When a potential customer messages your business on WhatsApp, the clock starts ticking. If they wait an hour for a reply, there is a good chance they have already messaged a competitor. A WhatsApp AI agent fixes that: it replies in seconds, answers common questions, qualifies the lead and passes serious buyers to your team, day or night.
In this guide I walk you through exactly how to set one up properly, using the official WhatsApp Business Platform, in a way that is reliable, compliant with Meta’s rules and genuinely useful to your customers. I will also cover the no-code route if you would rather not touch any code.
WhatsApp Business App vs WhatsApp Business Platform
Before anything else, it helps to know which WhatsApp product you are working with, because only one of them supports a proper AI agent.
- WhatsApp Business App is the free app on your phone. It offers greeting messages, away messages and quick replies, which is great for small teams, but it cannot connect to your own AI system or CRM.
- WhatsApp Business Platform (Cloud API) is the developer platform from Meta. It lets software send and receive messages automatically, which is what you need for a real-time AI agent, CRM integration and multiple team members.
Everything below uses the WhatsApp Business Platform. You can access it directly through Meta, or through an official Business Solution Provider (BSP) that adds a dashboard and no-code tools on top.
What you need before you start
- A Meta Business portfolio (formerly Business Manager) for your company.
- A phone number that can receive an SMS or call for verification and is not currently registered on the WhatsApp app (or that you are willing to migrate).
- Your business details for verification: legal name, address, website and documents. Business verification unlocks higher messaging limits.
- A clear list of the questions customers ask most, your prices or packages, service areas and the rules for when a human should step in.
Step 1: Create your Meta Business portfolio and verify your business
Go to Meta Business Suite and create a business portfolio if you do not already have one. Then start Business Verification in the security centre. Verification is not always required to begin testing, but it increases the number of people you can message and builds trust in your display name.
Step 2: Create a Meta developer app and add WhatsApp
Visit the Meta for Developers dashboard, create a new app, choose the business app type and connect it to your business portfolio. Add the WhatsApp product to the app. Meta will create a test phone number and a WhatsApp Business Account (WABA) so you can start experimenting immediately.
Step 3: Add and register your business phone number
In WhatsApp Manager, add your real business number, verify it with the code Meta sends, and set your display name. The display name must match your business, and Meta reviews it. Once approved, note down two values from the API Setup page: your Phone Number ID and your WhatsApp Business Account ID. Your AI agent will use these to send messages.
Step 4: Create a permanent access token
The token shown on the API Setup page expires after 24 hours, which is fine for testing but not for a live system. In your business settings, create a System User, assign it to your app and WhatsApp account, and generate a token with the whatsapp_business_messaging and whatsapp_business_management permissions. Store it securely, for example in your server’s environment variables, and never paste it into a website or share it in chat.
Step 5: Set up the webhook so you receive messages in real time
A webhook is the address Meta calls every time a customer sends you a message. In your app’s WhatsApp configuration, enter your callback URL (the endpoint on your server or automation tool) and a verify token you choose. Then subscribe to the messages field. From this moment, every incoming message arrives at your endpoint within a second or two.
Step 6: Build the AI agent brain
This is where the “AI” happens. When a message arrives, your system should:
- Identify the customer by phone number and look up any history in your CRM.
- Understand the intent: a price question, a booking, a support issue or a new enquiry.
- Answer from your own knowledge: give the AI model your FAQs, services, pricing rules and policies (often called retrieval-augmented generation), and instruct it to answer only about your business.
- Qualify the lead with a few natural questions such as budget, timeline and location.
- Decide on a handoff: if the customer asks for a person, sounds unhappy or looks like a high-value lead, notify your team and pause the AI.
Keep the agent focused. It should behave like a well-trained front-desk assistant for your business, not a general chatbot that answers anything. That is better for customers, and it matters for Meta’s policy, which I cover below.
Step 7: Send replies through the Cloud API
Your AI agent replies by sending a request to the WhatsApp Cloud API messages endpoint using your Phone Number ID and access token. Within 24 hours of the customer’s last message (the customer service window) you can send free-form replies. Outside that window, you must use a pre-approved message template, for example a follow-up or appointment reminder.
Good replies are short, friendly and specific. Use buttons or lists for common choices, such as “See packages”, “Book a call” or “Talk to a person”, so customers can reply with one tap.
Step 8: Connect your CRM and human handoff
An AI agent becomes truly valuable when it feeds your sales process. Save every conversation and lead to your CRM, tag the lead with intent and score, and alert the right team member instantly for hot leads. If you run ads, sending qualified-lead events back to Meta and Google helps your campaigns find more customers like them. I cover this in more detail on my sales and marketing automation page.
Step 9: Test, launch and improve
- Test with real questions from your team, including awkward ones, before going live.
- Check that the handoff works and that someone actually picks up.
- Launch to a small share of traffic first, such as one ad campaign or one landing page.
- Review conversations weekly, improve answers and add missing information to the knowledge base.
The no-code option
If you do not have a developer, many official Business Solution Providers offer built-in AI agents, shared team inboxes, templates and CRM integrations. You connect your number, upload your FAQs and set handoff rules from a dashboard. You pay the provider’s subscription on top of Meta’s messaging fees, but you can be live within days.
Rules you must follow in 2026
- Business-focused AI only. From 15 January 2026, Meta’s WhatsApp Business Solution Terms stop general-purpose AI assistants from running on the Business Platform. AI agents that support your own business, such as customer service, bookings, order updates and sales enquiries, remain allowed.
- Opt-in. Only message people who have contacted you or agreed to hear from you.
- Templates outside 24 hours. Proactive and follow-up messages need approved templates.
- Pricing changes. Meta moved to per-message pricing in July 2025 and continues to update its rates. Always check the official WhatsApp Business Platform pricing page for your country before you launch.
Frequently asked questions
Can I use the normal WhatsApp Business App for an AI agent?
Not for a custom AI agent. The app supports quick replies and away messages, but connecting your own AI and CRM requires the WhatsApp Business Platform, either directly through Meta or through a provider.
How fast does a WhatsApp AI agent reply?
Typically within a few seconds of the customer’s message, depending on how much information the agent needs to look up before answering.
Will customers know they are talking to AI?
Being transparent builds trust. A short line such as “I’m the virtual assistant for [your business], and I can connect you to the team at any time” works well and sets the right expectations.
How much does it cost?
Costs combine Meta’s per-message fees for templates (which vary by country and message type), any provider subscription, and the AI model usage. For most small and mid-sized businesses, the cost is small compared with the value of answering every lead instantly.
Want this set up for your business?
I help premium D2C, B2B, fintech and service brands connect their ads, WhatsApp, CRM and sales team so no enquiry goes unanswered. If you would like a WhatsApp AI agent built around your business, apply to work with me and I will review your current setup personally.