Table of Contents
- Key takeaways
- What is a customer service chatbot example?
- Why do the sample conversations follow these rules?
- Worked example: building a customer service chatbot for an online store
- Step 1: write the scope list
- Step 2: connect the content
- Step 3: write the opening message
- Step 4: set the rules
- Step 5: run a full test conversation
- Step 6: turn the test into a checklist
- Sample customer service chatbot conversations by industry
- 1. Ecommerce: a shipping question
- 2. Banking: a lost card
- 3. Healthcare: appointment preparation
- 4. Travel and hospitality: a late checkout
- 5. Insurance: filing a claim
- 6. Telecom and utilities: an outage
- 7. Education: an enrollment deadline
- 8. Software: a password reset
- Customer service chatbot template by industry
- Chatbot response examples for common moments
- How should a customer service chatbot decide to answer or hand off?
- What to decide when developing an AI chatbot for a customer service role
- Common mistakes in customer service chatbot conversations
- Which tool can build a customer service chatbot like these?
- Frequently asked questions
- Basics
- What is an example of a customer service chatbot?
- What does a customer service chatbot conversation look like?
- What is the difference between a chatbot example and a chatbot template?
- Writing replies
- What is a good chatbot response to "hi"?
- How long should a customer service chatbot reply be?
- What should a chatbot say when it does not know the answer?
- Handoff and limits
- When should a customer service chatbot hand off to a human?
- Can a customer service chatbot check order status?
- Can a healthcare chatbot give medical advice?
- Building one
- How do you build a customer service chatbot from these examples?
- Do you have to tell customers they are talking to a chatbot?
- Sources
Created by
Do not index
Created time
Sep 15, 2023 12:03 PM
A customer service chatbot example is a short, scripted conversation that shows how a bot greets a customer, answers from the business's own content, links the source, and hands off to a person when it should. This page walks through one full example for an online store, then gives sample conversations for eight industries you can copy.
TL;DR
- A good customer service chatbot conversation follows six moves: disclose the bot, set the scope, answer from a source, link the page, confirm the next step, and hand off cleanly.
- The worked example below builds one chatbot for an online store, from the scope list to a full 8-turn conversation with the rule behind each turn.
- Eight sample conversations cover ecommerce, banking, healthcare, travel, insurance, telecom, education, and software. Each one names its content source and its handoff rule.
- Content alone cannot answer questions about one customer's order, account, or claim. Those need a system that reads the record, or a person.
- In a 2026 Gartner survey, only 27% of customers said they would try a chatbot again after a bad experience, so a narrow scope that works beats a wide one that fails.
Key takeaways
Question | Short answer |
What makes a good chatbot reply? | The answer in the first sentence, two or three lines long, with a link to the page it came from |
What should the first message say? | That the visitor is talking to an AI assistant, plus three or four topics it can help with |
When should the bot hand off? | When the customer asks, when the content has no answer, when money, health, or a legal promise is at stake, or after two failed tries |
What can a content-trained bot not do? | Look up one customer's order, account, or claim without a connected system |
Which industries are covered here? | Ecommerce, banking, healthcare, travel, insurance, telecom, education, and software |
What is a customer service chatbot example?
A customer service chatbot example is a sample conversation between a customer and a bot that shows what the bot says, where its answer comes from, and when it passes the customer to a person. It is a script for one job, such as a return, a billing question, or a booking change.
Examples help in two ways. They show a team what "good" looks like before anything is built. They also become test cases: once the bot is live, the team can type the same customer messages and check the replies.
This page is about the conversations themselves. For real brands and their bots, see the companies using chatbots for customer service list and the support chatbot examples grouped by job. For B2B sales and service bots, see the B2B chatbot examples.
Why do the sample conversations follow these rules?
The sample conversations follow six rules because customers give a chatbot very few chances. Gartner surveyed 3,566 B2B and B2C customers in February and March 2026. Only 27% said they would try a chatbot again after a negative experience.
The same survey found that 87% of customers say access to a human agent is essential when companies use generative AI in customer service. Gartner's advice was to "prioritize reliability over reach".
Two other facts shape every example below:
- Disclosure is now a legal duty in the EU. Article 50 of the EU AI Act has applied since 2 August 2026. Providers must design chatbots so people are told they are interacting with an AI system, unless that is obvious.
