Meta Business Agent: what it does, and when you still need something else

What Meta Business Agent actually is

On June 3, 2026 Meta announced its own AI agent for businesses, and it is not a side project. In Meta's own newsroom post, a Business Agent can "Answer questions specific to your business", "Make product recommendations from a business catalog", "Book appointments and qualify incoming leads", "Let you decide when a team member steps in", and "Close sales." Meta states that "more than one million businesses are already using a Meta Business Agent on WhatsApp and Messenger."

It learns your business from knowledge sources you provide: FAQs, business information, product catalogs, files, and websites, plus custom connectors to your external APIs. On top of that sits the Meta Business Agent Platform, which Meta says "lets businesses connect to a growing suite of hundreds of systems like Shopify, Zendesk, and Shopee, giving Business Agents the ability to take action on behalf of the business."

Two details are worth reading precisely, because they are where most secondhand summaries go wrong. Meta's newsroom describes the agent across WhatsApp, Messenger and Instagram, while the Business Agent Platform developer documentation describes the WhatsApp integration and gates it by country and by business vertical, with retail and ecommerce among the approved ones. And "getting started is free" is Meta's phrasing for activation, not for running it, which the next section takes apart.

Everything above is a genuinely capable product from the company that owns the channel. Any vendor telling you otherwise is protecting a price, not informing you.

Turn Meta's agent on before you pay anyone, including us

This is the recommendation, stated first so nobody has to dig for it. If you are a small store, activate Meta's Business Agent, point it at your business information and your catalog, and run it for a month. It costs nothing to switch on and it will answer a large share of what customers ask you. A month of it will teach you more about your own chat volume and question mix than any comparison page, this one included.

Do that first for a practical reason as well as an honest one: you cannot evaluate whether a paid tool is worth $29 a month until you know what the free option already covers.

If you follow this advice and never buy anything from us, this page did its job.

What it costs, now that running it is not free

Free to activate is not free to operate, and the gap arrived quickly. Meta began charging for Business Agent messages on August 1, 2026, on a per-token basis, publishing "One global rate of $2.00 USD per 1M tokens" and the estimate that "One message typically consumes about 20,000-25,000 tokens, which translates to approximately 4-5 cents (USD) per message." Meta documents this under non-template message pricing.

Two dated facts to hold on to. Since August 1, 2026 Meta charges for its own Business Agent messages at $2.00 per 1M tokens, roughly 4 to 5 cents a message by Meta's own estimate. And Meta has said that "in the coming months, businesses will access the agent through paid subscription offerings, with options for businesses of every size." The free activation you get today is explicitly a starting price, not a permanent one. Separately, from October 1, 2026 Meta begins charging for ordinary service messages on WhatsApp, which lands on every tool including ours: the full rule set is in our WhatsApp Business API pricing guide.

Four to five cents sounds like nothing until you multiply it, and the multiplier is not the number of customers, it is the number of messages the agent sends. The method is three terms:

Agent messages per conversation, times conversations per month, times $0.04 to $0.05.

We will not invent your volume, because we do not know it and neither does anyone quoting you a figure. The table runs the arithmetic at three clearly illustrative volumes. Substitute your own, which is the only version of it that means anything:

Two caveats. The messages-per-conversation figure moves everything, and it is the one number only your own transcripts can give you: open your last twenty chats, count the replies your side sent, take the median. And Meta's 4 to 5 cents is a typical figure, not a ceiling: a longer answer over a bigger knowledge base consumes more tokens.

Set that against a flat subscription and the difference is the shape of the bill, not only its size. Ours is $29 a month for 500 conversations, $79 for 2,000, and $199 for 6,000, and it does not move when a conversation runs long. A metered bill rewards short answers; a flat one does not care. Neither of us can waive Meta's own per-template message fees, which sit under both models.

Where a separate tool still earns its price

Each of these is a specific workflow, not an adjective. If none of them describes a problem you actually have, you do not need us yet.

• COD confirm or cancel before dispatch, tagged on the order. The order lands, the customer gets the items, the total and two buttons, and the answer is written back onto the order in your store as a tag your fulfilment queue can filter on. The value is not the question, it is the tag: unconfirmed orders never enter the pick list, so the parcel that would have come back never ships. That is a commerce workflow with a state change at the end of it, not a conversation. Full mechanism on the COD confirmation page.
• Abandoned cart recovery with timed reminders and resume links. A cart goes quiet, a reminder goes out on your schedule, and it carries a checkout link that restores that exact cart rather than dropping the customer on your homepage. Getting this right means listening for the abandonment event in the store, holding a timer per customer, generating a valid resume URL, and stopping the sequence the moment they buy. See cart recovery.
• Order status messages driven by store events. Fulfilment created, shipment out for delivery, delivery failed. These are outbound messages triggered by something happening in your store while nobody is chatting, which is the opposite of an agent answering an inbound question. Details on order updates.
• Stores that are not on Shopify. Meta's connector list is a Shopify-and-friends list, and it is a good one. If you run WooCommerce, a POS, or you sell with no website at all, the choice is to build a custom connector to your own API or to use something that already speaks your stack out of the box.
• Owning the logic and the conversation data. An agent that lives inside the platform keeps its instructions, its knowledge and its history there, next to the channel it runs on. A separate tool keeps them in your account, and keeps the commercial record, the orders, the tags, the customer history, inside your store. This is not a security argument. It is about which relationships survive a change of vendor.

Where a separate tool does not earn its price

If your store is on Shopify and what you want is product questions answered from your catalog, day and night, in the customer's language, Meta's free agent may be all you need. That sentence has no second half. Product recommendation over a catalog is precisely the job Meta built the agent for, the Shopify connector is Meta's own, and you can have it running today without a card, a contract, or a demo call.

Two more cases where the answer is to buy nothing. If you take a handful of chats a week, no automation will pay for itself, and answering them yourself is cheaper and better. And if your product descriptions are one line long, an agent of any brand will keep saying it does not have that detail, so the afternoon you would spend evaluating vendors is better spent writing product data. The catalog versus AI agent guide covers which of those is your real bottleneck.

Side by side: Meta Business Agent and AiBayya

Verified on August 5, 2026 against Meta's newsroom post and its developer documentation, and against our own product. The rows where we lose are in the table on purpose.

The strategic question: the platform is also a competitor

Meta owns the channel, the rate card, the customer's address book, and now an agent that competes with the tools built on its platform. That is not a scandal, it is the ordinary shape of a platform business, and Apple, Amazon and Google have each done the same to their own ecosystems. It is still worth pricing into a decision, in three specific ways.

• The price is Meta's to set. Business Agent messages went from free to roughly 4 to 5 cents in under two months, and Meta has already announced subscriptions on top. Anything built directly on the agent inherits that pricing authority with no say in it.
• Eligibility is Meta's to set. The developer documentation gates the agent by country and by vertical. If your category is not approved, the free option is not cheap, it is simply not available to you.
• The roadmap is Meta's to set. Features Meta ships arrive for a million businesses at once, which is a real advantage. Features Meta chooses not to build do not get built by anyone else inside its agent.

The measured conclusion is not to avoid Meta's agent. It is to use it without making it the only place your customer relationships and your commerce logic live. Keep order records and tags in your store, keep your customer list somewhere you can export, and treat any single-vendor stack as a decision you may have to revisit. The same caution applies to us: ask the export question before you sign up, not after.

How to decide, in one afternoon

Stores that end up paying us do so because pile two kept filling up. Stores whose pile two stays empty are better off with the free agent, and we would rather write that down than be the vendor who left it out.

Meta Business Agent questions