Meta Launches Enterprise AI Agents Across WhatsApp, Messenger, and Instagram
Claire Weston
Meta on Wednesday launched Meta Business Agent for businesses worldwide, spanning WhatsApp, Messenger and Instagram, with over one million businesses already on early versions. The move marks Meta's first systematic attempt to turn tens of billions in AI spending into revenue — and a critical step in WhatsApp's shift from messaging tool to enterprise software platform.
What can this AI agent actually do?
Business Agent can be customized to a company's brand voice. It handles customer queries, screens sales leads, processes payments, manages bookings and places orders autonomously.
Complex issues are automatically escalated to human agents. Meta is also testing an "overnight briefing" feature that summarizes after-hours conversations and surfaces insights.
This means → it is no longer a rules-based chatbot but an agent that can "take action" — the loop from reply to transaction closes inside a single chat window.
How will Meta charge for it?
The product is free for now. In the coming months it will shift to paid subscriptions, tiered by business size. Small and mid-sized businesses will use a Meta One subscription plan.
Large enterprises will pay per token consumed, not a flat fee — a pay-as-you-go model consistent with Meta's existing per-message pricing for business chats.
Meta is simultaneously launching a Business Agent Platform that lets large enterprises build custom agents integrated with hundreds of third-party systems including Shopify, Zendesk and Shopee.
Where does WhatsApp monetization stand today?
WhatsApp's paid-messaging business hit an annualized revenue run rate above $2 billion in Q4 2025 — still a tiny fraction of Meta's roughly $200 billion in total revenue last year.
WhatsApp has over 3 billion active users, yet its long-standing privacy commitments have kept advertising off the platform.
In plain terms = WhatsApp has enormous reach but has never been a serious money-maker. The AI Agent launch opens an entirely new monetization channel for those 3 billion users.
What privacy trade-off is Meta making?
User-to-user messages remain end-to-end encrypted. Conversations with AI agents, however, are not encrypted — both Meta and the business can read them, and the data "may" be used to train AI models.
WhatsApp product head Alice Newton-Rex confirmed this policy in an interview.
This reflects a structural tension: the platform needs conversation data to improve agent performance, but that clashes directly with WhatsApp's core privacy brand. Put simply = making the AI agent smarter requires feeding it chat data — exactly the thing WhatsApp has long refused to touch.
Who are the competitors, and what is the key test?
Amazon and Microsoft have already shipped enterprise-grade agent products. OpenAI and Anthropic are accelerating their own enterprise push.
Meta's differentiator is WhatsApp's 3-billion-user messaging gateway — a business's customers are already there, with no need to migrate them to a new platform.
This means → the real test is not technology. It is whether Meta can convert enterprise willingness to pay into sustainable, high-scale software revenue.
As our models get better, your agent will take on more and more work and eventually help you run your whole business.
Mark Zuckerberg
CEO, Meta
(June 2026, video address at WhatsApp Conversations conference, London)
Content is for reference only, not financial advice.