Meta just put an AI agent inside WhatsApp. Rent it or build your own?

WhatsApp has spent a decade as plumbing. Businesses used it to push order confirmations, shipping updates, and the occasional support reply, and that was about the whole job. On June 3, Meta rewrote the job description. It dropped an AI agent into WhatsApp Business, available everywhere at once, that can answer customer questions, recommend products, book appointments, and qualify sales leads on its own.

Meta calls it the Business Agent, and it spent nearly two years quietly testing the thing in markets like India and Mexico before flipping it on for the rest of us. It runs in Instagram DMs too. The pitch to a small business is hard to argue with: point it at your catalog, and the channel three billion people already check every day starts selling and supporting for you, around the clock, in whatever language the customer happens to type in.

If you have read anything we have written about Telegram turning into an app store or Shopify letting merchants sell inside ChatGPT, this will sound familiar. The messaging app is becoming the storefront, and the agent is becoming the salesperson. But the decision underneath it is not whether to use this. It is whether to rent Meta's version or build your own. Those are very different bets, and the easy one is not always the right one.

What Meta actually shipped

The Business Agent is Meta's old customer-support bot grown up. By Meta's own description it answers questions, recommends products from your catalog, books appointments, qualifies leads, and hands the conversation to a human when it gets out of its depth. There is a daily-briefing feature in testing that summarizes the chats that came in overnight. The roadmap Meta has talked about goes further still: managing customer calendars, doing market research, and a platform for bigger companies to build custom agents that plug into Shopify, Zendesk, and Shopee.

Read that list again and notice what it is not. It is not a chatbot that deflects FAQs so your support team can go home earlier. It is Meta trying to make WhatsApp the software a small business runs its front desk on. That is a much bigger ambition, and it is the reason this deserves a few minutes of thought rather than a shrug and a free-tier signup.

The reason it is tempting is reach. WhatsApp has more than three billion monthly users and is the default messaging app in over a hundred countries. For a lot of businesses it is the channel customers actually open, as opposed to the email that goes unread or the app nobody downloads. An agent that lives right there, that a customer can message the way they message a friend, takes a real amount of friction out of the path to a sale. I get the appeal. I just think the appeal is also where the trap is hiding.

The free part is the on-ramp, not the deal

Meta is not running this out of generosity, and the way it charges tells you how it sees your business. The Business Agent is going into paid tiers of WhatsApp Business Premium, and Meta has said larger businesses will pay based on how many tokens the agent burns through. So the more your agent talks, the more you owe, on a meter Meta sets and reads.

And that sits on top of a bill you were already paying. WhatsApp's Business Platform was never free messaging. Since Meta switched to per-message billing in 2025, you pay for most business-initiated messages, and the price swings on category and country: a marketing message runs around two and a half cents in the US and north of twelve cents in Germany, with utility and support messages a good deal cheaper. Replies inside a 24-hour service window are free, and there is a 72-hour free window when someone reaches you from a click-to-WhatsApp ad. None of this is outrageous on its own. It is just worth seeing the whole picture before you wire your business onto it: you pay to send the messages, and now you pay again for the agent that writes them.

Here is the part I keep coming back to. The message fees are a toll you cannot dodge if you want to be on WhatsApp, because it is Meta's pipe and Meta charges for the pipe. Fine. The agent fee is a different animal. That is the part you can either rent from Meta or build yourself, and it is also the part that holds your actual logic: how you qualify a lead, how you handle a booking, what your business says back when a customer is upset at 11pm. Renting that out is a bigger decision than the per-token price makes it look.

Rent for reach, build for logic

Renting Meta's agent is the right call more often than founders like to admit, for one kind of business. If messaging is a support channel rather than your product, if your volume is modest, if the questions are mostly 'where is my order' and 'are you open Sunday,' then a configured Meta agent is faster and cheaper than anything you would build, and you should just use it. Spending six figures on a custom agent so you can answer store-hours questions is the sort of thing that sounds clever in a planning meeting and looks ridiculous a year later.

Building your own starts to make sense when the conversation is where your product actually happens. If you run a booking platform, a clinic, a commerce business with real inventory and pricing rules, the agent is not answering questions about your product. It is operating your product. At that point you want it wired into your systems, your live availability, your discount logic, your patient records, and you want to decide exactly how it behaves when things get strange. That is very hard to do as a tenant inside someone else's agent, configuring around the edges of choices Meta made for its own reasons.

If you want a quick gut check, ask what breaks when the agent gives a slightly wrong answer. If the worst case is a customer who waits a few minutes for a human to clarify your return policy, rent it and sleep fine. If the worst case is a double-booked surgery slot, a mispriced order that actually ships, or a refund that should never have fired, you are not looking at a support feature anymore. You are looking at software that runs your operations, and the most important conversations your business has should not live inside a tool you cannot see into or change.

There is a quieter reason to build, and it is the one I would weigh most. Meta's agent works on Meta's surfaces. Rent it, and you have an agent on WhatsApp and Instagram, and then you start the whole exercise over for Telegram, for your website, for the ChatGPT storefront everyone is suddenly talking about. Build your own brain once, behind your own systems, and you can point it at all of those channels, WhatsApp included, through its Business API. You rent the pipe either way. The question is whether you also rent the intelligence that runs through it.

The part that should make you slow down

When a customer talks to Meta's agent, Meta is sitting in the middle of that conversation. You may keep the sale, but you are renting the relationship, the data, and the next-best-thing-to-say from the company that owns the channel. We made the same point about AI agents booking appointments, and it keeps being true across every flavor of agentic commerce: the convenient layer in the middle is convenient for a reason, and the reason is not always you.

Then there is the security angle, which almost nobody prices in up front. An agent that reads customer messages and takes actions, books a slot, pulls up an account, starts a refund, is a new attack surface. A customer, or someone pretending to be one, can try to talk it into doing something it should not. Renting Meta's agent hands a chunk of that problem to Meta, which is a real point in renting's favor. Building your own means you own the defense. Neither option makes the risk vanish. Both of them make you decide, on purpose, who is holding it.

For a lot of teams the answer is going to be both, and that is fine. Rent Meta's agent to be present on WhatsApp by next week, because being reachable matters and the on-ramp really is cheap. Build your own where the logic and the relationship are worth owning, and aim both at the same backend. The mistake is not renting. The mistake is renting the part that should have stayed yours, your booking flow, your customer data, the actual conversation your business runs on, because the free tier was sitting right there and signup took ten minutes.

Meta has made the easy choice very easy, which is exactly the moment to ask what you are handing over to get it. If WhatsApp is just where you answer questions, rent the agent and get on with your day. If the conversation is where your product lives, that is not a feature to rent. That is the product.