How to Use a Chatbot for Customer Service: 4 Jobs to Roll Out

The four jobs a chatbot does in customer service every day, the order to roll them out, where it must hand off to a person, and five numbers that show whether it works.

How to Use a Chatbot for Customer Service: 4 Jobs to Roll Out
Created by
Do not index
Created time
Nov 2, 2023 03:39 PM
To use a chatbot for customer service, give it four jobs in this order: answer the questions your team gets every day, cover the hours nobody is online, triage conversations and hand the hard ones to a person, and point order or account questions to the page or person that can resolve them.
TL;DR
  • Start with one job. Train the chatbot on your help content and the 20 questions your inbox gets most, then test it on those 20 before launch.
  • Add after-hours cover next. A team that works 40 hours a week leaves 128 of the 168 hours in a week (76%) without a person online.
  • Write handoff rules before you widen the scope. In a 2026 Gartner survey of 3,566 customers, 87% said access to a human agent is essential.
  • Keep order and account questions honest. A chatbot trained on content can explain the returns policy, but it cannot see one customer's order unless it is connected to the order system.
  • Track five numbers every week: resolution rate, handoff rate, fallback rate, answer rating, and time to a human reply.

Key takeaways

Job
What the chatbot does
What it should leave to a person
Number to watch
1. Answer repeat questions
Answers from your help center, policies, and product pages
Questions your content does not cover
Resolution rate
2. Cover after-hours
Answers when the team is offline and sends an away message on handoff
Anything that cannot wait for the next working day
Handoffs answered by the next working day
3. Triage and hand off
Spots requests for a person, urgency, and frustration, then passes the transcript
Refunds, complaints, billing disputes
Handoff rate and time to a human reply
4. Order and account questions
Explains policies and links the tracking or account page
Any answer that needs a lookup the chatbot cannot do
Fallback rate on order questions
Summary graphic of the four jobs a chatbot does in customer service, rolled out in order: answer repeat questions, cover hours nobody is online, triage and hand off, and point order and account questions
Summary graphic of the four jobs a chatbot does in customer service, rolled out in order: answer repeat questions, cover hours nobody is online, triage and hand off, and point order and account questions

What does a chatbot do in customer service?

A chatbot in customer service answers customer questions in a chat window, around the clock, from content your team already wrote. A modern AI chatbot reads your help articles, policies, and product pages, then replies in plain sentences instead of fixed menu buttons.
The chatbot does not replace the team. It takes the repeat questions, so people spend their time on the conversations that need judgment. When it cannot help, it hands the conversation to a person with the transcript attached.
This guide covers the day-to-day jobs and how to roll them out. To build the chatbot itself from a help center, follow the customer service chatbot build tutorial. For the wider list of AI jobs in a service team, such as ticket routing and reply drafting, see customer service AI use cases.

Why should a chatbot start with a narrow set of jobs?

A chatbot should start narrow because customers give it few second 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.
Gartner's advice was to "prioritize reliability over reach". In practice, that means one job at a time, each tested before the next one starts. A chatbot that answers 20 questions correctly builds more trust than one that tries 200 and gets 40 wrong.

Which questions should a chatbot answer, and which should go to a person?

A chatbot should answer questions whose answer is written down and the same for every customer. A person should take questions that depend on one customer's record, carry a cost if answered wrong, or come from an upset customer.
Question type
Example
Who answers
Why
Policy or how-to
"What is your return window?"
Chatbot
The answer is on a page and is the same for everyone
Product detail
"Does the annual plan include phone support?"
Chatbot
The answer is on the pricing or product page
One customer's order
"Where is order 48213?"
Chatbot links the tracking page, or a person looks it up
The answer needs a lookup in the order system
Account change
"Change the email on my account."
Person, or a link to the account settings page
The change needs the customer's identity confirmed
Refund or billing dispute
"I was charged twice."
Person
A wrong answer costs money
Complaint
"This is the third time it broke."
Person
The customer needs a person, not an article
Decision tree for deciding whether a chatbot answers a customer question or hands it to a person, based on whether the answer is written down, depends on one customer's order, carries a cost if wrong, or comes from a frustrated visitor
Decision tree for deciding whether a chatbot answers a customer question or hands it to a person, based on whether the answer is written down, depends on one customer's order, carries a cost if wrong, or comes from a frustrated visitor

How to use a chatbot for customer service, job by job

Use a chatbot for customer service by switching on one job, checking it against a pass mark, and only then adding the next. Each job below has a setup, a check, and a "done when" line.

