Chatbot Customer Service Rulebook: Who Answers What, and When the Bot Steps Aside
Chatbot customer service works when everyone, the bot included, knows its role. This rulebook gives you a policy you can adapt in an hour: which questions the bot answers, which it passes to people, how fast your team replies, and the promises you make to customers about all of it.
โก In short
- Sort every common question into green (bot answers), amber (bot collects, person finishes) or red (person only).
- Write down your reply-time promises for chat, email and tickets, and show them to customers.
- Customers must always be able to reach a person, and the bot must say when one will reply.
- Review the rulebook monthly using real conversations, not guesses.

Why a rulebook
Most bot problems are policy problems
When a support bot frustrates customers, the cause is rarely the technology. It is usually that nobody decided what the bot should do in the situation that went wrong. Should it try to calm an angry customer or pass them on straight away? Can it promise a refund? Should it answer a question about a competitor? Without a written answer, the bot improvises, and so does your team.
A chatbot customer service rulebook fixes this. It is a short document, one or two pages, that the whole team agrees on. It sets out three things: which conversations the bot owns, which it shares with people, and which it never touches. It also records the promises you make to customers about reply times. Once written, it guides how you set up the bot, how your team works the inbox, and what you tell customers.
The rest of this page walks through each section of the rulebook with examples. At the end you’ll have everything you need to write your own. If you are still deciding whether you need a bot at all, start with chatbot vs live chat and come back.
Part 1 ยท The traffic lights
Sort every question into green, amber or red
Click each colour to see what belongs there and how the bot behaves.
The bot answers fully. Questions with a stable, factual answer that is the same for everyone.
- Opening hours, address, directions, parking
- Delivery areas, delivery times, shipping charges
- Return window and how to start a return
- Payment methods, cash on delivery, EMI options
- How to book, reschedule or cancel
- Product facts: sizes, materials, compatibility
Bot behaviour: answer in two or three sentences with the specific detail, then ask “Anything else I can help with?” Offer a person if the visitor asks or seems unsatisfied.
The bot starts, a person finishes. Questions that need a human decision or access to a specific record.
- “Where is my order?” when it needs a lookup
- Custom quotes and bulk orders
- Damaged or wrong items received
- Changes to an existing booking or subscription
- Account or billing problems
Bot behaviour: acknowledge the issue, collect the details a person will need (order number, email, photo description, dates), tell the customer who will reply and when, then hand over or create a ticket.
Straight to a person. Conversations where a bot reply could cause harm, anger or legal trouble.
- Complaints, especially repeat ones
- Refund disputes and chargebacks
- Health, safety, legal or financial advice
- Threats, abuse, or anything mentioning legal action
- Requests to delete personal data
- Anything involving a child’s safety
Bot behaviour: one short, calm line (“I’m passing this to our team now”), collect contact details if offline, and stop. No troubleshooting, no defending the company.
Part 2 ยท Reply-time promises
Write down how fast you’ll answer, then keep it
Customers forgive a slow reply if they were told what to expect. They do not forgive silence. Your rulebook should state a reply target for each channel, and your chat should tell customers that target honestly.
Suggested targets for a small team
– Live chat, during opening hours: first human reply within 2 minutes for amber and red chats. – Chat messages left while offline: reply by the next working morning. – Tickets: first reply within one working day; a fix or a clear update within three. – Email: within one working day.
These are starting points, not industry rules. Pick targets you can hit on a busy day, not a quiet one. Missing a promise does more damage than making a modest one. The article on live chat response time covers how to measure and improve your numbers.
Tell customers the truth
Set your opening hours so the chat shows when a person is available. Outside those hours, the away message should say exactly when you’ll reply and collect an email or phone number. “We’re back at 9 am tomorrow and will reply then” is a promise customers can plan around.
Part 3 ยท Handover rules
Exactly when the bot steps aside
The handover is the moment chatbot customer service either earns trust or loses it. Write these rules into your policy and set them up in your bot.
Always hand over when
1. The customer asks for a person, in any wording (“human”, “agent”, “real person”, “call me”). 2. The question is on your red list. 3. The bot has failed to answer the same question twice in one chat. 4. The customer uses angry language or capital letters, or mentions a previous unresolved contact. 5. The conversation involves money already paid and something going wrong.
