Meta is launching a business AI agent across WhatsApp, Messenger, and Instagram, priced on a token-based model. It's the company's first attempt to generate direct AI revenue outside its advertising machinery. The timing is deliberate: Q1 2026 delivered record sales of $56.3 billion, up 33% year-over-year, yet 98% of that still comes from ads. Meta Business Agent is the opening move in what could become a fundamental shift in how the world's largest social media company makes money.
98% from advertising: why Meta needs to diversify
$META generates almost all of its revenue from targeted advertising - a model that has worked exceptionally well, but is not without weaknesses. The advertising market is cyclical: in a recession, advertisers are the first to cut budgets, and this was pronounced in 2022, when Meta saw its first-ever year-over-year revenue decline. Although AI has helped companies streamline their advertising business, repeated attempts to sell digital and physical products directly to consumers and businesses have not yet taken off.
What Meta Business Agent can do
Unveiled at the Conversations conference in London on 3 June 2026, the tool extends existing corporate messaging with so-called agent capabilities - the assistant can not only reply but also perform actions such as booking appointments or closing sales on behalf of companies. The agent works across WhatsApp, Messenger and Instagram, integrates with hundreds of external systems, Shopify, Zendesk, Shopee, and will offer admins custom metrics and adjustable behavioural limits. The global launch was with free access; Meta has paid tiers in the pipeline.
Pricing model: tokens instead of reports
The monetization method replicates the model established by large AI providers: large enterprise customers will pay for an agent based on the number of tokens consumed, the basic units of text processed by the language model, just like when accessing Microsoft's OpenAI or Azure AI APIs. For Meta, this means revenue tied directly to AI usage, not the number of messages sent as before. Smaller businesses get access via a Business Premium subscription, while larger players on the WhatsApp Business Platform stay at the consumer rate.
Base: one million businesses as a starting point
More than a million businesses already use the older chatbot versions of these tools on WhatsApp and Messenger. The new agent appeals to this group with an expanded set of features - the ability to escalate queries to human staff, suggest products and autonomously close business cases. In total, WhatsApp Business serves over 200 million businesses worldwide, providing a base on which the Meta Token model can scale without having to acquire new customers. Meanwhile, the conversion success rate from the free to the paid version remains a key unknown.
Enterprise Solutions: new internal unit
The product alone is not enough - Meta is building an organizational infrastructure for the enterprise segment in parallel. The company is setting up an Enterprise Solutions unit that will send engineers and product managers directly to clients. At the same time, CTO Andrew Bosworth has launched an internal Agent Transformation Accelerator initiative in which employees move from directly performing tasks to supervising AI agents. This reorganisation is running alongside the redundancy of approximately 8,000 people. Meta is therefore not expanding capacity, but reallocating it towards AI.
CapEx $145 billion: an investment with no guarantee of return
The enterprise agent is not an isolated product, but part of a massive bet on AI infrastructure. Meta has raised its 2026 capex outlook to $125-$145 billion, up from its original range of $115-$135 billion. Record Q1 revenue of $56.3 billion confirms that the ad machine is working, but that's why investors are watching closely to see if new revenue streams can justify the growing spending base. Meanwhile, subscription and AI services are gaining higher valuation multiples in equity markets than advertising revenues, are more predictable and less sensitive to the business cycle, which is a major argument for diversification from a valuation perspective.