- A business answers for what its bot says. In Moffatt v. Air Canada, a British Columbia tribunal held the airline liable after its website chatbot gave wrong advice about bereavement fares. The tribunal wrote that it "makes no difference whether the information comes from a static page or a chatbot".
Worked example: building a customer service chatbot for an online store
This worked example builds one customer service chatbot for a fictional online shoe store. It shows the decisions in the order a team makes them: scope, content, opening message, rules, and a full test conversation.
Step 1: write the scope list
The scope list names the questions the bot answers on day one. For the shoe store, the team picks five topics that already have a help page:
- Shipping times and costs
- Returns and exchanges
- Sizing and fit
- Order changes before dispatch
- Store hours and contact details
Order status is left out of scope. Order status needs the store's order system, and this first version of the bot only reads content.
Step 2: connect the content
The bot is trained on the pages that answer those five topics: the shipping page, the returns policy, the size guide, and the contact page. The team fixes one contradiction before launch. The FAQ said returns were accepted for 30 days and the policy page said 60, so the FAQ was corrected to 60.
Step 3: write the opening message
The opening message says the visitor is talking to an AI assistant and names the topics:
- Bot: Hi, I am the store's AI assistant. I can help with shipping, returns, sizing, and changes to an order. What do you need?
Step 4: set the rules
The bot gets four plain-language rules:
- Answer only from the trained pages, and link the page used.
- If the pages have no answer, say so and offer a person.
- Never promise a refund, a discount, or an exception to the policy.
- Hand off when the customer asks for a person, is upset, or the bot has failed twice.
Step 5: run a full test conversation
The table below is the full 8-turn test conversation. Each turn shows what the customer typed, what the bot did, and which rule produced that reply.
Turn | Customer says | Bot replies | Rule applied |
1 | hi | Hi, I am the store's AI assistant. I can help with shipping, returns, sizing, and order changes. What do you need? | Disclose the bot and set the scope |
2 | can i return shoes i wore outside once | Returns are accepted within 60 days for unworn shoes in the original box. Shoes worn outside are not eligible. The full policy is on the returns page. | Answer in the first sentence and link the source |
3 | what if they hurt my feet though | The returns policy does not list an exception for worn shoes. A team member can look at your case. Do you want me to connect you? | No answer in the content, so offer a person |
4 | not yet. do these run small | The size guide says this style runs half a size small. Most customers order half a size up. | Answer from the size guide |
5 | ok can i swap my order to a 9.5 | Orders can be changed until they are dispatched. I cannot see your order, so a team member needs to make the change. Shall I pass this on with your request? | Account-specific request, so hand off |
6 | yes | Sure. Please share the email on your order so the team can find it. | Collect only the data the handoff needs |
7 | (types the email on the order) | Thanks. Your request and this conversation have gone to the team. They reply by email within the hours on the contact page. | Hand off with the transcript and set expectations |
8 | thanks | You are welcome. Is there anything else about shipping, returns, or sizing? | Confirm the next step and close |
The test passes because the bot never guessed. Turn 3 and turn 5 are the important ones: the bot had no policy to quote and no access to the order, so it offered a person both times.
Step 6: turn the test into a checklist
After launch, the same eight messages become a regression test. The team reruns them after every change to the content or the rules, and checks three things:
- Every factual reply links a trained page.
- The bot declines the worn-shoe exception and the order change.
- Every handoff carries the conversation to the team.
Sample customer service chatbot conversations by industry
The sample customer service chatbot conversations below cover eight industries. Every business is fictional and every conversation is illustrative. Each one uses the same template: the job, the conversation, the content it needs, and the handoff rule.
1. Ecommerce: a shipping question
An ecommerce chatbot answers the questions that fill an online store's inbox: shipping, returns, and sizing. This conversation handles a delivery date before purchase.
- Customer: if i order today will it arrive by friday
- Bot: Orders placed before 2pm ship the same day. Standard delivery takes 3 to 5 business days, so Friday is likely but not guaranteed. Express delivery arrives in 1 to 2 business days. Both options are on the shipping page.
- Customer: how much is express
- Bot: Express delivery costs $12. It is free on orders over $100.
- Content it needs: the shipping page with cutoff times, delivery windows, and prices.
- Handoff rule: hand off any question about a specific order that has already shipped.
More ecommerce patterns are in the ecommerce chatbot examples.
2. Banking: a lost card
A banking chatbot answers product and process questions and routes anything involving money or fraud to the bank. This conversation handles a lost card.
