A customer service conversation example is a full exchange between a customer and an agent, with both sides written out, that shows how a good reply opens, checks the facts, resolves the issue and closes. Below are six for online stores, across email, messaging and phone, each annotated with what the agent did well and why it works.
Many published scripts give only the agent’s lines. A real conversation shows what the customer says back and how the agent adapts. Single lines and reply templates live in our templates hub. Every conversation here is an illustrative example from a fictional store, written for this article, not a real customer exchange. [Placeholders] mark what to fill in from your own order data.
What does a good customer service conversation look like?
A good customer service conversation moves through the same stages on every channel: greet and identify, understand the problem, check the facts in your systems, resolve or set the next step, confirm the customer is happy with it, and close with what happens next and when. The wording changes by channel, the order does not.
| Stage | What the agent does | Example line |
|---|---|---|
| Open | Greets, gives a name, offers help | ”Hi [name], this is [agent] from [store]. How can I help?” |
| Understand | Lets the customer finish, then summarises it back | ”So the parcel arrived without the [item]. Have I got that right?” |
| Check | Looks up the order before answering, asks one question at a time | ”Let me check order [number]. One moment.” |
| Resolve | Gives the answer or action, or the next step with a date | ”I’ve sent a replacement. It ships on [date].” |
| Confirm | Checks it solves the problem | ”Does that work for you?” |
| Close | Recaps what happens next, who does it and by when | ”You’ll get tracking by email on [date]. Anything else?” |
The stages are the script. The words should sound like your team, which is the point of the section on reading from a script further down.
Customer service conversation examples for an online store
Here are three written conversations on email and messaging, from start to finish. Each shows both sides, the facts the agent checked and the moment it could have gone wrong. After each one, a short list covers what the agent did well. For live chat transcripts, see our chat examples, and for single messages, our email templates.
1. Change of delivery address after the order is placed (email)
Customer: Hi, I've just moved and need my order #[number] sent to a new address. Can you change it?
Agent: Hi [name], this is [agent] from [store]. Let me check where your order is. One moment.
Agent: Good news: the order is still [unfulfilled], so we can change the address. Please send the full new address in one reply: street, number, postcode and city.
Customer: [New address].
Agent: Thank you. I've updated order #[number] to [new address]. The delivery estimate is now [date range] from [carrier].
Agent: One note: if an order has already been handed to the carrier, we can't change the address ourselves. In that case I would contact [carrier] or arrange a return, and I would tell you straight away.
Customer: Perfect, thanks.
Agent: You're welcome. You'll get a new confirmation email within [time]. Reply here if it doesn't arrive.What the agent did well:
- Checked the order status before promising anything.
- Asked one question that gathered everything needed, so there was no back and forth.
- Confirmed the change in writing with a new delivery estimate.
- Was honest about the dispatched case without making it the customer’s problem.
2. Discount code did not apply at checkout (WhatsApp or SMS)
Customer: I paid full price on #[number]. My code [CODE] didn't work. Can I get the difference back?
Agent: Hi [name], [agent] here from [store]. Let me look at your order and the code. One moment.
Agent: I can see the code needs a [minimum spend] of [amount], and your basket was [amount] before delivery, so it didn't apply.
Customer: That wasn't shown at checkout.
Agent: You're right that it should have been clearer, and I'll pass that on. Because [the minimum spend was not visible at checkout], I've refunded [amount] to your original payment method as a goodwill gesture. It shows within [timeline].
Customer: Thanks, that's fair.
Agent: Glad that's sorted. Anything else I can help with?What the agent did well:
- Named the rule in one plain sentence rather than saying “that’s our policy”.
- Gave the amount and the timeline for the refund.
- Checked the code conditions against the order, then used the discretion your policy gives for unclear terms.
- Never asked for card details in the message.
3. Cancellation request after the order has shipped (email)
Customer: Please cancel order #[number]. I don't want it any more.
Agent: Hi [name], [agent] from [store] here. Let me check the status. One moment.
Agent: The order left our warehouse on [date] and is with [carrier], so I can't cancel it now. Here is what I can do instead.