Job 1: Answer the questions your team gets every day

Answering repeat questions is the first job because it carries the most volume and the least risk. Pull the last 100 chats or emails and list the 20 questions that come up most. These are usually hours, shipping, returns, pricing, and setup steps.
  1. Train the chatbot on the pages that answer those 20 questions: help articles, policies, the pricing page, and product pages.
  1. Write any missing answers as short pages or text snippets. A chatbot cannot answer what nobody wrote down.
  1. Ask the chatbot all 20 questions in your own words. Mark each answer correct, partly correct, or wrong.
  1. Fix the content behind every wrong answer, then run the 20 questions again.
Done when: at least 16 of the 20 answers (80%) are correct and every answer links or matches a real page. The click-by-click setup for this job is in the build tutorial.

Job 2: Cover the hours nobody is online

After-hours cover is the second job because the chatbot already answers the common questions well. A team that works 9 to 5, Monday to Friday, is online for 40 of the 168 hours in a week. The other 128 hours, 76% of the week, have nobody in the chat.
  1. Set the team's working hours and time zone in the chatbot.
  1. Choose what happens outside those hours: the AI answers, or the team answers later.
  1. Write an away message that says when a person will reply. Do not promise a reply time the team cannot keep.
  1. Check the overnight handoffs every morning and reply to each one.
Done when: every conversation handed off outside working hours gets a reply by the end of the next working day. The cost case for round-the-clock cover is in the 24/7 chatbot guide.

Job 3: Triage conversations and hand off the hard ones

Triage is the third job: the chatbot decides which conversations it keeps and which go to a person. Gartner found that 87% of customers say access to a human agent is essential when a company uses generative AI in customer service.
  1. Turn on handoff so a visitor can ask for a person in one click or one message.
  1. Write escalation rules in plain language. Name the topics that always go to a person, such as refunds, billing disputes, legal threats, and complaints.
  1. Make sure the transcript travels with the handoff, so the customer does not repeat the question.
  1. Decide who on the team picks up handed-off chats, and during which hours.
Done when: a visitor reaches a person in one step, and no test conversation about a refund stays with the bot. The rules for when to escalate are covered in chatbot best practices.

Job 4: Point order and account questions to the right place

Order and account questions are the fourth job because they need care. A chatbot trained on your content can answer "How long does shipping take?" It cannot answer "Where is my order?" unless it is connected to the order system and can confirm who the customer is.
  1. Train the chatbot on every policy that order questions touch: shipping times, returns, exchanges, and cancellations.
  1. For live order status, have the chatbot link the tracking page or the account page.
  1. For account changes and anything it cannot look up, hand off and ask the customer for the order number first.
  1. If your tool offers an order lookup, test it on 10 real orders before you switch it on for customers.
Done when: no test conversation ends with the chatbot guessing an order status. Every order question gets a policy answer, a link, or a person.

Where should you start? A decision matrix

Start with the job that removes the most repeat work without raising the cost of a wrong answer. For most teams, that is job 1.
If you need
Start with
Why
Fewer "same question again" tickets
Job 1: answer repeat questions
The top 20 questions usually carry the most volume
Answers for customers in other time zones
Job 1, then job 2 in the same week
After-hours cover only works once the answers are right
Faster replies on hard cases
Job 3: triage and hand off
The team sees the transcript and skips the first questions
Fewer "where is my order" chats
Job 4, starting with a link to the tracking page
A link resolves the question without an order lookup
A chatbot in a regulated service, such as healthcare
Job 1 with strict handoff rules
Keep the bot on public information and route anything personal to a person

How do you measure a customer service chatbot?

Measure a customer service chatbot with five numbers, read every week from the chatbot's conversation history and analytics.
Metric
How to calculate it
What a good trend looks like
Resolution rate
Conversations closed without a person, divided by all conversations
Rising week over week
Handoff rate
Conversations handed to a person, divided by all conversations
Stable, with handoffs on the topics you chose
Fallback rate
Conversations where the bot said it did not know, divided by all conversations
Falling as you add missing content
Answer rating
Helpful ratings, divided by all rated answers
Rising, with low ratings read one by one
Time to a human reply
Time from handoff to the first reply from a person
Inside the time your away message promises
Read 20 transcripts each week as well. The numbers show where the chatbot fails. The transcripts show why.

