LONDON — The small business had a familiar problem: customers asked the same five questions all day, and a human could not be in three conversations at once. The shop owner set up a WhatsApp Business Agent, configured it with the price list and opening hours, and watched it answer the questions, book appointments, and hand a confused customer to a human before the customer could get annoyed. The setup took an afternoon. The customer never noticed the difference.
Meta this week made that scenario available everywhere, launching its Business Agent platform globally on WhatsApp. The product lets any business deploy an AI agent that handles consultations, takes bookings, screens leads, and transfers conversations to human staff when needed, all inside the messaging app that more than two billion people use every day. The launch turns WhatsApp from a place where people talk to each other into a place where businesses talk to customers, with Meta’s AI doing the talking.
The timing is deliberate. Meta has spent two years building the infrastructure: business accounts on WhatsApp, catalog and payment tools, and the AI models to run conversations. The Business Agent is the layer that ties them together, and it is the first Meta AI product built to generate revenue directly from conversation volume. The company has said premium capabilities will eventually be folded into WhatsApp Business Premium, the paid tier for businesses, creating a direct line from agent usage to Meta’s income statement.
Meta’s route to AI revenue runs through distribution rather than enterprise sales. OpenAI sells its agents through corporate software contracts, with sales teams, pilots, and procurement cycles. Meta’s model is different: the businesses that will use WhatsApp agents already exist, already pay for business messaging tools, and can switch them on without a procurement department. Analysts said the comparison captures the two strategies that will define the consumer AI market, with OpenAI building a SaaS business and Meta building a payments business around conversation.
The revenue potential rests on WhatsApp’s scale. Two billion daily active users means that in much of the world, WhatsApp is the internet, the place where commerce, communication, and customer service all happen. A business in Indonesia, Brazil, or India does not need to build a website to sell; it needs a WhatsApp account. Meta’s agent plugs into that existing behavior, which is why the company describes the launch as turning on the lights rather than building the building.
Customer service is also the AI application with the clearest economics. Unlike creative tools or assistants whose value is diffuse, an agent that answers a support ticket costs a fraction of a human agent and is available around the clock. Meta is entering a market already served by startups selling customer-service AI, but it has advantages none of them can match: the platform where the conversations happen, the payment rails, and the identity system. Competitors build software on top of other companies’ networks; Meta owns the network.
The risks are as visible as the opportunity. AI agents in customer service have a failure mode that is uniquely damaging: they are confident, they are wrong, and they do so in front of a customer. A bad agent experience on WhatsApp is not a support ticket; it is a conversation shared with the customer’s family group chat. Meta has built escalation paths into the product, routing uncertain conversations to humans, but the reputational damage from a widely shared bad interaction is a real cost for both Meta and the businesses that use the platform.
Regulators are also watching. WhatsApp has been a target of privacy scrutiny for years, and an AI agent that collects customer data on behalf of businesses raises new questions about who owns that data and what it can be used for. Meta has said the agents operate under the same data rules as business messaging, with businesses responsible for their own customer information. Whether regulators accept that framing is an open question in the EU, where Meta’s data practices are already under pressure.
The product’s history shows how deliberately Meta has built toward this moment. WhatsApp Business began as a free tool for small merchants, then gained paid tiers for larger companies, and then added catalog and checkout features that turned the app into a commerce channel. Each step collected more business data and more business dependency, and the Business Agent is the point where Meta finally sells automation on top of that installed base. The company has been criticized for moving slowly on AI monetization compared with rivals, and the global launch is an answer to that criticism, delivered through the distribution channel only Meta has.
For businesses, the calculus is straightforward and is playing out in markets around the world: the agent costs less than the human, works more hours, and does not complain. The human remains in the loop for the conversations that matter, and Meta is betting that the handoff, when it happens, will be seen as a feature rather than a failure. The launch day was about the small shop with the price list. The coming year will show whether the model scales to the enterprise, and whether the world’s largest messaging platform becomes its largest customer service desk.