Agent: Option one: refuse the delivery and it comes back to us. Option two: when it arrives, return it under our returns policy and we refund you.
Customer: I'll return it. How long does the refund take?
Agent: Once the parcel is back with us, we refund your original payment method within [number] days. Here is the return label: [link].
Agent: I've noted the case, so you won't need to explain it again. Does that cover it?
Customer: Yes, thanks.What the agent did well:
- Turned a “no” into the best available option.
- Respected the customer’s right to return the item under your returns policy and the consumer law where you sell.
- Gave dates, not “soon”. For more on this scenario, see our refunds page.
Customer service telephone conversation example
A phone call needs the same stages plus three things text does not: verifying who is calling, telling the caller what you are doing during silences or holds, and reading the next step back before you hang up, because the caller has nothing in writing. The two calls below are written for human agents.
Phone callers often arrive stressed. In a CM.com survey of over 7,000 consumers (a vendor survey published in May 2023), phone was still the most preferred channel at 69%, but four in five respondents said they only call when it is really urgent. More than a third were nervous about picking up the phone, and just under half assumed they must be confrontational. The opening matters.
4. Charged twice for one order (phone, human agent)
Agent: Good morning, you're through to [agent] at [store]. How can I help?
Caller: I've been charged twice for one order.
Agent: I'm sorry about that, I'll look into it now. First, so I know I'm speaking to the right person, can I have your order number and the email address or postcode on the order? Please don't read out any card numbers.
Caller: It's #[number], and the postcode is [postcode].
Agent: Thank you, that matches. So you placed one order and see two payments on your statement. Have I got that right?
Caller: Yes.
Agent: May I put you on a short hold while I check the payment record? It'll take about two minutes.
Caller: Go ahead.
Agent: Thank you for waiting. The order shows [one capture and one pending authorisation / two captures]. [If authorisation: The second amount is a temporary hold that your bank releases within [time]. Nothing further is taken.] [If two captures: That is our error, and I've refunded the extra [amount] to the original card.]
Caller: When will I see that?
Agent: It should show on your statement within [timeline]. Your reference is [reference]. To recap: [action], by [date], and you don't need to do anything else. Does that cover it?
Caller: Yes, thanks.
Agent: Thanks for calling, [name]. Have a good day.What the agent did well:
- Verified the caller with order number plus email or postcode and told them never to read out card numbers.
- Signposted the hold and gave it a time.
- Explained what the record showed in plain words.
- Read back the reference and the next step before closing.
5. Caller asks for a manager (phone, human agent)
Caller: This is the third time I'm calling. I want a manager.
Agent: I can hear how frustrating that is, and I'm sorry you've had to call again. Here is what I can do right now: I have your case open and can [action]. I can also arrange a call from a team lead at a time that suits you.
Caller: I'd rather speak to someone now.
Agent: Understood. I can transfer you now, and I'll give them the full case first so you don't need to repeat it. There may be a short wait. Or I can have them call you by [time]. Which would you prefer?
Caller: Transfer me.
Agent: Thank you. I'll stay on the line until they pick up. Handover note: order #[number], [issue], what we've tried: [steps], what the customer wants: [outcome].What the agent did well:
- Never said “calm down”; acknowledged the impact instead.
- Said what it could do straight away, then offered a choice.
- Promised only what it controlled.
- Summarised the case for the next person.
These calls are for your human team. If phone is a big share of your volume, see our guide to voice AI.
How do you handle customer complaints? A script you can follow
Handle a complaint in five steps: hear the customer out without interrupting, show you understand, apologise for the impact, resolve it or agree the next step, then find out why it happened so it does not recur. The steps are the script. The wording should be yours.
