Meta launches AI agent for WhatsApp Business globally

On June 3rd, Meta activated its artificial intelligence agent for WhatsApp Business globally. This is not a chatbot with predefined responses: according to TechCrunch, the system charges companies based on token consumption, a billing model that until now was more characteristic of language model APIs than mass-market messaging platforms.

That matters for a concrete reason: WhatsApp has over 2 billion active users and is the dominant customer support channel in markets as different as Brazil, India, Spain, or Nigeria. Introducing a token-priced agent into that ecosystem is not a minor move.

What the agent offers and how pricing works

The WhatsApp Business AI agent allows companies to automate entire conversations with customers: answering inquiries, managing orders, offering post-sales support, or guiding purchase processes without human intervention. The difference from earlier solutions, such as the decision-tree chatbots Meta already allowed integrating, is that the system can maintain conversational context and respond to open-ended questions with reasonable coherence.

The economic model, however, is what stands out most. Meta is moving away from flat-rate or per-conversation-initiated pricing and shifting to billing based on processed tokens. For companies with low volumes or short conversations, this could be more economical. For those managing intensive support or longer conversations with much context, costs can become unpredictably high.

This approach is the same one already used by Anthropic, OpenAI, and Google with their APIs, and it has clear technical logic: the actual inference cost of a language model depends directly on tokens processed. But introducing that granularity to a messaging platform aimed at small and medium businesses, many of which lack a technical team that understands what a token is, adds a layer of complexity that can create friction.

Who is most affected by this change

Three distinct business profiles are most exposed:

• SMEs in emerging markets that use WhatsApp as their primary sales and support channel. For them, cost predictability is critical, and the token-based model can feel opaque.
• Ecommerce platforms and marketplaces that already integrated the WhatsApp Business API and must now recalculate their operating margins if adopting the agent.
• Agencies and developers building WhatsApp automation solutions for clients: they will need to update their pricing models and monitoring dashboards.

Conversely, companies with short, well-defined conversations, such as order confirmations, appointment reminders, or shipping notifications, will likely benefit from the new scheme compared to previous flat-rate pricing.

The competitive landscape to keep in mind

Meta is not entering this space alone. Twilio, Zendesk, Intercom, and dozens of specialized platforms have offered WhatsApp automation with third-party language models for years. Meta's advantage is obvious: it controls the channel, the platform's data, and can optimize integration at layers competitors could never access.

What remains unclear is which language model powers the agent and what level of customization it allows. If Meta opens that layer, for instance by allowing custom models or fine-tuning the agent's behavior through system instructions, the product becomes far more interesting for advanced integrators. If it remains a black box, many companies with specific requirements will continue preferring to build on the API with their own models.

---

From our perspective, Meta's move aligns with the direction the industry is taking: conversational agents are shifting from a differentiator to standard infrastructure. What will determine this launch's success is not the technology, but whether Meta can make the cost model understandable to the business profile that uses WhatsApp Business most: the one selling through chat without a CTO.