A WhatsApp AI chatbot is an assistant that answers customers on WhatsApp using a large language model (LLM), your own knowledge base and your business systems. It connects through the WhatsApp Business Platform (the Cloud API), replies freely within the 24-hour customer service window, uses approved templates for anything outside it and hands complex conversations to a human.
This guide covers how the platform works, Meta's rules in plain terms, the best use cases, how the AI side is built and what to know if your customers are in India.
What Do You Need to Run a WhatsApp AI Chatbot?
You need four things: access to the WhatsApp Business Platform, a verified business and phone number, an AI backend and a way for humans to take over.
WhatsApp Business app vs WhatsApp Business Platform
| WhatsApp Business app | WhatsApp Business Platform (API) | |
|---|---|---|
| Built for | Small businesses replying by hand | Medium and large businesses, or any business automating at scale |
| Automation | Greeting, away messages, quick replies | Full chatbots and AI agents via webhooks |
| Integrations | Minimal | CRM, helpdesk, ERP, e-commerce, payments |
| Multiple agents | Limited | Yes, through an inbox or helpdesk tool |
| Business-initiated messages | Manual broadcasts | Approved templates at scale |
For an AI chatbot, you need the platform. Meta now directs new integrations to the Cloud API, which Meta hosts, so you don't run your own WhatsApp servers. You can connect directly through Meta's developer tools or through a Business Solution Provider (BSP), which adds onboarding help, an inbox, analytics and billing in local currency.
Setup steps
- Create or use a Meta Business portfolio and complete business verification.
- Register a phone number for the API. Ask your provider whether an existing WhatsApp Business app number can be connected as is or must be migrated.
- Set your display name and business profile, which Meta reviews.
- Configure a webhook so incoming messages reach your chatbot backend.
- Create and submit message templates for approval.
- Connect your AI backend, knowledge base and business systems.
What Are WhatsApp's Rules for Business Messaging?
WhatsApp's rules center on consent and context: customers control when businesses can message them. Policies change, so always check Meta's current WhatsApp Business Messaging Policy before launch. The core concepts are stable. This section summarizes platform policies in general terms and is not legal advice.
The 24-hour customer service window
When a customer messages you, a 24-hour customer service window opens and resets with each new customer message. Inside the window, your chatbot can send any free-form reply: AI-generated text, images, documents, buttons and lists. When the window closes, you can't send free-form messages until the customer writes again.
Message templates
To start a conversation, or to message a customer after the window closes, you must use a message template approved by Meta. Templates fall into three categories:
- Marketing: offers, product announcements, re-engagement.
- Utility: updates about an existing transaction, such as order confirmations, delivery updates, appointment reminders and payment receipts.
- Authentication: one-time passcodes for login or verification.
Templates can include variables (name, order number, date) and buttons. Meta checks that the category matches the content; a promotion dressed up as a utility message may be re-categorized or rejected.
Opt-in and opt-out
You must get a customer's opt-in before sending them business-initiated messages. Collect it on your website, checkout, app, forms or in a WhatsApp conversation, and make clear that they will receive messages from your business on WhatsApp. Honor opt-out requests promptly. Respecting this also protects your number's quality rating: if many people block or report your messages, Meta can lower your quality rating and limit how many business-initiated messages you can send.
Pricing in general terms
Meta charges for template messages, with rates that vary by category and by the recipient's country. Free-form replies inside an open customer service window are generally free, and conversations that start from certain entry points, such as Click-to-WhatsApp ads, get extended free windows. The details have changed several times, so build your cost model from Meta's current rate card.
AI-specific policy
Meta's terms distinguish between businesses using AI to serve their own customers and providers offering a general-purpose AI assistant as the product itself. A chatbot that handles your support, bookings, orders and FAQs is the intended use. A general-purpose "ask me anything" AI service distributed through the Business Platform is not. Keep your bot focused on your business and review the policy as it evolves.
What Are the Best Use Cases?