This sequence is widely taught as H.E.A.R.D. (Hear, Empathise, Apologise, Resolve, Diagnose). The training firm Customer Service Group (CSG) describes it in its “Service Recovery the Disney Way” piece. We have not found it in Disney’s own documentation, so treat the Disney link as training folklore, not fact.
| Step | What you say | Why it works |
|---|---|---|
| Hear | ”Tell me what happened. I’ll listen first.” | The customer finishes, so you answer the real problem |
| Empathise | ”That would have annoyed me too.” | Shows you understood the impact |
| Apologise | ”I’m sorry the [item] stopped working.” | Owns the result without arguing about blame |
| Resolve | ”I’ll [refund or replace] it. Here’s the timeline.” | Gives an action and a date |
| Diagnose | Log the cause: carrier, warehouse, product description | The next customer is not hit by the same issue |
6. Faulty item complaint (email or phone)
Customer: The [kettle] I bought stopped working after [10] days. I'm really annoyed. This is the second time I've had a problem with you.
Agent: Hi [name], this is [agent] from [store]. Please tell me what happened and I'll listen first. [Pause until the customer finishes.]
Customer: It just won't heat up. I've followed the instructions.
Agent: That would have annoyed me too, especially after a previous problem. I'm sorry the kettle stopped working after only [10] days.
Agent: As the kettle is faulty, I can refund [amount] to your original payment method, or send a replacement now. Which would you prefer?
Customer: A refund, please.
Agent: Done. I've arranged a prepaid return label, sent to your email within [time]. Once the kettle is back with us, the refund arrives within [timeline].
Agent: I'll also log the fault against [product code] so our team can check the batch. Is there anything else I can help with?The agent followed each H.E.A.R.D. step in the table above, and the refund-or-replacement offer is the Resolve step. What the law requires for a faulty item depends on timing and on where you sell, so check the current rules for your case before you copy this offer.
For other damage cases, see our page on damaged items.
Customer service script examples for every stage of a call
A customer service script is a set of tested lines for the moments every conversation shares: the opening, verification, holds, transfers, saying no, not knowing the answer and the close. Use them as starting points and keep the facts in each line tied to the order in front of you.
Opening
"Good morning, you're through to [agent] at [store]. How can I help?"
Verification
"So I know I'm speaking to the right person, can I have your order number and the email or postcode on the order?"
"Please don't read out a card number. I don't need it."
Hold
"May I put you on hold for about [two] minutes while I check [system]?"
"Thank you for waiting. Here is what I found."
Transfer
"I'm going to pass you to [colleague], who handles [topic]. I'll give them the case first so you don't need to repeat it."
Saying no
"I can't [action], because [reason]. What I can do is [option]."
Not knowing
"I don't have that in front of me. Let me check [system] and come back to you by [time]."
Closing
"To recap: [action], by [date]. Is there anything else I can help with?"Three moments need script-specific wording, because they are where calls tend to go wrong. Replace the quick reflexes on the left with the lines on the right. This is our own guidance, not survey data.
| Avoid | Say instead | Why |
|---|---|---|
| ”Please hold." | "May I put you on hold for [two] minutes?” | Asks permission and gives a time |
| ”Let me transfer you.” (no context) | “I’ll pass you to [colleague] and give them the case first.” | The caller does not repeat themselves |
| ”I don’t know." | "Let me check [system] and come back by [time].” | Turns a dead end into a next step |
For general reply tone, our guide to better replies covers it. For ChatGPT prompts to train agents, see our AI prompts guide.
Should agents follow a customer service script word for word?
No. Script the order of the conversation and the facts that must be right, then let agents say it in their own words. Customers notice reading, and the survey evidence is old but consistent.
A 2014 Software Advice survey found that 69% of respondents said their customer service experience improved when the agent did not appear to be reading from a script, as reported by 3CLogic and CustomerThink. The primary Software Advice pages are not publicly retrievable, so these are the secondary reports. It is an older US consumer survey, so treat it as direction, not a benchmark.
| Placeholder | Where the fact comes from | Check before you send |
|---|---|---|
| [order number] and status | Store admin or order system | Is it the right order and the current status? |
| [carrier] and tracking | Fulfilment system or carrier | Does the tracking event match what you say? |
| [delivery estimate] | The carrier, not a guess | Is the date from the carrier? |
| [refund amount] and method | Order and payment record | Does it match what was paid, on which method? |
| [policy name] | Your returns or discount policy page | Is the policy current? |
| [stock] | Inventory | Is it live, per variant? |
Where does Engaige fit?
