your approved company knowledge, eliminating the risk of the AI making up policies.
Ensure Your KM System Meets Modern CX Needs
2. General Generative AI Models (The “Co-Pilots”)
The General Concept: These are the standalone AI tools that act as a creative sidekick for your agents. They are excellent for ad-hoc tasks, such as rephrasing a clumsy sentence, translating a message, or softening the tone of a difficult email. While they aren’t connected to your customer database, they are powerful engines for language processing.
Examples:
- ChatGPT (OpenAI) / Gemini (Google): Great for pasting a rough draft and using a prompt like “Rewrite this to be more empathetic” or “Summarize this long ticket history.”
- Grammarly: useful for real-time tone adjustments as agents type directly in their email client.
3. AI-Native Customer Support Platforms
The General Concept: Many modern ticketing systems now have “AI layers” built directly into the dashboard. These platforms often allow admins to save “Prompt Macros.” Instead of an agent typing out a full prompt, they click a button (e.g., “Expand,” “Shorten,” or “Formalize”), and the system runs a pre-set prompt in the background to modify the text.
Examples:
- Intercom / Zendesk AI: These platforms include AI features that help agents draft and refine replies inside the ticket. Check each vendor’s current documentation for what is included in your plan.
- Salesforce Service Cloud: Lists AI-powered sentiment analytics among its service operations features, according to Salesforce’s own service operations page.
How do you combine prompts with verified knowledge?
Generative AI can speed up customer support, but only if someone steers it. The AI prompts for customer service above are instructions that turn a generic bot into a useful assistant.
However, even the best prompt can fail if it lacks accurate information. That is why combining structured prompts with verified knowledge (via tools like Knowmax) is the pairing to aim for in modern CX. It lets agents move faster while staying accurate and empathetic.
The goal is to clear the clutter so agents have time to build the human connection. Start with these prompts, adjust them to your voice, and let your team focus on the customer while AI handles the words.
Want to Build a Smarter, Faster Support Operation
Frequently asked questions about AI prompts for customer service
They are reusable instructions that tell a generative AI tool what to write for a given support situation, such as an order status update, an apology for a delay or a refund confirmation. They fix the structure, the facts to include and the tone, so replies stay consistent whoever sends them.
Name the situation, list the facts the reply must include, and set the tone. Then add humanizing instructions, such as acknowledging the customer’s frustration first, being clear about what happens next and personalizing the reply with the customer’s name and order details.
A prompt is only as good as the information the AI can access. If it is not given your actual policy, it may invent an answer. Knowledge management tools store approved SOPs, policies and guides, so the AI draws on verified content.
No. The prompts here are designed to support agents by handling the typing and keeping replies consistent, so the agent can focus on solving the problem and on the customer.
A good prompt names the role, gives the facts of the case, states one task, sets constraints from policy and fixes the tone and length. For example: “You are a support agent. The order is delayed at the carrier. Write a reply under 120 words that acknowledges the delay, gives the status and does not promise a date.”
They can use general AI tools to draft and refine replies, provided the tool is approved by the organization for customer data and a person checks every reply before it is sent. A general tool does not know your policies, so supply the facts in the prompt or use a tool connected to your knowledge base.
Do not paste card numbers, passwords, security answers or full identity numbers. Use placeholders for names and account details unless the tool is approved for customer data, and keep regulated information inside approved systems.
Give the AI the policy text in the prompt, or use a tool that retrieves approved articles and cites them. Tell it to say when the policy does not cover the case, and have the agent check every promise, amount and date against the record before sending.

