A Million Businesses Now Let Meta Answer Their Messages

What Meta shipped

Meta Business Agent is an AI agent that answers your customers for you across WhatsApp, Messenger and Instagram. It handles questions, recommends products from your catalogue, qualifies leads, and can carry a sale to the point of purchase.

The numbers in Meta's own announcement are the part to sit with.

Over one million businesses are already running one on WhatsApp and Messenger. Not piloting. Running.

More than one billion threads with businesses happen every day across the three apps.

It is now open to businesses of any size, globally, set up in minutes, replying in the customer's local language in your tone of voice. It is free to activate, with paid tiers coming.

Meta is also making these businesses easier to find inside WhatsApp, by name in the search bar or by someone sharing a contact card in a chat. That last detail matters more than it sounds, and I will come back to it.

The thing nobody is saying out loud

Two weeks ago Google announced Search agents that phone businesses on a customer's behalf. Now Meta has an agent that answers on the business's behalf.

Put those together and you get the actual story of 2026. Both ends of the sales conversation are being automated at the same time, by different companies, for different reasons.

The customer sends an agent to gather facts. The business answers with an agent that supplies facts. Neither party is present for the exchange that decides the purchase.

That is not dystopian, it is just what happens when messaging volume exceeds human capacity. But it does change what you are competing on. When two machines exchange information, the winner is not whoever is more persuasive or more charming. It is whoever gave their machine better information.

Why the honest verdict is "yes, switch it on"

I am generally sceptical of automation that touches the customer relationship. Here I am not, for one boring reason.

The alternative is not a thoughtful human reply. The alternative is no reply until Tuesday. Most small businesses cannot answer Instagram DMs at 9pm on a Saturday, which is exactly when a lot of people message. An agent that answers accurately in ninety seconds beats a person who answers well in two days, because the second customer has already booked with somebody else.

This is the same mechanism as the five minute rule. Speed of response dominates almost every other variable in whether a conversation happens at all.

It also lands hardest in exactly the markets where messaging is the default channel rather than a fallback. If your buyers already contact businesses on WhatsApp, this is not a new channel, it is the main one getting an upgrade. That argument is set out in WhatsApp marketing.

What to fix before you switch it on

An agent is only as good as what you feed it, and the failure mode is worse than silence. A confidently wrong answer, sent instantly, in your brand voice, is a problem you have now published.

Your catalogue and prices. The agent recommends from your catalogue. Gaps and stale prices become confident misinformation. Same discipline as product data for agentic commerce.

The questions you actually get. Go through the last hundred messages and list the real questions. Not the ones you wish people asked. Delivery times, coverage areas, whether you take a particular job, minimum spend.

The boundary. Decide in advance what it must never do: quote outside a range, promise a date, discount, or answer anything with a legal or medical edge. Write the handover rule and test that it triggers.

The handover itself. When it escalates to a person, does anyone actually get the message? An escalation into an unwatched inbox is worse than no escalation, because the customer was told help was coming.

The strategic risk worth naming

Here is the part that should give you pause, and it is not about quality.

Meta is making businesses discoverable inside WhatsApp by name in the search bar. Read that again. A discovery surface is being built inside a messaging app, on top of an agent Meta provides, with paid tiers coming.

That is a familiar shape. Free reach, mass adoption, then the reach becomes a product you rent. It happened to Facebook Pages and it is the reason organic reach is what it is today, which is the argument in rented reach, owned audience.

Use the agent. It is genuinely useful and free right now. Just do not let it become the only way your customers can reach you. Keep the phone number, keep the email list, and keep the relationships somewhere you own.

What this does not replace

Agents handle the transactional layer: availability, price, spec, booking. They do not build the preference that decides which business gets messaged in the first place.

Somebody still has to want to contact you rather than a competitor, and that comes from reputation, recommendation and recognition built long before the conversation. Automating the reply does not automate being chosen. If anything, as the reply layer commoditises, the choosing layer is worth more.

Frequently asked

Will customers know they are talking to AI? Meta positions it as a business agent rather than a person. Be straightforward about it. Buyers are more tolerant of a fast, honest bot than a slow, evasive one.

Does it work outside English? Meta says it replies in the customer's local language, which is a significant advantage for anyone selling across markets.

Should we still pay a human to do this? For high value or sensitive purchases, yes, with the agent handling qualification and out of hours. For availability and price questions, the agent is better because it is instant.

What about the free tier ending? Assume it will. Build the workflow so the value is in your data and your process rather than in the free plan.

Can Fixora set this up? Yes. Catalogue and price data cleaned first, the real question list, the boundary rules, and a handover that lands somewhere a human is actually watching. Tell us your channels and get a fixed scope quote within 48 hours.