Meta Business Agent: What It Is, How It Works, and What It Costs for Meta's WhatsApp AI Agent

Meta is bringing artificial intelligence directly into the conversations between businesses and customers. On June 3, 2026, at its Conversations event in London, Meta introduced Meta Business Agent, an AI agent designed to represent businesses on WhatsApp, Messenger, and Instagram, answering questions and taking part in sales and customer service conversations.

The launch matters because it shows where WhatsApp Business is heading: the conversation is no longer limited to exchanging messages or automating simple replies.

There is an important distinction that does not always come through clearly, however. On one hand, Meta offers an experience that certain businesses can set up directly from the WhatsApp Business app. On the other, it introduced Meta Business Agent Platform, an infrastructure aimed at organizations that need to build, customize, and integrate their agents at a larger scale.

Understanding that difference, and what each one costs, is key to knowing what this technology can really do and what kind of implementation each business needs.

What is Meta Business Agent?

Meta Business Agent is an artificial intelligence agent developed by Meta that can represent a business in conversations with customers on WhatsApp, Messenger, and Instagram.

Unlike automation based only on predefined replies, the agent uses business information to understand questions and generate answers adapted to the context of each conversation. Meta explains that it can learn from existing content, such as Facebook posts, past WhatsApp conversations, website information, and product catalogs.

Meta's bet is significant given the scale of the ecosystem it is being built into: at launch, the company said that more than a million businesses were already using a Business Agent on WhatsApp and Messenger, within an ecosystem with more than a billion active conversations per day across WhatsApp, Messenger, and Instagram.

What can Meta Business Agent do?

Customer service

For frequent questions or situations that do not require immediate involvement from a person, the agent can answer questions about hours, location, products, services, and prices. It can also run 24 hours a day and hand cases off to a person when the conversation calls for it.

Sales

The agent can understand what someone is looking for, ask questions to get more information, recommend products, and share relevant commercial information. Meta also indicated that Business Agent can qualify leads, book appointments, and close sales, although these functions depend on availability, the market, and the specific configuration of the agent.

For example, if someone writes "I'm looking for a product for my business, but I don't know which one I need," the agent can identify the need, ask additional questions, recommend an available option, and collect the data needed to continue the process.

Operations

In enterprise deployments, Meta Business Agent Platform is designed to connect with hundreds of systems and tools. Meta publicly mentions platforms such as Shopify, Zendesk, and Shopee, with the goal that agents can not only converse but also act on behalf of the business within defined limits. That is where the real difference begins between responding intelligently and taking part in a commercial operation.

How does Meta Business Agent work?

At a conceptual level, the process is basically this: customer writes → the AI interprets the request → it checks the business's knowledge → it generates a response → it collects information or recommends an action → the conversation continues or is handed off to a person.

The central piece is the knowledge that feeds the agent. Meta Business Agent can learn automatically from sources linked to the business, such as the Facebook page, past WhatsApp conversations, and the website, and the business can add catalogs, price lists, documents, and other sources.

This means the quality of the agent does not depend only on the AI model used: it also depends on how clear, up to date, and consistent the information it receives is. If a price list is outdated or the catalog has incomplete information, the AI will have a less reliable basis for answering. Deploying an agent should not consist only of "turning on the AI"; it also requires defining what the official sources of information are.

How is Meta Business Agent configured and controlled?

For eligible businesses, setup is done from the WhatsApp Business tools, with no coding required: the business turns on the agent, lets it learn from existing information, adds extra sources, tests the responses, and finally enables it to interact with customers.

But letting an AI represent a business requires control, which is why Meta documents four main areas:

1. Knowledge control

The business decides what information the AI will use to learn about the business.

In addition to certain existing sources, data such as website information, products, and price lists can be added.

2. Personality control

It is also possible to define how the agent should express itself.

The business can provide instructions related to tone and edit responses to help future interactions better represent the brand's way of communicating.

3. Audience control

Meta lets you configure who the AI will serve.

Depending on the available options, the agent can respond to everyone, only to new customers, or to people who arrive from ads.

4. Handoff control

Not every conversation should stay in the hands of artificial intelligence.

The business can set situations or topics in which the agent must transfer the conversation to a person. It can also step in manually whenever it considers it necessary.

Before turning it on, Meta lets you test the agent in a sandbox environment and fix incorrect information, unnatural responses, or situations that should end in a handoff to a person.

Meta Business Agent vs. Meta Business Agent Platform

This is probably the most important distinction for understanding Meta's proposal.

Meta Business Agent in the WhatsApp Business app aims to offer a simple way to bring AI into conversations, for simpler operations.

Meta Business Agent Platform, on the other hand, is infrastructure for building, customizing, and deploying agents at scale and connecting them to business systems, with controls, guardrails, and measurement aimed at organizations with more complex needs.

