WhatsApp Business API chatbot: what it does, what it costs and how to launch it

October 29, 2026 · 8 min read

A WhatsApp Business API chatbot is a program connected to your WhatsApp number through Meta's official API that answers customers automatically, looks up real data such as stock, orders or available slots, and hands the conversation to a person when it should. You pay for development once, a monthly running cost, and Meta's per-message fees, which depend on the country and type of message. This guide covers what it can do, how the rules work, what drives the cost and how to launch without annoying your customers.

WhatsApp Business app vs WhatsApp Business API

Meta offers two products with similar names, and they solve different problems.

If you get a few dozen messages a day and one or two people handle them, the app is fine. Once several people share the same number, the same questions keep coming in, or you want WhatsApp connected to your systems, you need the API.

You can access the API directly through Meta's Cloud API or through a Business Solution Provider (BSP) that adds its own tools and pricing on top. Either way you need a Meta Business account, a verified business and a phone number for the API. Meta's rules about using the same number on the app and the API have changed over time, so check the current options before you migrate a number your customers already know.

What an AI WhatsApp chatbot for business can do

Old WhatsApp bots were button menus: press 1 for sales, 2 for support. Customers hated them. An AI chatbot understands free text ("do you have the blue one in medium?", "where's my order, I bought it Tuesday") and, more importantly, can look up the answer in your systems instead of reciting an FAQ.

Connected to your data, typical tasks include:

• Inventory and prices. Check real stock and prices by product, variant and location, and send a link to buy.
• Order status. Find the order by phone number or order ID and report its status and tracking, straight from your store or shipping provider.
• Bookings and appointments. Show available slots from your calendar, book, confirm, reschedule and send reminders.
• Quotes and lead qualification. Ask the right questions, calculate a price or range, and pass a qualified lead to sales with the full context.
• Payments. Send a payment link, confirm when it's paid and update the order.
• Post-sale. Returns, warranty claims and how-to questions, with photos collected and a ticket created for your team.
• Internal use. Staff asking for stock, prices or a sales report from their phone.

The key design rule: the AI should only state facts it got from your systems or your approved content. When the answer isn't there, it says so and offers a person. That's what keeps it from inventing prices or policies.

Human handoff

A chatbot that can't hand off is a wall between you and your customers. Plan the handoff before anything else:

• When to hand off. The customer asks for a person, the bot can't find the answer, the customer is upset, the sale is large, or the topic is sensitive (complaints, refunds, health, legal).
• Where it goes. A shared inbox where your team sees the whole conversation, not a new chat where the customer has to repeat everything.
• What the bot says. Be honest about timing: "A person will reply within business hours, usually in under an hour." Don't promise "right away" at 11 p.m.
• Getting the bot out of the way. Once a person takes over, the bot stays quiet in that conversation until the agent closes it.

Templates and the 24-hour window

These two rules shape everything you can do on the API, so it's worth understanding them early.

The customer service window

When a customer messages you, a 24-hour window opens. Inside it, your bot or your team can reply freely with any content. Every new message from the customer restarts the 24 hours.

Template messages

Outside that window, you can only start a conversation with a template message that Meta has approved in advance. Templates are grouped into categories, mainly marketing (offers, announcements), utility (order updates, appointment reminders, account alerts) and authentication (one-time codes). The category affects both the price and how strictly Meta reviews it, and a template submitted as utility that reads like marketing may be recategorized or rejected.

You also need the customer's opt-in to message them first, and customers can block or report you. A high block rate can lower your number's quality rating and limit how many messages you can send, so treat proactive messages as something customers should actually want.

How much a WhatsApp chatbot costs

The total WhatsApp chatbot cost has four parts:

1. Development. Conversation design, integrations with your systems, the handoff inbox, testing and launch. This is the biggest variable: a bot that answers from your FAQ and catalog is much simpler than one that reads stock from your ERP, books appointments and takes payments. We quote each project at a fixed price after a free discovery call, based on the integrations and scope.
2. Meta's fees. Meta charges per message or conversation depending on the category of message and the customer's country, and its pricing model has changed more than once in recent years. Rates for marketing messages are typically higher than for utility or authentication messages, and replies inside the customer service window have been treated differently from templates. Don't plan around numbers from an old blog post; check Meta's current rate card for the countries you serve.
3. AI usage. Each answer the AI writes has a small usage cost that grows with conversation volume and length. For most businesses it's modest compared with staff time, but it should be in the budget.
4. Hosting and maintenance. Servers, monitoring, updates when Meta or your systems change, and improving answers based on real conversations. Maintenance is a monthly plan, quoted according to the size of the system.

If you use a BSP or an inbox platform, add its subscription. Some charge per agent seat, some add a markup on Meta's fees.

To decide if it's worth it, count how many conversations your team handles per month, what share are repetitive (stock, prices, order status, hours, bookings) and how long each takes. If a third of those can be resolved automatically, the saving in staff time is usually easy to compare with the costs above. Faster answers at night and on weekends also bring sales you currently lose, which is harder to measure but often the bigger gain.

Common mistakes

• Menus instead of conversation. Forcing customers through numbered options when they've already typed their question.
• A bot with no data. An AI that only knows a generic FAQ will answer "please contact us" to anything useful. Connect it to stock, orders or calendars, or keep it narrow.
• No way out. No clear path to a human, or a handoff that loses the conversation history.
• Letting the AI improvise. Prices, discounts, delivery times and policies must come from your systems or approved text, never from the model's guess.
• Template spam. Sending marketing templates to people who didn't ask for them. It costs money, gets you blocked and can hurt your number's quality rating.
• Launching to everyone on day one. Skipping a pilot with real but limited traffic.
• Nobody reads the conversations. The first weeks of transcripts are the best source of improvements. Someone needs to review them.

Launch checklist

1. Meta Business account verified, and the phone number for the API decided (new number or migrated).
2. Display name and business profile approved, with hours, address and website.
3. List of the 20 to 30 most common questions, taken from real chats, with correct answers.
4. Integrations tested against real data: stock, orders, calendar, payments.
5. Rules written down for what the bot must never answer on its own.
6. Handoff working, with the full history visible to agents and business hours set.
7. Templates submitted and approved for the proactive messages you need (order updates, reminders).
8. Opt-in collected where you'll send proactive messages, and an easy way to opt out.
9. Privacy notice updated to mention automated answers and what data you store.
10. Pilot with a share of traffic or a limited schedule, then review transcripts daily for the first two weeks.
11. Basic metrics in place: conversations, share resolved without a person, handoff time, sales or bookings from WhatsApp.

We design and build these bots as part of our WhatsApp chatbot and AI automation services, connected to the systems you already use, with maintenance and hosting after launch.

Frequently asked questions

Keep reading