WhatsApp Bot for Ecommerce: How a WhatsApp Chatbot Works

What does a WhatsApp bot need to run?

Six conditions, all set or framed by Meta.

1. A number connected to the WhatsApp Business Platform

Meta describes the platform as a suite of products with business services and APIs that allow businesses to send or receive messages on WhatsApp, either directly or through a Solution Provider (platform terms, as of October 7, 2026). A third-party bot goes through this platform, not through the Tools tab of the WhatsApp Business app.

2. The 24-hour customer service window

When a customer messages or calls, a 24-hour timer starts, and each new message or call from them resets it to 24 hours. While the window is open, the bot can send free-form messages, which Meta calls service messages: text, image, reply buttons, list. When it closes, only pre-approved templates go through (Meta's documentation on service messages, as of October 7, 2026).

3. Message templates to write outside the window

A template is a message prepared in advance, with variables such as the first name or the order number. Meta reviews it automatically when it is created and after each edit, and it can only be sent once approved. Each template is categorized as marketing, utility or authentication (Meta's documentation on templates, as of October 7, 2026). For a bot, templates reopen a conversation: a follow-up after 24 hours of silence, an order update, a cart reminder. Our WhatsApp Business template examples are ready to submit.

4. Customer opt-in

Meta requires businesses to obtain opt-in before messaging people on WhatsApp. The request must clearly state that the person is opting in to receive communication from the business, state the business's name and comply with applicable law. Meta lists several supported methods: SMS, a website, by phone, in person or on paper (Meta's documentation on opt-in, as of October 7, 2026). To apply it in Europe, read our guide to WhatsApp opt-in and GDPR.

5. A way out to a human

The WhatsApp Business Messaging Policy (last updated September 23, 2026, as of October 7, 2026) allows automation when responding during the 24-hour window, on one condition: prompt, clear and direct escalation paths must be available. It lists them: in-chat human agent transfer, phone number, email, web support on the business website, in-store visits, support form. The same policy says not to ask people to share full-length payment card numbers, financial account numbers or personal ID card numbers.

6. A message budget

Meta charges for delivered messages from the business to the customer, never for the customer's messages. Templates are charged by category and by the recipient's country. A bot's free-form replies are service messages: since October 1, 2026, each business phone number has 1,000 free service messages per month, and Meta charges from the 1,001st (WhatsApp Business Platform pricing page, updated September 30, 2026, as of October 7, 2026). The subscription of the tool you choose comes on top.

Does Meta allow AI agents on WhatsApp?

Yes, for a bot that answers a business's own customers: the messaging policy allows automation (point 5 above), and Meta's pricing page states that service messages can be powered by a third-party AI solution or by people. What the platform terms prohibit (section 4.7, as of October 7, 2026) is AI providers using the platform to offer an AI assistant when that assistant is the primary functionality made available, rather than an incidental or ancillary one. The same section adds that such technologies may be made available to businesses in certain countries. A store's bot answers about that store's products, orders and returns.

What does an online store use a WhatsApp bot for?

Meta illustrates its pricing with a sample conversation between a store and a customer. It shows where the bot steps in, and where it stops.

Example taken from the WhatsApp Business Platform pricing page, as of October 7, 2026. In Meta's example, the automated replies come from its own agent, Meta Business Agent. With a third-party agent, they are service messages.

In a store, a bot covers five situations:

• Pre-purchase questions: size, color, availability, delivery time. The bot must be able to read the store's catalog and policies.
• Order tracking: "where is my order?". The bot must be connected to the orders, otherwise it can only point to the tracking page.
• Returns and exchanges: a closed journey (order, item, reason), which suits a rule-based flow.
• Replies to messages the store sent: when a customer answers a cart reminder or an order notification, the 24-hour window opens and the bot can reply freely.
• Handoff to an agent: complaint, refund, payment problem. The bot passes the conversation on with what has already been said.

How do you set up a WhatsApp bot, step by step?

1. List the questions you receive. Go through the conversations of the last few weeks (WhatsApp, email, website chat) and sort them: those with a fixed answer, those that depend on an order, those that call for advice.
2. Choose your access to the WhatsApp Business Platform. Directly, if you have a technical team, or through a tool that handles it. The comparison linked in the introduction details the options.
3. Connect the number and complete the profile. The WhatsApp Business Messaging Policy asks for a business profile with customer support contact information and at least one of the following: email address, website address or telephone number.
4. Collect customer opt-in. Add the request where you collect the phone number: at checkout, on the site, in a form.
5. Write the flows and give the AI agent its sources. One flow per closed journey. For the agent: the catalog, the shipping and return policies, the answers to common questions.
6. Set the handoff to a human. Plan at least three triggers: the customer asks for it, the topic is sensitive (complaint, refund, payment), or the bot does not have the answer. Decide who receives the conversation, and during which hours.
7. Prepare the message templates. The ones used to write after 24 hours: conversation follow-up, order update. Submit them to Meta before launch.
8. Test with real questions. Reuse real wording, typos included. Introduce the bot as an automated assistant from its first message: Article 50 of the EU AI regulation provides that people are informed that they are interacting with an AI system, unless this is obvious from the context (text of the regulation on EUR-Lex, read on October 7, 2026).
9. Open in stages. Start with one type of question, reread the conversations, then widen the scope.

How do you measure the results of a WhatsApp bot?

There is no "good" rate that holds for every store. Your point of comparison is your own numbers from the weeks before the launch, then how they move week after week.

The statuses, the quality rating and the 1,000-message free tier come from Meta's documentation (messages webhook, service messages, pricing), as of October 7, 2026. The other rows are calculations to run on your own data.

Every week, reread the conversations the bot did not resolve. A question left without an answer is content to add, a misunderstood question is a flow to fix, and a customer who gets annoyed is a handoff rule to trigger earlier.

Where does Kanal fit in?

Kanal is a WhatsApp marketing and sales app for Shopify stores that runs on the official WhatsApp Business Platform. Its AI chatbot learns your product catalog, your policies and your brand voice. It answers product questions on WhatsApp, handles order tracking, guides returns and transfers complex queries to your team with the full conversation history. Its automations are built in a visual flow builder (triggers, conditions, actions) tied to Shopify events, and every conversation lands in a shared team inbox.

On Meta's rules, Kanal collects opt-ins at checkout and on your site, uses approved message templates and handles opt-outs. To see it answer from your own catalog, book a demo.