Worked example: one month of chatbot numbers

A worked example shows how the five numbers turn into a to-do list. Take a store whose chatbot handles 1,200 conversations in one month.
Bar chart of a worked example: 1,200 chatbot conversations in one month, with 780 resolved by the chatbot, 264 handed off to a person, and 156 ending in a fallback answer
Bar chart of a worked example: 1,200 chatbot conversations in one month, with 780 resolved by the chatbot, 264 handed off to a person, and 156 ending in a fallback answer
  • Resolved: 780 of 1,200 conversations, a resolution rate of 65%.
  • Handed off: 264 conversations, a handoff rate of 22%. Over 22 working days, that is 12 handoffs a day for the team.
  • Fallback: 156 conversations, a fallback rate of 13%. These are the questions the content does not answer yet.
The fallback group is the fastest win. If the team writes pages for the three most common missing topics and that halves the fallback group, 78 more conversations resolve. The resolution rate then rises from 65% to 71.5% (858 of 1,200), with no change to the handoff load.

Common mistakes when using a chatbot for customer service

The most common mistake is launching every job at once. The chatbot then fails on the hardest questions first, and customers stop using it.
  • Hiding the route to a person. Customers who cannot reach a person leave the chat and email instead, so the work comes back.
  • Letting the chatbot guess order status. A confident wrong answer about an order costs more than a link to the tracking page.
  • Never updating the content. Prices, policies, and shipping times change. Refresh the trained content whenever the source pages change.
  • Reading only the dashboard. Low ratings and fallback answers explain themselves only in the transcripts.
  • Promising instant replies at night. The away message should name the real reply time.

Where SiteGPT fits

SiteGPT is a chatbot that answers customer questions from your own content: website pages, sitemaps, files, and connected sources such as Google Drive, Notion, and YouTube. It answers in 95+ languages, and its auto refresh keeps trained content in step with your site on the Growth plan and above.
The four jobs map to settings SiteGPT documents. Human support is on every plan. The AI escalates by default 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. You can replace those defaults with your own rules, and the Follow Working Hours mode sets an away message for after-hours handoffs. Growth adds knowledge gaps, a weekly email digest, and chat integrations such as Slack and Crisp. Scale adds AI conversation insights and the Zendesk AI Agent.
The limits are real. Starter has basic analytics and no chat integrations. SiteGPT has no Intercom or HubSpot integration today. For live order status, the chatbot should link your tracking page or hand off to a person.
Starter costs $59 a month billed monthly, or $39 a month billed yearly ($468 a year). Growth costs $129 a month billed monthly, or $79 a month billed yearly ($948 a year). Every self-serve plan starts with a 7-day free trial. It needs a credit card, and you can cancel before it ends. To compare SiteGPT with other tools first, see the best chatbots for customer service.

Frequently asked questions

Getting started

How do you use a chatbot for customer service?

You use a chatbot for customer service by giving it one job at a time. Start with the 20 questions your team answers most, then add after-hours cover, handoff rules, and order or account routing. Test each job before you add the next.

What should a customer service chatbot answer first?

A customer service chatbot should answer the questions whose answers are written down and the same for everyone. Hours, shipping times, return windows, pricing, and setup steps are the usual first set.

How long does it take to roll out a customer service chatbot?

A focused rollout of job 1 takes about a week: a day to list the top 20 questions, a day or two to train and fix content, and a few days of testing. Each later job adds roughly a week of setup and checks.

Handoff and after-hours

Can a chatbot hand a conversation to a person?

Yes. A customer service chatbot should let a visitor ask for a person in one click or one message, and it should pass the transcript along. Write rules for the topics that always go to a person, such as refunds and complaints.

What should a chatbot do when nobody is online?

When nobody is online, the chatbot should keep answering the questions it can, and send an away message when a visitor asks for a person. The away message should say when the team will reply.

Orders and accounts

Can a chatbot tell a customer where their order is?

A chatbot can give order status only if it is connected to the order system and can confirm who the customer is. A chatbot trained only on content should link the tracking page or hand off with the order number.

Should a chatbot change account details?

A chatbot should not change account details unless it can confirm the customer's identity. Link the account settings page or hand the conversation to a person.