How the handover should sound
The bot should acknowledge the issue, say what happens next and ask only for the details the person will need. It should never make the customer repeat their story. Compare:
Part 4 ยท Team roles
Who does what on the human side
Even a two-person chatbot customer service team needs clear roles, or chats fall between chairs.
Inbox owner (per shift)
Watches the chat during set hours, picks up handovers, replies within your target. On a small team this rotates daily.
Ticket owner
Checks the ticket queue every morning, assigns or solves each ticket, and chases anything older than three days.
Rulebook owner
Reads a sample of bot conversations weekly, updates answers, and proposes changes to the traffic lights each month.
Escalation contact
The person (often the owner) who handles red-list complaints, refund disputes and anything legal.
Tools that help: saved replies for common human answers, turning chats into tickets for anything that needs follow-up, and the Android app for whoever is on shift away from a desk.
Part 5 ยท Tone
One voice for bot and team
Customers notice when the bot sounds cheerful and robotic and the team sounds tired and formal, or the other way round. Your rulebook should describe one voice for both.
Write three words and three rules
Three words for how you sound, for example “warm, direct, practical”. Then three rules that make the words concrete:
- Use the customer’s name once, early, if you have it.
- Lead with the answer; explain after.
- Apologise once for a real problem, then focus on fixing it.
Things neither bot nor team should say
“Your call is important to us.” “As per our policy.” “Kindly do the needful.” “I understand your frustration” (said without doing anything about it). Replace them with plain, specific language. The guides on positive scripting and how to sound human in live chat have many examples you can borrow.
Give the bot a name, honestly
It’s fine to give the bot a friendly name, but don’t pretend it’s a person. “I’m the shop assistant bot; I can answer most questions or get you a person” sets expectations correctly and makes the handover feel natural.
Part 6 ยท The review
A monthly rulebook check in 30 minutes
Tick each step as you go. Run this on the first working day of each month.
Writing yours
A fill-in-the-blanks rulebook template
Copy the outline below into a document, fill in the blanks for your business, and share it with everyone who touches customer service. Keep it to two pages; a rulebook nobody reads helps nobody.
- Purpose: “Our bot answers simple questions instantly, any time, so our team can spend time on problems that need a person.”
- Green list: the questions the bot answers fully (list 10โ20).
- Amber list: the questions where the bot collects details and a person finishes, with the details needed for each.
- Red list: the topics that go straight to a person.
- Handover triggers: the five rules above, adapted to your business.
- Reply-time promises: your targets for chat, offline messages, tickets and email.
- Roles: inbox owner, ticket owner, rulebook owner, escalation contact, with names.
- Voice: three words, three rules, and a short list of banned phrases.
- Review: who runs the monthly check and when.
Once written, set up the bot to match: green answers in its knowledge, amber and red handover rules, opening hours and away message, and alerts for the inbox owner. On LiveChatBot all of this runs on the free plan; the AI Agent handles the green list, and tickets, saved replies and the Android app cover the human side.
Frequently asked questions
Can chatbot customer service replace my support team?
For a small business, no, and it shouldn’t try. The bot takes over repeat factual questions so your team has time for problems that need judgement, empathy or access to records.
What percentage of questions should the bot handle?
There’s no universal figure. Track how many chats end without a handover in your first month, and grow the green list steadily from there.
Should the bot apologise?
Once, briefly, for a real problem, followed by a concrete next step. Repeated apologies without action annoy customers more than none at all.
How do I stop the bot promising things we can’t deliver?
Keep refunds, discounts and exceptions on the amber or red list, and only give the bot facts and policies it can quote, never decisions.
Where do I start if I have no rulebook today?
Pull your last 50 customer messages, sort them into green, amber and red, and set up the bot for the green ones first.
Keep reading
Related guides
Chatbot Button Design: Placement, Labels and Custom ‘Ask Us’ Buttons That Get Clicked
Chatbot Artificial Intelligence Examples: 8 Annotated Conversations That Show What the AI Is Actually Doing
Chatbot Application Ideas by Department: 14 Real Jobs a Website Bot Can Take Over
More to read
More guides on this topic
Chatbot Platform, Inside Out: Trace One Customer Message Through Page Rules, Bot Answers, Handover and Tickets (Free Message Path Tracer)
Best Chatbot for Service Businesses: Plan Who Answers When, and Turn After-Hours Chats Into Booked Jobs (Coverage Planner)
Put your rulebook to work
Set up the green list, handover rules and opening hours on livechatbot.in. Free plan, no card needed.