Guide

How to create a WhatsApp chatbot

To create a WhatsApp chatbot, define one customer task and prepare reliable answers. For automatic menu steps, connect your business number to chatbot software on the WhatsApp Business API through Meta's Cloud API, directly or through a provider. Build the flow and test human handover. The free app offers greeting and away messages; its quick replies are selected and sent by a person.

How to create a WhatsApp chatbot

This guide gives you a practical sequence from preparation to everyday use. If you want to see the product capabilities while planning, the WhatsApp chatbot page covers menus, AI replies, keyword answers and human takeover in Kwamle Media. The work below is about deciding what those tools should say and do for your own business.

What you can start for free

The WhatsApp Business app's greeting message, away message and quick replies are free. They help welcome people, explain staff availability and reuse familiar answers. They do not turn the app into a chatbot that understands requests, follows a menu through an order and assigns exceptions to staff. In particular, a quick reply still needs a person to choose and send it.

TaskFree WhatsApp Business app toolsChatbot on WhatsApp Business API
Welcome and availabilityGreeting and away messagesWelcome and next step configured in software
Common questionsA person selects and sends a saved quick replySoftware responds using configured rules or AI instructions
Menus and ordersStaff read the customer's choice and continue manuallySoftware follows configured steps and hands exceptions to staff

A chatbot on the API needs software. Meta may charge for some messages according to their type and the customer's country; check Meta's current pricing and the software's terms before choosing. Kwamle Media has a 7-day free trial, which lets you try its software. A trial is different from a promise that API messaging will always be free.

Start with the app if it solves the problem you wrote down. If it cannot, use the app and API comparison to check whether the additional setup matches your daily work.

How to create a WhatsApp chatbot, step by step

1. List the questions you answer every day

Read recent conversations and copy the customer's actual wording: “Are you open?”, “Can I collect it?” or “How do I change my order?”. Group questions that need the same answer. Beside each group, write the information staff use and who can confirm it when it changes.

Separate stable facts from things that need checking. Your collection location may be settled, while today's product availability is not. The chatbot should only give a definite answer when it has reliable information to use. Keep this working note somewhere your team can update; it becomes the basis for saved replies, menu choices and AI instructions.

2. Decide what stays human

Mark the conversations a person must handle before writing automated replies. Complaints, payment disputes, unusual orders and requests for a person need a clear route to staff. At a clinic, keep the chatbot to practical information and leave medical questions to the appropriate team.

For each boundary, write what the assistant should say and who takes over. “A colleague will check whether we can make that change” is useful only if a colleague is responsible for reading it. Keep these boundaries visible during setup, so a helpful-sounding answer does not quietly become an unauthorised promise.

3. Start with the app's free tools and understand their limits

Set a greeting that names the business and explains how customers can ask for help. Write an away message that reflects when staff can reply. Prepare quick replies for stable information such as collection arrangements or the details needed to investigate an order.

Use those replies in normal conversations and notice where people still get stuck. An away message acknowledges a request; it does not answer a follow-up question. A quick reply saves typing; it does not decide which answer fits. If your team can finish the remaining work comfortably, you may already have the setup you need. Keep the useful wording even if you later move to a chatbot.

4. Choose the API when the conversation needs it

Consider the WhatsApp Business API when several people need to coordinate replies, or when you need menus, AI answers and guided order steps through software. The number connects through Meta's Cloud API, directly or through a provider. The software supplies the inbox and conversation tools; connecting a number alone does not build the bot.

Choose a setup that supports your planned workflow and human takeover. Check the current official instructions for your business account and number before connecting, especially if the number already works in the app. Do not assume that app access or old conversations will carry across unchanged. Prepare the conversation before changing the way customers reach you.

5. Draw the menu and the steps

Sketch a short welcome menu using customer language: place an order, opening hours, talk to the team. Under each choice, write the next question, the answer it needs and how the customer finishes. Let someone who changes their mind correct a detail without being pushed towards an order they no longer want.

Ask only for information needed for that request. A person asking when you open should not have to give a delivery address. An order path should end with a summary and an explicit confirmation, while the human path should reach the inbox. The guide to building a flow without code develops this planning exercise further.