Measurement

How do you know if a customer service chatbot is working?

A customer service chatbot is working when its resolution rate rises, its fallback rate falls, and customers who ask for a person reach one in the time the away message promises. Read 20 transcripts a week to see the reasons behind the numbers.

What is a good chatbot resolution rate?

A good resolution rate is one that rises week over week on the jobs you switched on. Compare the chatbot with its own past numbers, because the mix of questions differs from one business to the next.

Sources

All sources were opened on 3 November 2026.

Give Your Customers The Experience That They Deserve

Create A Chatbot In Minutes, Today

Create Your Chatbot Now

Written by

Bhanu Teja P
Bhanu Teja P

Founder @ SiteGPT.ai & SourceSync.ai

Related posts

Customer Support Automation: 8 Workflows by Ticket Type
•

Customer Support Automation: 8 Workflows by Ticket Type

Eight customer support automation workflows by ticket type, from auto-replies and routing to AI answers and handoff. Each lists its trigger, action, tool and what can go wrong.

Customer Support Chatbot Service for Websites: How to Pick
•

Customer Support Chatbot Service for Websites: How to Pick

How to pick a customer service chatbot for your website by who answers chats today and where your answers live, with prices for Tidio, SiteGPT, Crisp and Zendesk.

Customer Service AI Chatbot Development Service: Build or Buy
•

Customer Service AI Chatbot Development Service: Build or Buy

Should you hire a chatbot development service or use a no-code platform? Four routes compared on first-year cost, three-year cost and time to launch, from an agency to SiteGPT.

Customer Support Chatbot: Benefits, Limits, and ROI Math
•

Customer Support Chatbot: Benefits, Limits, and ROI Math

The business case for a customer support chatbot: six benefits, five limits, the cost per answered question, a break-even worked example, and when to skip a chatbot.

16 Companies Using Chatbots for Customer Service (2026)
•

16 Companies Using Chatbots for Customer Service (2026)

16 well-known companies that use chatbots for customer service, what each chatbot does, the results each company published, and the pattern a smaller business can copy.

Customer Service Chatbot Example: Sample Conversations by Industry
•

Customer Service Chatbot Example: Sample Conversations by Industry

A worked customer service chatbot example for an online store, plus sample conversations for eight industries with the content and handoff rule each one needs.

Best AI Customer Service Software (2026): 6 Tools by Job
•

Best AI Customer Service Software (2026): 6 Tools by Job

AI customer service software does three jobs: AI agents inside a helpdesk, knowledge-trained answer bots and full AI agent platforms. This guide ranks six tools by job, with prices and G2 ratings checked on 3 November 2026.

Customer Service AI Chatbot Solution for Websites: How to Choose
•

Customer Service AI Chatbot Solution for Websites: How to Choose

A customer service AI chatbot for websites answers visitors from your own content and hands off to a person. Compare the four types, pricing units and first-year costs.

How AI Improves Customer Experience: 6 CX Metrics It Moves
•

How AI Improves Customer Experience: 6 CX Metrics It Moves

How AI improves customer experience across six CX metrics, from first response time to customer effort, with evidence from NBER, Klarna and Gartner.

Personalized AI Customer Support Solutions: How They Work
•

Personalized AI Customer Support Solutions: How They Work

Personalized AI customer support works at four levels: content, page, person, and live account. See what each level needs, which tools fit, and where to start.

AI Chatbot for Customer Service: 7 Best Platforms (2026)
•

AI Chatbot for Customer Service: 7 Best Platforms (2026)

The best AI chatbot for customer service depends on where your answers live. Seven platforms ranked, with prices and G2 ratings read on 18 October 2026.

AI Tech Support: 5 Best Technical Support Chatbots (2026)
•

AI Tech Support: 5 Best Technical Support Chatbots (2026)

Five AI chatbots for technical support, ranked on docs-grounded answers, ticketing, engineering escalation, price, and G2 ratings. Zendesk leads for ticketing, and SiteGPT is the flat-priced answer layer in front of the queue.

9 FAQ and Customer Service Chatbot Examples by Job (2026)
•

9 FAQ and Customer Service Chatbot Examples by Job (2026)

Nine real FAQ and customer service chatbot examples grouped by support job, with what each bot does, published results, sample replies, and two public failures.

