Your customers already message you on WhatsApp. A WhatsApp AI chatbot answers them in seconds, at any hour, from your own information. Here is how it is set up on the official platform and what it is good for.
In many markets, WhatsApp is where customers already talk to businesses. They message a clinic to move an appointment, a shop to ask about a size, a builder to send a photo of the problem. Most of those messages wait until someone picks up the phone. A WhatsApp AI chatbot answers them in seconds, from your own information, and passes the tricky ones to your team with the conversation attached.
What is a WhatsApp AI chatbot for business?
A WhatsApp AI chatbot for business is software connected to your business number through Meta's WhatsApp Business Platform that reads incoming messages and replies automatically, using a language model and your own content. It answers questions, shares order updates and booking links, and hands the chat to a person when it should. It needs the official platform, not the free WhatsApp Business app on a phone.
The AI part is what separates it from older WhatsApp bots that offered a numbered menu. Customers type the way they normally would, in their own words and often their own language, and the bot works out what they want. Built properly, it answers from your documents and systems rather than from the model's memory, which is what keeps it from inventing prices or policies.
WhatsApp Business app vs WhatsApp Business Platform
These are two different products, and the difference decides whether a real chatbot is possible at all.
| Aspect | WhatsApp Business app | WhatsApp Business Platform |
|---|---|---|
| What it is | An app for small business owners to chat with customers by hand | APIs for sending and receiving messages from your own software |
| Automation | Greeting and away messages | Full control: AI replies, lookups, bookings, handover rules |
| Who replies | You and your team, by hand | Your bot and your team, through a shared inbox or your own app |
| Messages you start | One to one, typed by hand | Approved templates, sent at scale within messaging limits |
| Meta's fees | Messages sent from the app are free | Charged per delivered message, by category and country |
You do not always have to give up the app. Meta now lets a business connect an existing WhatsApp Business app number to the platform through a partner, keep using the app for one-to-one chats, and have recent message history synced between the two. Messages sent from the app stay free; messages sent through the API are charged.
WhatsApp chatbot use cases that pay for themselves
The best WhatsApp chatbot use cases share a pattern: lots of conversations, most of them similar, where a fast answer is worth more than a perfect one.
| Use case | What the bot does | Where a person steps in |
|---|---|---|
| FAQs and pre-sales questions | Answers about products, prices, delivery areas and policies from your content | Unusual requests, negotiations |
| Order status | Looks up the order in your system and replies with the latest update | Lost parcels, refunds |
| Appointments | Shares open slots, books, sends reminders and handles reschedules | Clinical or complex questions |
| Leads from click-to-WhatsApp ads | Greets, asks qualifying questions, books a call or sends a quote form | Taking the qualified call |
| After-sales and support | Collects photos, order numbers and symptoms before routing | Diagnosis and fixes |
| Updates and alerts | Sends confirmations, delivery notices and payment reminders as templates | Exceptions and disputes |
Clinics and dental practices, salons, property agents and trades get the most out of booking and reminders. Shops get the most out of order status and product questions. Any business running click-to-WhatsApp ads should have something answering those chats instantly, because the person who clicked is interested now, not tomorrow.
The platform supports more than plain text, which helps a bot stay short and clear. Reply buttons offer up to three choices, list messages show a menu of options, and WhatsApp Flows present structured forms for things like capturing a lead or collecting booking details, all inside the chat.
The WhatsApp rules every chatbot has to follow
WhatsApp is stricter than a website chat window, and the rules shape how the bot is designed.
- The 24-hour customer service window. When a customer messages or calls you, a 24-hour window opens, and it resets with each new message. Inside it, your bot can reply with normal messages. Outside it, you can only send pre-approved template messages.
- Templates for anything you start. Reminders, order updates and offers sent outside the window must use templates, each categorised by Meta as marketing, utility or authentication.
- Opt-in first. You may only message people who have agreed to hear from you on WhatsApp.