A planning sketch for this example:

  • Place an order: product and quantity → fulfilment details → checked summary → customer confirmation → staff check availability.
  • Hours and location: checked business information → return to menu.
  • Talk to the team: reason for contacting staff → handover with conversation history.

Keep a route back to the menu and allow corrections before confirmation. Send unrecognised requests to staff. In the free app, a person reads the choice and replies; automatic branching needs configured chatbot software.

6. Write the AI instructions and keyword triggers

Describe the business, its tone, the facts the assistant can use and the questions it must pass on. Add examples of a useful answer and an honest uncertainty. For a bakery, an instruction could say: “Explain our collection arrangements using the supplied information. Ask the team to check availability and custom decoration. Never promise that an order has been accepted before staff confirm it.”

Use keyword triggers for fixed information, such as “hours” or “location”, and write those answers exactly as customers should see them. Test words in context: “change my collection location” needs different handling from “where is your location?”. In Kwamle Media, flows and keyword answers sit alongside AI replies; choose the method that fits the request. Fluent wording should never substitute for missing facts.

7. Confirm orders inside the chat

Before asking for confirmation, show the relevant product, options, quantity, collection or delivery details, and the checked amount payable. Invite corrections. A customer asking about a product has not yet placed a confirmed order, and silence is not confirmation.

For orders from a Kwamle Media store, the customer can confirm in the conversation before staff prepare the order. The order confirmation page explains that step. Keep customer confirmation separate from checks the business still owes, such as whether it can meet a custom request. If the customer changes a detail, make sure the final summary reflects it before anyone starts packing.

8. Plan the handover to a person

Give the customer an obvious way to ask for staff throughout the conversation. Name the colleague or role responsible for taking over, and write an honest message about staff availability. Do not tell customers that somebody is checking if nobody owns the inbox.

In Kwamle Media, a teammate can take over in the unified inbox, read the conversation history and continue while the bot stops answering that chat. Practise that change with your team. The human handover guide helps identify requests that deserve personal attention. The first human reply should address the unresolved question instead of asking the customer to start again.

9. Test with a colleague before going live

Ask a colleague who did not write the flow to act as a customer. Let them type naturally, skip the menu, correct an earlier detail and ask for something you do not offer. Watch both sides of the conversation: a smooth customer screen can hide an unclear task for staff.

Check the menu paths, the fixed answers and the AI responses against your approved information. Ask for a person during an order and confirm that the bot goes quiet when the colleague takes over. Try a request outside opening hours and verify that it sets an honest expectation. Fix the failed path, then test it again before opening the chatbot to customers.

10. Follow WhatsApp's messaging rules

Get clear customer permission, or opt-in, for the updates you plan to send. Explain their purpose, record the customer's choice and respect requests to stop. An enquiry about collection is not permission for unrelated offers, and an approved template does not replace that permission.

A customer's message opens a 24-hour API reply window. Outside it, business messages require approved templates under WhatsApp's current rules. Check the current official documentation for your intended messages, and prepare suitable templates before planning follow-ups. Do not assume a saved reply is an approved template or that sending your own message restarts the window. Apply these checks to staff messages as well as automation.

11. Measure the conversations and improve

Review whether customers reach the answer they need, where they repeat themselves and which requests end up with a person. For orders, check whether details are clear, corrections survive into the summary and unconfirmed requests stay out of preparation. Look at actual conversations before deciding what to change.

Keep a simple record of resolved requests, pending questions and human handovers. Kwamle Media provides analytics in a shared dashboard. Use it alongside the conversation history and your team's notes to understand the work. Choose a recurring problem, update the relevant wording or branch, and test that path again. Keep human handovers available while improving the bot; a request reaching a person can be exactly the right outcome.

Worked example: a bakery's collection order

Imagine a bakery preparing a chatbot for collection enquiries. The team knows the collection arrangements, but staff must check availability and custom decorations. Its short menu is:

  • Order for collection: choose a product and describe the collection request.
  • Hours and location: receive the bakery's checked practical information.
  • Talk to the bakery: ask a person about a change, a problem or a special request.