‍
Neither option is necessarily better than the other; they respond to different levels of complexity. And capabilities keep evolving: WhatsApp's official page states that Meta Business Agent is available to eligible businesses in select markets, and that some features may still not be available to all of them.

When is Meta Business Agent enough, and when do you need a more integrated operation?

A relatively small business may mainly need to answer frequently asked questions, show products, give prices, collect customer data, and stay available outside business hours. For those cases, a simple Meta Business Agent setup can cover a good part of the need.

The scenario changes when WhatsApp is part of a broader commercial operation: checking the CRM, updating opportunities, validating inventory in real time, generating quotes, booking time slots based on real availability, distributing conversations across teams, measuring which campaigns generate sales, or recovering opportunities that stopped responding.

At that point the conversation stops working in isolation and needs to connect with other systems and processes, the kind of scenario in which specialized platforms like Atom can be part of the architecture, making it possible to deploy AI agents on WhatsApp connected to a CRM, calendars, and other tools in the sales process.

The question should not simply be which agent answers better, but what needs to happen operationally as a result of each conversation.

What should you evaluate before deploying an AI agent on WhatsApp?

Before choosing a technology, it helps to have clear answers to these questions

A business should have clear answers to a few fundamental questions:

1. What goal should the agent fulfill? Support, qualifying opportunities, selling, scheduling, customer service, or a combination of processes.
2. What information does it need to check? Products, prices, policies, customer data, inventory, history, or other sources.
3. What actions should it be able to take? Collecting information is not the same as creating an opportunity, booking an appointment, or updating a system.
4. What tools does it need to connect to? CRM, e-commerce, calendar, support platform, quoting system, or other applications.
5. When should a person step in? Handoff should be part of the agent's design, not just a response to an error.
6. How will it preserve context? When a conversation passes to a rep, it should be possible to understand what the customer asked, what the AI answered, and what needs to happen next.
7. How will results be measured? It is not enough to know how many messages the artificial intelligence answered. It also matters how many conversations moved toward a quote, an appointment, an opportunity, or a sale.
8. How will knowledge be controlled? Sources must be kept up to date, and there must be a process to review responses and correct behavior.

‍

These decisions have more impact on the final result than simply turning on an AI feature.

How much does Meta Business Agent cost?

When Meta introduced Business Agent in June 2026, it launched it free of charge during a testing period. That period has ended, and since August 1, 2026 Meta charges for its use at $2 USD per million tokens processed, a single rate that bundles AI processing and message delivery. In practice, this works out to roughly 4-5 cents of a dollar for a simple conversation, and up to about 24 cents for a longer, more complex sale.

In addition, starting October 1, 2026 Meta will also begin charging for service messages that do not go through the AI agent (answered by a person or by third-party tools), eliminating the free window that existed for that type of message.

As with any Meta pricing policy, terms may keep changing; it is worth checking the official WhatsApp Business Platform documentation before projecting costs for a specific operation.

Meta Business Agent shows where WhatsApp is heading

The goal is no longer just to automate replies. AI is starting to take part in conversations where it needs to understand an intent, use business knowledge, recommend a solution, collect information, and decide what the next step should be. Meta Business Agent Platform takes that idea a level further, with agents able to connect to enterprise systems and execute actions within a larger infrastructure.

For businesses, the important decision is no longer just whether to use AI on WhatsApp, but what role that AI should play within the sales process, for example, what it needs to know or which systems it should access.

Once those answers are clear, WhatsApp stops being just a channel where messages arrive and becomes an integrated part of the sales process.

Frequently asked questions about Meta Business Agent

What is Meta Business Agent?

It is a Meta artificial intelligence agent able to represent a business in customer conversations on WhatsApp, Messenger, and Instagram, where it answers questions, requests information, and takes part in sales and service conversations using knowledge provided by the business.

What can Meta Business Agent do on WhatsApp?

Answer questions about the business, products, or services, give pricing information, recommend options, collect customer data, and transfer conversations to people. Meta also announced lead qualification, appointment booking, and sales, although some capabilities vary by business and market.

What is the difference between Meta Business Agent and Meta Business Agent Platform?

Meta Business Agent is set up directly from the WhatsApp Business app for eligible businesses. Meta Business Agent Platform is aimed at enterprise deployments that need to build, customize, and deploy agents at scale, connecting them to external systems with enterprise-level controls.

Can Meta Business Agent connect to a CRM?

The enterprise version, Meta Business Agent Platform, is designed precisely to connect to enterprise systems, with a growing network of integrations (Meta mentions examples such as Shopify, Zendesk, and Shopee). Whether it can integrate with a specific CRM depends on the implementation available.

If your business needs to take that next step in its operation, find out how to deploy AI agents connected to your sales process on WhatsApp with Atom.

Book a demo with our team