- Messaging limits. A new business portfolio can reach 250 unique customers outside the service window in a moving 24-hour period. Verifying your business is one route to 2,000, and from there the limit rises automatically while your messages stay high quality.
- No general-purpose AI assistants. Meta's platform terms prohibit using WhatsApp to offer an AI assistant as the main product. A bot that helps customers with your business, where the AI is incidental to your service, is the normal case.
That last rule is set out in section 4.7 of the Meta Terms for WhatsApp Business Platform, last modified in September 2026. A support or booking bot for your own customers sits comfortably inside it.
How to set up a WhatsApp chatbot, step by step
- Decide the job. List the twenty questions or tasks the bot will handle, the ones it must hand over, and the ones it must refuse. This takes a day and saves weeks.
- Prepare your business on Meta. You need a business portfolio in Meta Business Suite. Verify the business early, because verification is also one route to a higher messaging limit.
- Choose the number. Use a new number for the platform, or connect your existing WhatsApp Business app number through a partner if you want to keep using the app as well.
- Choose how you connect. Use Meta's Cloud API directly, or go through a partner. Solution Partners can bill you for messaging; with Tech Providers you add your own payment method and Meta bills you directly.
- Build the bot. Incoming messages arrive at your server as webhooks. The bot searches your content, checks your systems where needed, asks the model for a reply within your rules, and sends it back through the API.
- Write and submit templates for every message you will start: confirmations, reminders, follow-ups. Category matters for pricing, so describe each one honestly.
- Set up handover. Decide which topics, words or customer requests send the chat to a person, and where your team will pick it up.
- Test with real questions, then launch in stages: a handful of customers first, then everyone.
The middle step is where the quality comes from. A bot that searches your own help pages and policies before replying, an approach called retrieval-augmented generation (RAG), will hold up in front of customers. A bot that answers from the model's general knowledge will not. Our WhatsApp chatbot development work starts with that retrieval layer before any conversation design.
WhatsApp Business API pricing: what Meta charges
Meta's fees sit on top of whatever you pay to build and run the bot. As checked on Meta's WhatsApp Business Platform pricing page in October 2026, the main rules are these.
- Meta charges per delivered message, and only for messages from your business to customers. Messages customers send you are free.
- The rate depends on the message category (marketing, utility, authentication, service) and the country code of the recipient's number.
- Service messages, the normal replies your bot sends inside the 24-hour window, are free for the first 1,000 each month per business phone number. From 1 October 2026, Meta charges for service messages beyond that.
- When a customer starts the chat from a click-to-WhatsApp ad on their phone and you reply, Meta opens a free entry point window of up to 7 days, during which your messages are not charged.
- Utility and authentication messages get lower rates at higher monthly volumes, and Meta may change rates only on the first day of a quarter.
For a small business, the practical takeaway is that customer-started support conversations are cheap, and marketing templates are where the bill grows. Our guide to what an AI chatbot costs covers the build, model usage and hosting that sit alongside these fees. To price your own volumes, the WhatsApp API pricing calculator uses Meta's official rates for every country.
Meta Business Agent or your own AI chatbot?
Meta now offers its own AI agent for WhatsApp, the Meta Business Agent, announced in June 2026. Meta says it answers questions about your business, recommends products from your catalogue, books appointments, qualifies leads and lets you decide when a team member steps in, and its Agent Platform connects it to systems such as Shopify and Zendesk. On the WhatsApp Business Platform, Meta bills its replies by tokens instead of as service messages: one global rate of $2.00 per million tokens on its pricing page in October 2026, which Meta estimates at roughly 4 to 5 cents a message.
It is a reasonable option for a business that only needs WhatsApp and is happy inside Meta's tools. A custom bot makes more sense when the same assistant must work on your website, email or phone line as well, when you want to choose the model and see exactly how retrieval works, when your content lives in systems that need careful syncing, or when you want the code and data in your own accounts. We compare the two for your case as a normal part of scoping AI chatbot development, before you commit to either.