The conversation below illustrates confirmation of the requested details. In a real order, show the complete order summary, including the checked amount payable, before asking the customer to confirm it.

Assistant: Hello, this is the bakery's automated assistant. Would you like to order for collection, check our hours and location, or talk to the bakery?
Customer: I'd like a chocolate cake for collection on Saturday.
Assistant: A colleague will check availability. Would you like a message on the cake?
Customer: Please write “Happy birthday Maya”.
Assistant: Your request is a chocolate cake for collection on Saturday, with “Happy birthday Maya”. Are those details correct? The bakery still needs to confirm availability.
Customer: Yes, those details are correct.
Baker: Hi, I can see your request and the cake message. I'll check availability and agree the collection arrangements with you here.

The assistant gathers useful information without accepting a promise it cannot check. The baker continues from the details already supplied. If the customer asks about a different collection day, the team should update the request and check it again rather than rely on the earlier confirmation.

Put the first version into daily use

Start with a familiar conversation whose information is ready. Tell the team who checks the inbox, who updates business facts and who handles an unanswered question. Keep the menu and fixed answers in language your customers recognise. Add more paths when actual conversations show a useful need, rather than filling the menu with everything the business does.

In Kwamle Media, use the no-code flow builder for the steps, AI instructions for open questions and the shared inbox for exceptions. Review the wording when collection arrangements, services or products change. A working chatbot is a maintained part of customer service: the welcome, the order check and the person taking over should all describe the same business practice.

The rules in numbers (checked November 2026)

  • A customer's message opens a 24-hour customer service window: replies inside it are free-form; outside it, only an approved template message can be sent. Meta
  • Templates have three categories: marketing, utility and authentication. Meta
  • For messaging charges and the monthly free allowance, see WhatsApp Business API.
  • Business messaging must follow WhatsApp's business policy. WhatsApp

Common questions

Can I create a WhatsApp chatbot for free?

You can use the app's greeting messages, away messages and quick replies for free. That is not a complete chatbot: staff select quick replies and handle menu choices. An API chatbot needs separate software. Check its terms and Meta's messaging charges; a software trial does not make API messages free.

Does the WhatsApp Business app include a full chatbot?

Its built-in tools help with greetings and saved answers. They do not by themselves provide the menu, AI and guided order workflow described here. Use them first if staff can comfortably handle the remaining conversation.

Do I need to write code?

Kwamle Media supports a no-code flow builder and AI instructions written in ordinary language. You still need to prepare reliable information, define the steps and test the result. Connecting the account is only part of the work.

Can the chatbot accept orders without staff checking everything?

It can guide customers and support confirmation of orders from your Kwamle Media store inside the chat. Decide which standard orders the business can fulfil and which requests require a person. The bot should not invent availability, payment status or acceptance of a special request.

What happens when a customer asks for a person?

Provide a clear human route and assign someone to watch it. In Kwamle Media, the teammate can read the history and take over while the bot stops replying in that conversation. Tell the customer honestly when staff are available.

How do I create a WhatsApp chatbot for my business?

Choose a task such as order enquiries, write checked answers and mark requests for staff. Connect your business number through Meta's Cloud API, directly or through a provider, and configure the menu and replies in chatbot software. Test corrections, unknown questions and human takeover before using it with customers.

How do I activate a chatbot in WhatsApp Business?

There is no switch in the free app that turns greetings or quick replies into the chatbot described here. For that, connect the number through Meta's Cloud API, directly or through a provider, then configure and test the flow in chatbot software. If you only need greeting and away messages, set up those app tools instead.

Solutions

WhatsApp Business · Official WhatsApp connection · Broadcast campaigns · Customer follow-up · Restaurants and cafés · Cafes and bakeries · Clinics and practices · Dental clinics and practices · Pharmacies · Hotels and vacation rentals · Pet shops and veterinary clinics · Car service centres · Travel agencies and tour organisers

Free tools: WhatsApp link generator

Updated

ShareXLinkedInWhatsAppTelegram

Try it on your own channels

Start your free trial and set up your first automation in minutes.