Meta is no longer treating artificial intelligence as just another feature layered onto Facebook, Instagram or WhatsApp.
The company is building something far more ambitious around Muse, its new consumer AI agent — and Mark Zuckerberg has now revealed the business model that could eventually turn the rapidly growing assistant into a major source of revenue.
The idea is remarkably simple: keep Muse free for a huge amount of usage, let the AI do things for people, and then take a small fee when those actions result in transactions.
That means Meta may not need to sell Muse as a traditional subscription product to make serious money from it.
Instead, the company is positioning Muse as a digital middleman between consumers and businesses.
At Meta Connect on Wednesday, Zuckerberg said the company expected Muse to generate revenue over time by taking a small fee from transactions completed through the AI assistant. Meta’s chief AI officer Alexandr Wang subsequently outlined a growing network of retail, travel, payments and productivity integrations designed to give Muse the ability to act rather than merely answer questions.
The change is important because it moves the AI monetization debate into a new territory.
For much of the generative-AI boom, the standard model has been subscriptions. Consumers pay a monthly fee for additional usage or more advanced models, while companies sell AI capabilities to businesses through enterprise contracts.
Meta is proposing something different.
Make the agent available at low or no upfront cost, build a gigantic user base, then monetize the economic activity taking place through the agent.
In theory, the more useful Muse becomes, the more opportunities Meta has to earn.
Imagine a user asking Muse to find a cheaper flight.
Muse searches available options, selects an itinerary, interacts with the travel provider and completes the booking.
Or imagine a consumer asking Muse to find a replacement for a household product. The agent searches retailers, compares products and prices, chooses an option and processes the purchase.
Under Meta’s model, the AI itself becomes part assistant, part shopping platform and part transaction broker.
And Meta is already building the infrastructure required for that vision.
The company announced connections with PayPal as well as retailers including Walmart, Best Buy, Gap, Sephora, Wayfair and American Eagle. Expedia is being integrated for travel, while Instacart is being added for groceries. Meta has also announced integrations with Shopify and Stripe, giving Muse access to a much larger ecosystem of merchants and payment infrastructure.
That list matters because an AI agent becomes much more valuable when it can actually complete an action.
A chatbot can tell someone where to buy a product.
An agent can buy it.
A chatbot can explain how to book a trip.
An agent can potentially do the booking.
That difference is at the heart of Meta’s strategy.
Muse is being developed as an agent that can operate across services and, with user permission, perform tasks in the background. Meta has also announced computer-use capabilities that allow Muse to interact with applications on a Mac, opening the possibility that the assistant could manage files, software and other digital workflows without requiring the user to remain at the keyboard.
That makes the transaction-fee model potentially broader than shopping alone.
A future Muse user could ask the assistant to negotiate a bill, book travel, purchase goods, arrange appointments or complete other tasks where money changes hands.
The more transactions Muse touches, the more potential revenue Meta can extract from a small take rate.
The challenge is scale.
A tiny fee only becomes meaningful when the volume of transactions is enormous.
Meta therefore has a strong incentive to maximize usage first and worry about monetization later.
That is exactly what Zuckerberg appears to be attempting.
Muse launched on September 8 and has gained traction unusually quickly. Yahoo Finance reported that the app had reached the top of both Apple’s App Store and Google Play Store, while Sensor Tower data showed roughly 560,000 daily active users after only 11 days. Other reporting has since cited more than 2.5 million downloads within its first two weeks.
Those early numbers do not prove long-term success, but they help explain why Meta is moving aggressively.
The company has already begun surrounding Muse with a much larger ecosystem.
Muse can connect with email and calendars, while Meta is developing an email address for the agent itself. Users are expected to be able to put Muse into email threads or forward messages for the assistant to handle.
Meta is also bringing Muse to its smart glasses, meaning users could eventually interact with the agent simply by speaking while walking around.
That creates a potentially powerful distribution advantage.
Instead of opening an app whenever they need assistance, users could interact with Muse from whatever device they are already using.
The strategy also links directly with Meta’s hardware ambitions.
At Connect, the company introduced new Ray-Ban Meta glasses, camera-free audio glasses, Meta VR Glasses and the Muse Charm — a compact device designed to provide direct access to Muse without relying on a smartphone or glasses.
The more places Muse appears, the less it resembles a standalone application.
It starts looking like a platform.
And platforms are where the economics become interesting.
Meta has spent enormous amounts of money building AI infrastructure, hiring researchers and developing its models. For investors, the long-term question has increasingly shifted from whether Meta can build advanced AI to whether those investments can generate returns.
The transaction model gives the company one possible answer.
Instead of asking users to pay $20 or $100 every month, Meta could allow large numbers of consumers to use Muse for free and monetize the actions the assistant performs.
Meta already has a subscription tier. The basic Muse experience is free, while heavier users can pay $20 or $100 per month, according to Yahoo Finance reporting. But Zuckerberg’s comments indicate that transaction fees are intended to become a separate and potentially larger monetization path.
There is an important economic advantage to this approach.
A subscription puts a ceiling on how much revenue Meta can receive from a particular customer each month.
A transaction model does not have the same limitation.
A highly active user could generate dozens or hundreds of economically meaningful interactions.
More importantly, Meta does not necessarily have to charge the user directly.
That could reduce friction and make the service easier to adopt.
But it also introduces difficult questions.
How large will Meta’s fee be?
Who ultimately pays it?
Will retailers absorb the cost, pass it to consumers or negotiate lower fees in return for additional sales?
And how will Meta ensure that Muse’s recommendations remain useful rather than becoming influenced by which merchants are willing to pay?
Those questions could become increasingly important if Muse grows into a major shopping gateway.
There is also a trust problem.
An AI assistant that merely answers questions has limited access to a user’s financial life.
An agent that can make purchases, send emails, interact with websites and operate software has far more power.
That means Meta has to convince consumers that Muse will respect instructions, protect sensitive information and make decisions transparently.
Reuters has separately reported that Meta has been testing a “human concierge” system in which contractors handle certain phone calls made through Muse, raising questions inside the company about privacy and the handling of sensitive information. Meta has said the testing is designed to improve safety and privacy before broader release.
Those concerns are not minor.
For transaction-based AI to work at scale, users have to trust the system enough to let it spend their money.
The commercial opportunity could be enormous, but so is the responsibility.
Meta’s gamble is therefore bigger than building another AI assistant.
It is attempting to change how people interact with the internet.
Today, humans search, compare, click, fill out forms and complete transactions.
Tomorrow, Meta wants Muse to handle more of those steps.
That would shift value from the screen interface to the AI agent performing the task.
For Meta, the attraction is obvious.
The company already has one of the world's largest consumer platforms. If it can connect that audience to an AI assistant that can perform real-world transactions, a tiny percentage taken from enormous economic activity could eventually become a substantial business.
The critical variable will be adoption.
Muse has to remain useful enough that people trust it with increasingly important tasks. Businesses have to participate. Payment systems have to work smoothly. Regulators will inevitably pay attention as AI agents begin handling more commercial activity.
And competitors are pursuing similar ideas.
OpenAI, Google, Microsoft, Anthropic and others are all developing increasingly capable AI agents.
Meta therefore has something to prove.
Its strategy is not simply to build a smarter chatbot.
It wants to build an AI that sits between people and the businesses they interact with — and quietly collects a small fee every time the digital world gets something done.
The biggest question for investors is no longer whether Muse can make conversation.
It is whether Muse can become the checkout counter for the AI era.