B2B Chatbot Examples: 8 Real Companies and Their Use Cases
•

B2B Chatbot Examples: 8 Real Companies and Their Use Cases

Eight B2B chatbots at seven named companies, grouped by job: lead qualification, demo booking, account-based engagement, customer service, and documentation help.

AI-Enabled Customer Experience: Journey Map and Maturity Model
•

AI-Enabled Customer Experience: Journey Map and Maturity Model

What AI-enabled customer experience means at each stage of the customer journey, with a four-stage maturity model, sourced evidence, and a guide to where to start.

Conversational AI for Better Customer Experiences: 5 Platforms (2026)
•

Conversational AI for Better Customer Experiences: 5 Platforms (2026)

How to choose conversational AI for customer experience: five platforms ranked with best-for labels, verified pricing, a side-by-side table and a decision matrix.

Customer Support Chatbot: What It Is and How It Works
•

Customer Support Chatbot: What It Is and How It Works

A customer support chatbot answers customer questions in a chat window and hands off to a person when it cannot help. This guide covers how it works, the four types, what it can and cannot handle, and when to hand off.

What Is Customer Chat Support? Benefits of Live Chat for Customer Service
•

What Is Customer Chat Support? Benefits of Live Chat for Customer Service

Customer chat support is real-time text help from a person, a chatbot, or both. This guide covers the benefits of live chat for customer service, wait-time benchmarks from 2 billion chats, and how to choose between in-house agents, a BPO provider, and a chatbot.

Customer Service AI Use Cases: 12 Chatbot and AI Examples
•

Customer Service AI Use Cases: 12 Chatbot and AI Examples

12 customer service AI use cases grouped by who they help: customers, agents, and managers. Each comes with what it needs, the metric to watch, and where to start.

Intelligent Customer Engagement: What It Is, Examples, and Metrics
•

Intelligent Customer Engagement: What It Is, Examples, and Metrics

What intelligent customer engagement means, its four parts, examples by channel, the metrics that show it works, and the six ways it goes wrong.

How Much Does a Chatbot Cost? Prices by Business Size
•

How Much Does a Chatbot Cost? Prices by Business Size

Chatbot prices by business size in 2026, from free plans to enterprise quotes, with yearly totals and the cost compared with one customer service rep.

Live Chat vs Chatbot: What Is the Difference?
•

Live Chat vs Chatbot: What Is the Difference?

Live chat puts a person on the other end of the chat window. A chatbot answers on its own. Definitions, a difference table, and examples of each, side by side.

Conversational AI Examples: 11 Real Use Cases and Applications (2026)
•

Conversational AI Examples: 11 Real Use Cases and Applications (2026)

11 real conversational AI examples in five types, the use cases and applications each one covers, the results companies published, and how a small business can apply the same pattern.

Machine Learning in Customer Support: How to Build a Chatbot
•

Machine Learning in Customer Support: How to Build a Chatbot

The four machine learning jobs inside a customer support chatbot, three ways to build one, a worked example with the cost and test math done, and a build or buy decision matrix.

24/7 Customer Support Chatbot: Setup, Handoff, and Cost Math
•

24/7 Customer Support Chatbot: Setup, Handoff, and Cost Math

How a 24/7 customer support chatbot covers nights and weekends: what it should answer, how handoff works when nobody is online, and the staffing math.

ChatGPT for Customer Service: How to Set Up a Chatbot
•

ChatGPT for Customer Service: How to Set Up a Chatbot

Three real ways to use ChatGPT in customer service, with setup steps, monthly costs from 1,000 to 40,000 answers, and a table for choosing one.

What Is An Advantage of Using Online Chat for Customer Support?
•

What Is An Advantage of Using Online Chat for Customer Support?

What is an advantage of online chat for customer support? Instant assistance, seamless communication, and enhanced customer satisfaction await!

Customer Engagement Plan: Template, Steps, and Example
•

Customer Engagement Plan: Template, Steps, and Example

A customer engagement plan lists who you engage, when, on which channel, who owns it, and how it is measured. This guide has 7 steps, a template, a worked example, and metric formulas.

Chatbot vs Conversational AI: What Is the Difference?
•

Chatbot vs Conversational AI: What Is the Difference?

A chatbot is a program people chat with. Conversational AI is the technology that lets software understand free-form language and keep context. This guide compares the two and shows which one to use.