- Customer: i lost my debit card
- Bot: I am sorry. You can freeze the card right now in the mobile app under Cards, then Freeze card. To report it lost and order a new one, call the number on the lost card page, which is open 24 hours.
- Customer: i think someone used it
- Bot: For possible fraud, please speak to the fraud team now. I am connecting you to a person, and this conversation goes with you.
- Content it needs: the lost card page, the app help pages, and the fraud contact page.
- Handoff rule: hand off at the first mention of fraud, a disputed charge, or a balance question.
The CFPB's report on chatbots in consumer finance describes customers stuck in "doom loops" with no offramp to a human. The second bot reply above is the offramp.
3. Healthcare: appointment preparation
A healthcare chatbot answers administrative questions and never gives medical advice. This conversation handles preparation for a clinic visit.
- Customer: do i need to fast before my blood test tomorrow
- Bot: The clinic's blood test page says some tests need 8 to 12 hours of fasting and others do not. Your appointment letter says which applies to you. If you are unsure, the clinic's nurse line can confirm.
- Customer: i also have chest pain since this morning
- Bot: Chest pain needs urgent attention. Call your local emergency number now. I am an AI assistant and cannot assess symptoms.
- Content it needs: the patient preparation pages, opening hours, and the contact page.
- Handoff rule: route any symptom to emergency or clinical staff at once, without trying to answer.
A US clinic covered by HIPAA that lets patients share health details in the chat needs a signed business associate agreement with its chatbot vendor. The HIPAA compliant AI chatbot guide compares the options.
4. Travel and hospitality: a late checkout
A hotel or travel chatbot answers policy questions around a booking: check-in times, baggage, and changes. This conversation handles a late checkout request.
- Customer: can i check out late on sunday
- Bot: Standard checkout is 11am. Late checkout until 2pm costs $30 and depends on availability. The front desk confirms it on the morning of departure.
- Customer: can you book it for me
- Bot: I cannot change bookings. I can pass your request to the front desk with your name and room number. Would you like that?
- Content it needs: the hotel policies page and the amenities page.
- Handoff rule: hand off any request that changes a reservation or a charge.
5. Insurance: filing a claim
An insurance chatbot explains how to file a claim and what documents are needed. It never says whether a claim is covered. This conversation handles a car accident claim.
- Customer: how do i file a claim for a car accident
- Bot: You can file a car claim online or by phone. You will need your policy number, the date and place of the accident, photos of the damage, and the other driver's details if there was one. The steps are on the claims page.
- Customer: will my policy cover the rental car
- Bot: Rental car cover depends on your specific policy. A claims adviser can check your policy and confirm. Do you want me to connect you?
- Content it needs: the claims guide, the document checklist, and the policy summary pages.
- Handoff rule: hand off every coverage question. Coverage depends on one customer's policy, and a wrong answer creates a promise.
6. Telecom and utilities: an outage
A telecom or utility chatbot answers outage, billing, and setup questions. This conversation handles a broadband outage.
- Customer: my internet is down
- Bot: Sorry about that. First, check the service status page for outages in your area. If there is none, restart the router by unplugging it for 30 seconds. The full steps are on the troubleshooting page.
- Customer: did that, still down
- Bot: Thanks for trying those steps. The next step is a line test, which a technician runs. I am passing you to the support team with what you have already tried.
- Content it needs: the troubleshooting guide and the service status page.
- Handoff rule: hand off after the documented steps fail, and pass along the steps already tried.
7. Education: an enrollment deadline
A school or university chatbot answers admissions, fees, and deadline questions. This conversation handles an application deadline.
- Customer: when is the deadline for fall applications
- Bot: The fall application deadline is listed on the admissions dates page. Late applications are reviewed only if places remain. Do you want the list of required documents?
- Customer: yes
- Bot: You need a completed form, transcripts, and two references. International applicants also need proof of English. The checklist is on the admissions page.
- Content it needs: the admissions dates page, the fees page, and the document checklist.
- Handoff rule: hand off questions about one applicant's file or a decision.
The bot points to the dates page instead of stating the date. If the date changes and the content refreshes, the reply stays correct.
8. Software: a password reset
A software company's chatbot answers setup, account, and billing questions from its help center. This conversation handles a password reset.
- Customer: i cant log in
- Bot: You can reset your password from the sign-in page with Forgot password. The reset email arrives within a few minutes. Check the spam folder if it does not.