What a good WhatsApp chatbot conversation looks like
A customer messages a salon at ten at night: 'Can I get a balayage this Saturday, and how much is it?' The window is now open. The bot finds the price in the salon's own price list and quotes it as written, then sends a list message with the Saturday slots still free. The customer taps one. The bot books it in the salon's calendar and replies with a confirmation and the cancellation policy, quoted from the policy page.
Two days later the window has closed, so the reminder goes out as an approved utility template, with a button to confirm or reschedule. On the morning itself the customer replies 'running ten minutes late'. That message reopens the window, the bot acknowledges it in one line and alerts the stylist. No member of staff typed anything, and every message stayed inside the rules.
Mistakes that get WhatsApp chatbots switched off
- Unofficial tools that automate a personal or app number. They are cheap until the number is banned.
- Marketing templates sent to people who never asked for them. Meta only raises your messaging limit while your messages stay high quality, so blocks and reports keep you small.
- No way out. Customers who type 'agent' or 'human' and get another menu stop using the channel.
- Long replies. On a phone screen, three short sentences beat a paragraph. Set a length limit.
- Ignoring the window. A bot that tries to send a normal follow-up the next day will fail; plan templates for every message the business starts.
Want a WhatsApp AI chatbot built on the official platform?
Tell us what your customers message you about and where the answers live. We will set out the setup, the templates you need and Meta's likely fees for your volume, then quote the build of your custom WhatsApp chatbot as a fixed price, with every account in your name. You can start a project here, and we reply within one business day.
FAQ
Frequently asked questions
Not directly through the consumer ChatGPT app. You can, however, build a WhatsApp chatbot on Meta's WhatsApp Business Platform that calls an OpenAI model, or a model from another provider, through its API. The bot receives messages from WhatsApp, searches your own content for the answer, asks the model to write the reply within your rules, and sends it back. The model is one component; the useful part is the retrieval, the rules and the handover around it.
No. Businesses can connect to Meta's Cloud API directly and build their own bot. Partners are useful if you want help with onboarding, a ready-made inbox for your team, or a single invoice. Solution Partners can bill you for Meta's message fees, while with Tech Providers you add your own payment method and Meta bills you. If you use a partner, ask for its platform fee in writing, separate from Meta's fees.
It can send messages to many customers at once using approved template messages, but only to people who have opted in to hear from you on WhatsApp. Marketing templates are charged per delivered message and are subject to messaging limits that grow with good-quality sending. Customers who block or report your messages hurt that quality. Treat broadcasts as useful updates people asked for, such as reminders or order news, rather than a free mailing list.
Yes, if it is built to. The platform delivers images, documents, audio and video to your server. A bot can transcribe voice notes with a speech-to-text model before answering, and can read images with a vision-capable model, such as a photo of a damaged product or a scanned form. It is worth deciding up front which media the bot handles itself and which go straight to a person, especially for anything sensitive.
They should. Telling customers they are chatting with an automated assistant, and showing how to reach a person, builds more trust than pretending otherwise. Customers are generally fine with a bot that answers quickly and correctly. What frustrates them is a bot that cannot understand them and will not let them escape. A clear greeting, short answers and an obvious route to a human are the three things that matter most.
A focused WhatsApp AI chatbot over content you already have usually takes a few weeks to build and test. Business verification on Meta and template approvals run alongside the build, so start them early. Bots that look up orders, book appointments or connect to several systems take longer, because every integration needs testing. Launching to a small group of customers first lets you fix gaps before everyone sees it.
It should say so and hand the conversation to a person, with the full chat and a short summary attached. Your team picks it up in a shared inbox or in your own help desk, replies within the customer service window, and the bot steps back. The unanswered questions are worth reviewing weekly, because each one points to a missing document or a rule worth adding, which is how the bot improves after launch.
Written by
Kavion Solutions Team
The senior team behind Kavion Solutions. We build websites, apps, AI tools and cloud setups, and run the SEO and ads that bring in customers, so everything here comes from work we have actually shipped.
Want this done for you?
Start a project