Blog

AI & Machine Learning

WhatsApp AI Assistants for Customer Service: Setup, Limits and Costs

WhatsApp AI Assistants for Customer Service: Setup, Limits and Costs

Run a WhatsApp AI assistant through the WhatsApp Business Platform: answer inside the 24-hour window, use approved templates outside it, and plan per-message fees.

Run a WhatsApp AI assistant through the WhatsApp Business Platform: answer inside the 24-hour window, use approved templates outside it, and plan per-message fees.

8 min read
Portrait of Omer Faruk Ilhan.
Umur Duman
Operations Lead

Blog

AI & Machine Learning

WhatsApp AI Assistants for Customer Service: Setup, Limits and Costs

Run a WhatsApp AI assistant through the WhatsApp Business Platform: answer inside the 24-hour window, use approved templates outside it, and plan per-message fees.

8 min read
Portrait of Omer Faruk Ilhan.
Umur Duman
Operations Lead
Dithered illustration of a marble hand posting a folded note into a carved lion's-mouth letterbox while a reply with a red wax seal slides back out.
logo of Facebook
logo of X
logo of LinkedIn
logo of Reddit
logo of Telegram
Contents
No headings found on page

To run an AI assistant on WhatsApp for customer service, connect it to the WhatsApp Business Platform, let it answer freely while a customer's 24-hour service window is open, and use pre-approved templates for anything you send outside that window. Since October 2026 Meta charges per message for service replies too, beyond a free monthly allowance.

Picture a clinic that gets most of its booking questions on WhatsApp. The receptionist answers between patients, so replies arrive hours later and some never arrive. The clinic adds an AI assistant that answers opening hours, prices and preparation questions, books slots into the calendar and passes anything medical to a person.

The first week goes well. Then the team tries to send next-day reminders to everyone with an appointment, and the messages fail, because free-form text can't go to someone who hasn't written in the last 24 hours. Nobody had read the template rules.

What is a WhatsApp AI assistant?

A WhatsApp AI assistant is software that reads incoming WhatsApp messages, answers them with an AI model using your business's own information, and takes simple actions such as booking or looking up an order. It connects through Meta's WhatsApp Business Platform, the API version of WhatsApp for businesses.

The WhatsApp Business Platform is Meta's interface for sending and receiving WhatsApp messages from your own software rather than from a phone. An assistant on it usually has three parts:

  • The model, which reads the message and writes the reply.

  • Your knowledge, such as prices, policies, opening hours and the order or booking system it may query.

  • The rules, which decide what it may say, what it may change and when it hands over to a person.

Meta now also offers its own option. Its developer documentation describes the Meta Business Agent as a way to "configure an AI agent that handles customer conversations 24/7 in your brand voice on WhatsApp", with the ability to hand control back to your app for specific events or flows. The choice is between Meta's agent and one you build or buy from a solution provider.

How do you set one up?

You need a WhatsApp Business account on the platform, a phone number registered to it, message templates approved for anything outside the service window, and an assistant connected to your systems. The order below avoids most rework.

  1. Pick the calls it should take. List the questions customers actually send and choose two or three with clear answers and a clear action, such as bookings or order status.

  2. Register the number and verify the business. Verification also raises how many people you can message first (see the limits below).

  3. Write the knowledge it may use. One source for prices, policies and hours, kept current by a named person.

  4. Connect the systems it acts on. Calendar, order system or CRM, through an automation layer your team can maintain.

  5. Create your templates. Reminders, confirmations and follow-ups that may go out after the window closes.

  6. Set the handover rules. Decide which topics go straight to a person and how the customer reaches one.

  7. Test with real conversations. Replay past chats, check every write the assistant makes, then go live on one topic first.

What is the 24-hour customer service window?

The customer service window is a 24-hour period that opens when a customer messages or calls you, during which you may send any reply without a template. Meta's pricing documentation says that if the user writes again before it expires, "the timer resets to 24 hours."

A fragmented marble face watching a sundial's shadow, with one open shutter lit red in a wall of closed ones

Outside that window the rules change. Meta's template guide states: "Template messages are the only type of message that can be sent to WhatsApp users outside of a customer service window." A template is a pre-written message, with optional fields, that Meta reviews before you can send it.

Template categories

Each template must be categorised as one of three types:

  • Marketing: offers, announcements and anything promotional.

  • Utility: updates about something the customer already did, such as a booking reminder or a shipping notice.

  • Authentication: one-time passcodes for logins and verification.

Templates are reviewed automatically when you create or edit them, and you can send one only once its status is approved. For an AI assistant this means the model writes freely inside the window, but anything it sends later has to fit a template you prepared in advance.