- Customer: the email never came and i have a demo in 10 min
- Bot: That is urgent, so I am passing you to the support team now with this conversation. They can verify your account and send a new link.
- Content it needs: the help center articles on sign-in, billing, and setup.
- Handoff rule: hand off when the documented fix fails and the customer is under time pressure.
IT and internal help desk patterns are covered in the IT support chatbot guide.
Customer service chatbot template by industry
This customer service chatbot template lists, for each industry, the questions to cover first, the content to train on, and the point where a person takes over. Use the row closest to your business as a starting scope list.
Industry | Questions to cover first | Train on | Hand off when |
Ecommerce | Shipping, returns, sizing | Shipping page, returns policy, size guide | The question is about one order |
Banking | Lost cards, app steps, fees | Help pages, fee schedule, contact pages | Fraud, disputes, or balances come up |
Healthcare | Hours, preparation, insurance accepted | Patient information pages | Any symptom is mentioned |
Travel and hospitality | Check-in, baggage, amenities | Policies and amenities pages | A booking or charge must change |
Insurance | How to claim, documents, policy types | Claims guide, document checklist | Coverage for one policy is asked |
Telecom and utilities | Outages, bills explained, setup | Troubleshooting guide, status page | Documented steps fail |
Education | Deadlines, fees, documents | Admissions and fees pages | One applicant's file is involved |
Software | Sign-in, billing, setup | Help center articles | The fix fails or the customer is blocked |
Ready-made prompt and flow templates are in the chatbot template guide and the chatbot templates library.
Chatbot response examples for common moments
Chatbot response examples are most useful for the moments every bot meets, whatever the industry. The table compares a weak reply and a strong reply for six of them.
Moment | Weak response | Strong response |
The customer says "hi" | Hello! How can I help you today? | Hi, I am an AI assistant. I can help with orders, returns, and sizing. What do you need? |
The content has no answer | I think the warranty is probably two years. | I could not find that in the warranty information. A team member can check. Shall I connect you? |
The question is off topic | Here is a poem about shoes. | I can only help with questions about this store. Is there anything about an order I can help with? |
The customer is angry | I understand. Please rephrase your question. | I am sorry this has been frustrating. I am passing you to a person now with this conversation. |
The customer asks for a person | Before that, can I try to help? | Of course. I am connecting you now. The team replies within the hours on the contact page. |
The conversation ends | Goodbye! | Glad that helped. If you have another question about shipping or returns, ask any time. |
Opening messages for specific jobs are collected in the welcome message section of the support chatbot examples. Tone and persona rules are in the chatbot design guide.
How should a customer service chatbot decide to answer or hand off?
A customer service chatbot should answer when the trained content covers the question and nothing important depends on one customer's record. It should hand off in every other case. Four questions decide it.
If the customer... | Then the bot... | Why |
Asks for a person | Hands off at once | Asking them to try the bot again is the doom loop |
Asks something a trained page answers | Answers and links the page | This is the work the bot is for |
Asks something no page answers | Says so and offers a person | A guess can become a liability |
Asks about one order, account, or claim | Reads the record through a connected system, or hands off | Public content cannot answer a private question |
Raises fraud, a symptom, or a coverage question | Routes to a person with the transcript | A wrong answer costs money, health, or a promise |
Is upset or the bot failed twice | Hands off with the transcript | A third attempt rarely works |
More handoff rules, and the metrics to watch after launch, are in the chatbot best practices guide.
What to decide when developing an AI chatbot for a customer service role
When developing an AI chatbot for a customer service role, decide six things before writing a single reply. Each one maps to a step in the worked example above.
- Scope. List the questions the bot answers on day one. Leave out anything that needs account data the bot cannot read.
- Sources. Pick the pages and documents the bot is trained on, and fix contradictions between them.
- Disclosure and opening. Write a first message that says the bot is an AI and names its topics.
- Rules. Write the rules for declining, for commitments the bot must never make, and for handoff.
- Handoff path. Decide who receives escalated conversations, through which tool, and in what hours.
- Test set. Write 10 to 20 real customer messages, including ones the bot should decline, and rerun them after every change.
Which customer service tasks a chatbot can take on in each department is covered in the customer service AI use cases guide. Cross-industry use cases are in chatbot use cases.
Common mistakes in customer service chatbot conversations
The most common mistakes in customer service chatbot conversations break one of the six moves. Five show up often.