Engaige (that’s us) isn’t a phone system or a script library: it’s the e-commerce AI support agent that works inside the help desk your text conversations land in, and it works alongside these examples for email, chat, WhatsApp, SMS and social DMs. Engaige AI configures and improves the agent. At MR MARVIS, it handles over 120,000 questions a year and resolves over 60% of the conversations it’s involved in.
What sets Engaige apart here is transparent reasoning. A script hides why an agent said what it said. Engaige shows, for every answer, which systems it checked (help desk, store, carrier) and why it replied as it did, with the source: no black box, so you can see, trust and improve it. On Shopify, Engaige combines live Admin API data, pricing across markets and SKUs, inventory, orders and customers, with the marketing content from your storefront.
Hi, where is my order? It's been eight days and the tracking hasn't moved.
Shopify
Order 48219 · fulfilled
Picqer
Still processing Contradicts Shopify
DHL
Returned to depot · 3 delivery attempts
Settles it: the parcel is not with the customer
Reasoning…
Not deliverable. Reship to a confirmed address, booked in Picqer.
Suggested reply
Your parcel went back to the depot after three delivery attempts, so it will not arrive. I have booked a reshipment to the address you confirmed, arriving Thursday.
That visible reasoning also shows where information is missing or unclear. MR MARVIS uses it to improve its workflow.
Within the AI we can see the reasoning, and how it comes to a response. That shows us where information is missing or unclear, and helps us improve our workflow.
It learns from how your team actually talks, not from a fixed script. You go live in Agent Assist in hours: Engaige drafts replies and suggested actions inside your helpdesk, and nothing closes without a human. Every reply your team sends becomes training data, and Engaige AI proposes policy changes for you to approve. You then flip ticket types to autonomous one at a time, most brands in about a week, up to 90% of supported tickets.
On phone, plainly: Engaige is text-first by design: email, chat, WhatsApp, SMS and social DMs, no voice AI. A live call gives an AI seconds to reason; text gives it time to read the order, check your policy and act. If phone is a big share of your mix, a voice specialist is the honest answer. For the reasoning behind that, see conversational support.
That is the short version. The full overview is in our templates hub. To see it on your own tickets, see a demo. For where an agent fits next to people, read AI agents, and for the cost side, the ROI guide.
Frequently asked questions
What is an example of a good customer service conversation?
One that moves through the stages: open, understand, check the facts, resolve, confirm and close. In example 1, the agent checks the order status first, asks one question that gathers the full address, confirms in writing and is honest about the dispatched case.
How do you start a customer service phone call?
Greet the caller, give your name and the store’s, and offer help. Then verify who is calling with an order number plus email or postcode, and tell them not to read out a card number. Summarise the problem back before you act.
What is a customer service script?
A set of tested lines and a conversation order for the moments every call shares: opening, verification, holds, transfers, saying no, not knowing and closing. Treat it as a starting point and tie each fact to the order in front of you.
How do you handle a customer complaint on the phone?
Hear the customer out, show you understand, apologise for the impact, resolve it or agree a dated next step, then diagnose the cause so it does not recur. Read the next step back before you hang up.
Should customer service agents read from a script word for word?
No. Script the order and the facts, not every word. In older US Software Advice surveys, most respondents said the experience improved when the agent did not seem to be reading from a script.
What should you never say to a customer?
Avoid “that’s our policy”, “calm down”, a bare “I don’t know” and “please hold” with no time. Name the rule and the option, acknowledge the impact, say when you will come back and give the length of the hold.
Can an AI agent learn from your customer service conversations?
Yes, with Engaige. Every reply your team sends becomes training data, and Engaige AI proposes policy changes for you to approve. It shows its reasoning with the source.
Does Engaige handle phone calls?
No. Engaige is text-first by design: email, chat, WhatsApp, SMS and social DMs, no voice AI. If phone is a big share of your mix, a voice specialist is the honest answer, and these phone examples are for your human team.