What does a WhatsApp AI assistant cost to run?

Running one costs Meta's per-message fees, plus the AI model or agent fees, plus whatever your solution provider charges. Meta has charged per message since 1 July 2025, and rates differ by country and by message category.

Meta's pricing documentation lists what changed and what stays free:

  • Service messages: from 1 October 2026 Meta "resumed charging, on a per-message basis" for service messages, the free-form replies inside the window. Each business phone number gets "a free tier of 1,000 service messages per month."

  • Utility templates inside the window: these were free before 1 October 2026; from that date they are priced at the regular rate.

  • Marketing, utility and authentication templates: charged per message at the rate for the recipient's market and the template's category. Meta publishes the rate cards on its WhatsApp Business website.

  • Free entry point: when someone messages you from a click-to-WhatsApp ad, the window "may remain open for up to 7 days", and messages in it are free, except messages sent by the Meta Business Agent.

  • Meta Business Agent: "one global rate of $2.00 USD per 1 million (M) tokens." Meta adds that one message typically uses 20,000 to 25,000 tokens, which it puts at about 4 to 5 US cents per message.

If you build your own assistant, the model's cost depends on the provider and on how much context each reply carries. Long chat histories and large knowledge files make every reply more expensive, so keep what the model reads short and relevant.

What limits apply on WhatsApp?

Two kinds of limits matter: how many people you can start conversations with, and what Meta allows on the platform at all. Both are easy to miss until a launch is blocked.

Messaging limits

Meta's messaging limits cap "the maximum number of unique WhatsApp user phone numbers your business can deliver messages to, outside of a customer service window, within a moving 24-hour period." The limit is shared by all numbers in your business portfolio. New portfolios start at 250, and the levels run 2,000, 10,000, 100,000 and unlimited. You reach 2,000 by verifying the business or by sending enough high-quality template messages; higher levels follow automatically when quality stays high and you use at least half your limit over seven days. Replies inside open windows don't count against it.

Policy rules

The WhatsApp Business Messaging Policy (last updated 23 September 2026) sets three rules that shape an assistant's design:

  • Opt-in. You may contact people only if they gave you their number and you "received opt-in permission" to message them.

  • Escalation. Automated replies must come with "prompt, clear, and direct escalation paths", such as a transfer to a human agent, a phone number, email or a support form.

  • Opt-out. You must respect any request to stop messages, whether it arrives on WhatsApp or elsewhere.

The AI Providers clause

Meta's terms for the WhatsApp Business Platform prohibit "AI Providers" from using the platform when AI is "the primary (rather than incidental or ancillary) functionality being made available." This is aimed at companies offering general-purpose AI assistants on WhatsApp. A business using AI to answer its own customers is a different case, and the terms let you retain an AI provider as your solution provider, but read the clause with your provider before you build. If you serve people in the EU, our post on EU AI Act Article 50 for chatbots covers telling users they are talking to AI. This is general information, not legal advice.

When should the assistant hand over to a person?

Hand over whenever the customer asks for a person, when the topic is on your never-automate list, or when the assistant isn't confident it has the right answer. The handover should carry the conversation with it, so the customer never repeats themselves.

A marble hand turning a revolving wheel in a convent wall to pass a scroll tied with a red ribbon to a second hand on the other side

Handover checklist

  • The customer can type "agent" or "person" at any time and reach one, as the policy's escalation rule expects.

  • Complaints, refunds above a set amount, medical, legal and account-security questions go to a person by default.

  • The person sees the full chat history and what the assistant already did.

  • Outside staffed hours, the assistant says when a person will reply and logs the request as a task.

  • Every conversation, its outcome and any changes the assistant made are written to the CRM.

  • Someone reviews a sample of conversations each week and updates the knowledge and rules.

How we approach it at Vizio AI

We start from the conversations, not the bot. With the client we export a sample of recent WhatsApp chats, sort them by topic, and pick one topic with a clear answer, a clear action and enough volume. The handover rules are written with the people who answer WhatsApp today, because they know the edge cases.

We then design the templates and the window logic before any model work, so reminders and follow-ups don't fail after launch. The assistant connects to the calendar, order system or CRM through an automation layer the team can maintain; our comparison of n8n, Zapier and Make explains how we choose. Before go-live we replay real chats, check every write, and set a weekly review. The same method sits behind our guide to AI voice agents for service businesses. This is part of our DX Studio work and our support for small and mid-sized businesses.

Sources

FAQs.

01

Can I use the WhatsApp Business app instead of the platform?

