WhatsApp Automation 6 min read
How to Integrate an AI Chatbot With WhatsApp Business API
To integrate an AI chatbot with WhatsApp, you connect your business number to Meta’s WhatsApp Business Platform, point an AI agent at your own content, and decide when a person takes over. Meta’s Embedded Signup lets you do all of it without a developer.
This guide walks through each step as it works in 2026, using ZynfoAI as the example. If you want the overview first, start with our guide to the WhatsApp AI chatbot.
Why you need the Business Platform instead of the app
The WhatsApp Business app on your phone has greeting messages, away messages and quick replies. It can’t be connected to an outside AI agent. For that you need the WhatsApp Business Platform, where Meta’s Cloud API passes each incoming message to your software and sends its replies back.
| WhatsApp Business app | Business Platform (Cloud API) | |
|---|---|---|
| Who replies | You, on your phone | An AI agent, plus your team on the questions it cannot answer |
| Automation | Greeting and away messages, quick replies | Answers from your content, order lookups, lead capture |
| Messages you start | Sent by hand | Approved templates, sent as campaigns (on the Pro plan in ZynfoAI) |
| Setup | Install the app | Embedded Signup from your chatbot tool |
Before you start
- A Facebook or Meta business login with access to your business portfolio.
- The phone number customers will message, and a way to verify it.
- Your best content: delivery, returns and pricing pages, product guides, policies.
- A list of topics that should always go to a person.
Step 1: Connect your number with Embedded Signup
Embedded Signup is Meta’s own sign-up flow, opened from inside your chatbot tool. In it you:
- Log in with your Facebook or Meta business credentials.
- Select your business portfolio, or create one.
- Select your WhatsApp Business account, or create one.
- Enter and verify your business phone number.
- Enter the display name customers will see.
When you finish, the number is connected to ZynfoAI and the WhatsApp Business account stays under your own Meta business portfolio. If a developer has already set up the Cloud API for you, ZynfoAI also accepts those credentials entered manually.
Step 2: Train the agent on your content
The agent can only be as accurate as what you give it. In ZynfoAI, add:
- Website pages. ZynfoAI finds pages through your sitemap and you choose which to import. Starter allows up to 50 knowledge sources.
- Files. PDF, Word and Excel documents such as warranty terms or price lists.
- Google Drive. The Google Docs and Sheets you pick.
- Shopify. With the ZynfoAI Shopify app, your product catalog and order tracking.
When your content changes, press Resync on that source so the agent works from the new version.
Then write short instructions. WhatsApp replies should be two or three sentences. Ask the agent to offer a button or list when there’s an obvious next step. WhatsApp’s reply buttons allow up to three choices, such as “Track order”, “Returns” and “Talk to the team”.
Step 3: Decide what goes to your team
Write down which questions the agent should not answer: damaged items, complaints, anything about money. In ZynfoAI, a question the AI agent cannot answer becomes a help-desk ticket for your team, with the WhatsApp conversation attached, so nobody asks the customer to repeat themselves.
Live chat handoff, where a teammate takes over the WhatsApp chat itself and the AI stops replying, is on the Pro plan. Read more about human handoff.
Step 4: Test, then go live
Open the Playground and ask the questions you get most, typed the way customers type them: “where’s my order 1042”, “do u deliver sunday”. Read each answer as a customer would. Where an answer is wrong or vague, fix the source content rather than the instructions.
Once you go live, the agent answers messages to your number, and the questions it could not answer arrive as help-desk tickets. Read the first week of tickets and fill the gaps in your content.
Step 5: Plan the messages you start
Inside the 24-hour customer service window, replies can be free-form. Anything you start outside it, like a back-in-stock note or a promotion, has to be a Meta-approved template, sent only to people who opted in. ZynfoAI lists your templates as Approved, Pending or Rejected and sends campaigns to your contacts. The free WhatsApp template generator helps you draft them.
What it costs
There are two separate bills. ZynfoAI costs from $39 a month, with no per-resolution fees and a 7-day free trial with no card; your team’s replies don’t count toward the AI reply allowance. See pricing.
Meta bills its WhatsApp charges directly. Under its per-message pricing, you pay when a template message is delivered, at a rate set by category and the recipient’s country. Since 1 October 2026, Meta also charges per message for free-form replies inside the 24-hour customer service window. Our WhatsApp Business API pricing guide has a worked example. Estimate your monthly spend with the WhatsApp pricing calculator.
Common mistakes
- Going live before the content is ready. If your returns page is vague, the answers will be too.
- No way to reach a person. Always let customers ask for someone, and make sure someone is there in working hours.
- Long replies. WhatsApp is read on a phone. Keep answers short and offer buttons.
- Treating templates as an afterthought. Template review takes time. Submit the ones you need before you plan a campaign.
For the wider picture, from use cases to costs, read the WhatsApp AI chatbot guide. Ready to connect your own number? See how ZynfoAI for WhatsApp handles answers, Shopify orders and help-desk tickets.
Tools and guides
Put this into practice
AI WhatsApp Reply Generator
Generate high-converting, professional WhatsApp sales and support replies with AI.
SolutionWhatsApp AI Chatbot Platform
Automate customer support and sales over official WhatsApp Business API.
Free ToolWhatsApp Chat Widget Generator
Build a custom floating WhatsApp button to embed on WordPress or Shopify.
FeatureAI with Human Handoff
Escalate complex chats to live agents with complete conversation context.