- Hiding that it is a bot. In the EU, disclosure has been required since 2 August 2026.
- Stating policies without a link. The customer cannot check the answer, and the team cannot spot a wrong one.
- Making promises. Refunds, discounts, coverage, and exceptions belong with a person or a fixed process.
- Blocking the route to a person. "Before that, can I try to help?" after a clear request is the doom loop.
- Hard-coding facts that change. Dates, prices, and hours belong on a page the bot reads, so a content refresh keeps replies current.
Which tool can build a customer service chatbot like these?
A customer service chatbot like the examples above can be built with three kinds of tool, and the right one depends on whether the bot must read customer records.
If you need | Pick | Why |
Answers from your website, help center, and documents, with a handoff to a person | A content-trained AI chatbot | It covers the policy and how-to questions in all eight examples |
Order status, account changes, or claim lookups inside the chat | An AI agent inside your helpdesk or store platform | It can read and act on the customer record |
A fixed sequence that must never vary, such as an eligibility check | A rule-based flow builder | Every step is scripted and auditable |
Patient details shared in the chat | A vendor that signs a business associate agreement | HIPAA covered entities need one before patient data reaches the vendor |
The buyer's guide to customer service AI chatbot solutions compares named products in each category.
SiteGPT is one content-trained option. It trains on website pages, files, and sources such as Google Drive, Notion, Dropbox, OneDrive, Box, GitHub, and YouTube, and supports 95+ languages. Human support is available on every plan. By default the AI escalates when the visitor asks for a person, when the situation is urgent, or when the visitor is frustrated after an answer that did not help.
SiteGPT has real limits for these examples. It does not read order or account records, so the order change in turn 5 of the worked example would still go to a person. It has no visual flow builder and no Intercom or HubSpot integration today. A HIPAA business associate agreement is available on the Enterprise plan only, and training content must stay free of patient data even then.
Starter costs $59 a month billed monthly, or $39 a month billed yearly ($468 a year). Paying yearly saves $240 a year on Starter, about 34%. Every self-serve plan has a 7-day free trial that requires a credit card.
Frequently asked questions
Basics
What is an example of a customer service chatbot?
An example of a customer service chatbot is an online store's assistant that answers shipping, returns, and sizing questions from the store's help pages and hands order changes to a person. The worked example on this page shows a full 8-turn conversation for that bot.
What does a customer service chatbot conversation look like?
A customer service chatbot conversation starts with a message that says the bot is an AI and names its topics. Each reply gives the answer first, links the source page, and asks one short follow-up question. The conversation ends with a handoff or a closing check.
What is the difference between a chatbot example and a chatbot template?
A chatbot example is a finished sample conversation that shows what the bot says. A chatbot template is a blank starting point, such as a prompt or a flow, that a team fills in with its own details.
Writing replies
What is a good chatbot response to "hi"?
A good chatbot response to "hi" says the bot is an AI assistant and names three or four things it can help with, then asks what the customer needs. For example: "Hi, I am an AI assistant. I can help with orders, returns, and sizing. What do you need?"
How long should a customer service chatbot reply be?
A customer service chatbot reply should give the answer in the first sentence and stay within two or three lines, with a link to the page that has the detail.
What should a chatbot say when it does not know the answer?
A chatbot that does not know the answer should say it could not find the information and offer to connect the customer with a person. It should never guess.
Handoff and limits
When should a customer service chatbot hand off to a human?
A customer service chatbot should hand off when the customer asks for a person, when the trained content has no answer, when money, health, or a legal promise is at stake, when the customer is upset, or after two failed tries.
Can a customer service chatbot check order status?
A customer service chatbot can check order status only if it is connected to the order system. A chatbot trained only on website content can explain shipping policies but cannot see one customer's order.
Can a healthcare chatbot give medical advice?
A healthcare chatbot should not give medical advice. It should answer administrative questions, such as hours and appointment preparation, and route any symptom to emergency services or clinical staff.
Building one
How do you build a customer service chatbot from these examples?
Build a customer service chatbot from these examples by writing a scope list, connecting the pages that answer those questions, writing an opening message and rules, and testing with real customer messages. The AI FAQ chatbot guide covers the build step by step.
Do you have to tell customers they are talking to a chatbot?
In the EU, yes. Article 50 of the EU AI Act has applied since 2 August 2026 and requires that people are informed they are interacting with an AI system, unless that is obvious.
Sources
- Checked on 18 October 2026.
Written by