The WhatsApp Business app is built for a person replying from a phone, and it suits a small team answering by hand. An AI assistant that reads every message, queries your systems and writes to your CRM needs the WhatsApp Business Platform, which gives your software access through an API. Most businesses reach it through Meta's own tools or a solution provider.

02

Do customers have to opt in before the assistant messages them?

Yes for any message you start. Meta's policy says you may contact people only if they gave you their number and agreed to receive messages from you. When a customer writes to you first, the assistant can reply inside the 24-hour window. Reminders and follow-ups after that need both the opt-in and an approved template.

03

Can the assistant send payment links or take orders?

It can send links and collect order details in the chat, as long as the actions run through your own systems and the customer can see what they are agreeing to. Keep a person approving refunds, unusual orders and anything above a set amount, and log every order the assistant creates so mistakes are easy to find and reverse.

04

How do I stop the assistant giving wrong answers?

Give it one short, current source for prices, policies and hours, and tell it to hand over when the answer isn't there instead of guessing. Test it on real past conversations before launch, review a sample of chats every week, and fix the source rather than the prompt when an answer is wrong. Wrong answers usually trace back to stale information.

05

Does a WhatsApp assistant work in several languages?

Most current AI models can read and reply in many languages, so one assistant can usually answer customers in their own language. Templates are different: they are fixed text, so you need an approved template in each language you send reminders and follow-ups in. Check replies in every language you plan to support before launch, especially prices, dates and policy wording, which are where translation mistakes cost most.

01

Can I use the WhatsApp Business app instead of the platform?

The WhatsApp Business app is built for a person replying from a phone, and it suits a small team answering by hand. An AI assistant that reads every message, queries your systems and writes to your CRM needs the WhatsApp Business Platform, which gives your software access through an API. Most businesses reach it through Meta's own tools or a solution provider.

02

Do customers have to opt in before the assistant messages them?

Yes for any message you start. Meta's policy says you may contact people only if they gave you their number and agreed to receive messages from you. When a customer writes to you first, the assistant can reply inside the 24-hour window. Reminders and follow-ups after that need both the opt-in and an approved template.

03

Can the assistant send payment links or take orders?

It can send links and collect order details in the chat, as long as the actions run through your own systems and the customer can see what they are agreeing to. Keep a person approving refunds, unusual orders and anything above a set amount, and log every order the assistant creates so mistakes are easy to find and reverse.

04

How do I stop the assistant giving wrong answers?

Give it one short, current source for prices, policies and hours, and tell it to hand over when the answer isn't there instead of guessing. Test it on real past conversations before launch, review a sample of chats every week, and fix the source rather than the prompt when an answer is wrong. Wrong answers usually trace back to stale information.

05

Does a WhatsApp assistant work in several languages?

Most current AI models can read and reply in many languages, so one assistant can usually answer customers in their own language. Templates are different: they are fixed text, so you need an approved template in each language you send reminders and follow-ups in. Check replies in every language you plan to support before launch, especially prices, dates and policy wording, which are where translation mistakes cost most.

Read more.

Dithered illustration of a marble hand posting a folded note into a carved lion's-mouth letterbox while a reply with a red wax seal slides back out.
WhatsApp AI Assistants for Customer Service: Setup, Limits and Costs
Umur Duman
8 min

Run a WhatsApp AI assistant through the WhatsApp Business Platform: answer inside the 24-hour window, use approved templates outside it, and plan per-message fees.

Dithered cutaway illustration of a Baroque cathedral, a marble hand below ground setting a red cornerstone beneath the foundations while the spire rises above.
Sizing an AI Budget for 2027: Fund the Foundations, Not Just the Pilot
Orhan Gazi Yalcin
7 min

Plan a 2027 AI budget around a few use cases with owners and success criteria, and fund the data, integration, governance and running costs they depend on.

Dithered illustration of a marble hand guiding a pantograph across a Baroque drafting table, the copying arm drawing a larger outline in characters, its tracing point pink.
Can AI Design Tools Replace a Product Designer? Claude Design, Figma Make and v0
Umut Durgut
7 min

No. Claude Design, Figma Make and v0 turn a prompt into screens and working prototypes fast, but a designer still decides what to build and why it works.

Think AI-first.
Move future-fast.

Let’s build what’s next! From product ideas and intelligent workflows to data-driven systems and scalable growth operations.

Have a product idea, AI workflow, data system, or growth challenge that needs clearer execution?

Let’s shape the right system around it.

Think AI-first.
Move future-fast.

Let’s build what’s next! From product ideas and intelligent workflows to data-driven systems and scalable growth operations.