The best WhatsApp chatbot use cases are frequent, time-sensitive conversations where customers already prefer chat over a call or email.
- Customer support: order status, returns, refunds, account questions, troubleshooting.
- Lead capture and qualification: respond instantly to ad clicks and website leads, ask qualifying questions and book meetings.
- Appointments: book, reschedule and confirm, with reminders as utility templates.
- Commerce: product questions, catalog browsing, cart recovery and cash-on-delivery confirmation; see AI for retail and e-commerce.
- Collections and payments: payment reminders and links, receipts and status checks.
- Internal use: field teams checking stock, submitting reports or getting answers from SOPs.
How Is the AI Side Built?
A production WhatsApp AI chatbot combines an LLM with your knowledge, your systems and clear rules. The flow for each incoming message:
- Receive. The webhook delivers the message (text, voice note, image, location or button reply).
- Understand. Voice notes are transcribed with speech-to-text; images can be read by a multimodal model.
- Retrieve. The bot searches your knowledge base (policies, product data, FAQs) for relevant passages. This retrieval-augmented generation (RAG) keeps answers accurate and current; see RAG for business.
- Act. If the customer needs something done, the bot calls tools: look up an order, check a slot, create a ticket, generate a payment link.
- Reply. The LLM writes a short, WhatsApp-friendly reply, using buttons or lists where a choice is needed.
- Log. The conversation is stored in your CRM or helpdesk with tags and a summary.
When the bot moves from answering to doing (booking, updating, refunding), it is closer to an AI agent than a classic chatbot. Our guide to AI agents vs chatbots explains the difference.
Design tips for WhatsApp
- Keep messages short. One idea per message; use buttons for choices.
- Use WhatsApp's interactive features (reply buttons, list messages and WhatsApp Flows for forms) instead of asking customers to type structured data.
- Confirm critical details back before acting: dates, amounts, addresses.
- Restrict what the bot may promise: discounts, refunds and delivery dates should come from your systems, not the model.
- Test with real past conversations and track resolution rate, handoff rate and customer ratings.
How Should Human Handoff Work?
Handoff should be fast, obvious and carry context, so the customer doesn't have to repeat themselves.
- Triggers: the customer asks for a person, the bot is unsure, the topic is sensitive (complaints, health, legal) or a high-value action needs approval.
- Routing: send the conversation to the right team in a shared inbox or helpdesk, with a summary and the data the bot already collected.
- Timing: handoff within the 24-hour window lets an agent reply freely; outside business hours, tell the customer when to expect a reply.
- Return: once the human resolves the issue, the bot can take over again for routine follow-ups.
What Should Businesses in India Know?
India is WhatsApp's largest market, and for many Indian customers WhatsApp is the default way to contact a business. That changes the design priorities.
- Languages. Expect Hindi, English, Hinglish and regional languages, often in Roman script. Test the bot on real mixed-language messages, not just clean English.
- Voice notes. Many customers prefer sending voice notes. Transcription that handles Indian accents and code-mixing matters; see our guide to voice AI in Indian languages.
- Payments. Eligible businesses can collect payments inside WhatsApp in India, including UPI, through supported payment partners. Otherwise, send secure payment links.
- Cash on delivery. COD confirmation and address verification over WhatsApp are high-value, low-risk automations for e-commerce.
- Data protection. The Digital Personal Data Protection Act, 2023 requires consent with clear notice, or another lawful ground the Act allows, for processing personal data. Keep consent records and limit what you store.
- Pairing with SMS and calls. Some customers still respond better to SMS or a phone call; a voice AI agent can cover the calls while WhatsApp handles chat.
Launch a WhatsApp AI Chatbot With Aaga
Aaga sets up the WhatsApp Business Platform, builds the AI on your knowledge base and connects it to your CRM, helpdesk or store, with templates, opt-in flows and human handoff included. Explore our WhatsApp services and AI chatbot development, or contact us to start with a free consultation and a scoped pilot on one high-volume conversation.
