An AI chatbot for WhatsApp is an automated assistant that answers your customers inside the world’s most used messaging app, understands natural language thanks to models like GPT or Claude, and escalates only complex cases to a human. WhatsApp has over 2 billion users and business messages get open rates above 90% — far beyond email. This guide covers how it works, what you need to build one, tools with real prices, use cases by industry and the mistakes worth avoiding.
What is an AI chatbot for WhatsApp
It is a program connected to the WhatsApp Business API that receives your customers’ messages, interprets them with a language model and replies instantly with your business information: opening hours, prices, order status or appointment booking. Unlike the old rigid menu bots, it understands questions written any way the customer likes.
The key is combining three pieces: the official WhatsApp Business API (required for automation; the regular app does not allow it), an AI model that interprets and writes, and your knowledge base (catalog, FAQ, policies) the bot pulls answers from. Without that base, the bot improvises — and a bot improvising with customers is a liability.
How to build it step by step
The short path for a small business has four steps: sign up for the WhatsApp Business Platform (via a provider like 360dialog or Twilio), connect a workflow tool, plug in the AI model with your information, and run two supervised weeks before letting it fly solo.
1. Get WhatsApp Business API access
You go through a Business Solution Provider. Meta charges per conversation (cents per service conversation in most countries). You need to verify your company in Meta Business Manager.
2. Connect the AI workflow
The most common routes: n8n or Make (WhatsApp node + Claude/GPT node, full control and minimal cost), or managed platforms like Landbot or ManyChat that handle everything for a monthly fee. To launch without code, pick the second; to scale cheap, the first.
3. Knowledge base and tone
Prepare a document with your real FAQ, catalog and policies, plus a prompt that sets your brand tone and the rules: what it may answer, what it must hand off to a human, and what it must never promise.
Tools compared
| Option | Best for | Difficulty | Monthly cost |
|---|---|---|---|
| n8n + Claude/GPT | Full control, cheap scaling | Medium | $20-60 + API |
| Make + WhatsApp | No-code visual workflows | Low-medium | $9-30 + API |
| Landbot | Fast launch without a developer | Low | from ~$45 |
| ManyChat | Marketing campaigns on WhatsApp | Low | from ~$15 |
| Custom build | High volume + ERP/CRM integrations | High | $1,500-5,000 upfront |
Use cases that actually work
The highest-ROI deployments share a pattern: the bot handles tier 1 and humans keep the delicate cases. A clinic confirming and reminding appointments slashes no-shows; an online store answering “where is my order” frees the team from its most repeated question; a restaurant taking bookings stops losing customers outside opening hours. Typical running cost (Meta conversations + AI model) sits between $50 and $150 a month for a small business.
The expensive mistake is the opposite: automating complaints, negotiations or medical/legal topics. There the bot should say “let me connect you with a person” — and actually do it.
Frequently asked questions about AI chatbots on WhatsApp
Is it allowed to automate WhatsApp with a bot?
Yes, using the official WhatsApp Business API and respecting privacy law: disclose that it is a bot, offer a human handoff, and never message people who did not consent. Automating with the regular app or unofficial mods can get your number banned.
How much does an AI WhatsApp chatbot cost?
A small business can start at $50-150 a month (Meta conversations + AI model + workflow tool). A turnkey build by a freelancer usually runs $1,500-3,000.
Do I need to know how to code?
No: Make or Landbot work with visual interfaces. Coding is only needed for deep ERP/CRM integrations, and a freelancer can cover that.
Can the bot sell, or only answer questions?
Both: it can show your catalog, recommend products based on what the customer asks and send the payment link. WhatsApp-assisted sales convert especially well for local businesses and mid-ticket ecommerce.
Which AI model is best for WhatsApp?
For customer replies, fast and cheap models like Claude Haiku or GPT-4o mini cover 90% of cases at a fraction of the cost. Save the big models for complex conversations or sales copy.
Conclusion
An AI chatbot on WhatsApp is today’s best effort-to-result automation for small business customer service. To start this week:
- Get WhatsApp Business API access through an official provider
- Build the flow with Make or n8n and connect Claude or GPT to your FAQ
- Define what hands off to a human and run 2 supervised weeks
- Measure: response time, % solved by the bot, and bookings/sales generated
Keep going with our guides on AI for business, AI agents and AI tools for small business.
