Muse Aims for Transaction Fees but Revenue Path Unclear

Meta says its new AI agent will pay for itself by skimming a little off the top whenever it saves you money. That sounds simple enough, but the math gets murky once you factor in user trust, partner demands, and the fact that most people still have no idea what an “agent” actually does.

Meta’s Muse has been live for barely a month and it’s already sitting at the top of the App Store with millions of downloads. The company has spent the last few weeks pitching it as something genuinely new: an AI that answers questions and also acts on your behalf, booking travel, buying inventory, managing cash flow, and handling back-office tasks for small businesses. Meta President Dina Powell McCormick told Jim Cramer on “Mad Money” that the company is actively exploring transaction fees and paid subscriptions as ways to turn that popularity into profit. Mark Zuckerberg said much the same at Meta Connect, promising that Muse will remain free for a huge number of tokens with the expectation that over time Meta will profit by taking a small fee from transactions. The logic is tidy on a whiteboard. In practice, it’s tangled up in questions about revenue splits, user consent, and whether a company that makes nearly all of its money from advertising can really pivot to commerce commissions.

McCormick was direct about the transaction model. “Muse right now helps a lot of people save money,” she said. “And so we’re thinking through… the potential of taking a fee when we help you save money.” On its face, this is a performance-based revenue model. If Muse negotiates a cheaper flight, finds a discount on retail inventory, or locates a better rate on software, Meta keeps a slice of those savings. Zuckerberg framed it similarly, telling developers that Muse will be free for a huge number of tokens but will eventually charge a small fee tied to completed transactions.

The issue is that “small fee” and “savings” are both moving targets. Consumers don’t have a clear benchmark for what counts as a saving large enough to justify a surcharge. If Muse saves a small business owner $50 on a supply order but charges $8 for the service, the net benefit is real but thin. If the alternative is spending ten minutes doing the price comparison manually and keeping the full $50, the agent’s value collapses. For the model to work at scale, Muse needs to produce savings that are large enough, frequent enough, and frictionless enough that users don’t feel nickel-and-dimed every time the AI clicks “buy” on their behalf.

The subscription layer adds another variable. McCormick noted that increased engagement could lead to recurring pricing models. “So the more you use it, we may have opportunities, obviously, for subscriptions as well,” she said. But subscription fatigue is already crushing consumer software, and small business owners in particular are flooded with monthly charges for everything from accounting software to email marketing. Layering another subscription on top of Shopify, QuickBooks, Stripe, and Slack may be a harder sell than Meta expects, especially when the free tier is designed to be generous enough to hook users in the first place.

Meta wants to monetize Muse through a combination of transaction fees and, eventually, subscriptions while keeping substantial usage free. The opportunity is largest if the agent can reliably complete valuable consumer purchases or reduce small-business workload, but the model depends on clear attribution, transparent incentives, strong controls, and enough user trust to justify access to sensitive data. Meta’s distribution can accelerate experimentation, but it cannot by itself prove that Muse will become a durable and profitable business.

For Muse to collect transaction fees, it needs deep integration into the places where people actually spend money. Meta has lined up an impressive roster of partners. The agent supports PayPal and connects with retailers including Walmart, Best Buy, The Gap, Sephora, Wayfair, and American Eagle. It has moved into travel through Expedia and grocery through Instacart. For small businesses, Muse hooks into Shopify, Salesforce’s Slack, Intuit QuickBooks, Stripe, Canva, Asana, Notion, Klaviyo, Dropbox, Zoom, and Meta’s own advertising and social accounts.

That breadth is necessary but not sufficient. Every integration represents a potential negotiation over who gets paid what. If a user buys a jacket through Muse at Walmart, it’s unclear whether Meta is earning an affiliate commission, a transaction fee assessed to the user, a revenue share with Walmart, or some combination of the three. Zuckerberg called it a “small fee from transactions,” which suggests the cost is passed to the consumer or skimmed from the retailer’s side. Neither party may be thrilled about that. Retailers already run thin margins on e-commerce, and consumers who discover that the “free” AI agent is actually adding a surcharge to every purchase may abandon the tool quickly.

McCormick emphasized that Muse operates as an active agent. “It’s not just giving you back research or feedback,” she told Cramer. “It’s actually acting on behalf of what I’m saying.” The application can carry out commercial transactions, though it requires explicit approval for sensitive actions such as purchases. That approval step is critical for safety, but it also interrupts the agentic promise. An active agent that pauses for permission on every meaningful action is, functionally, a very smart recommendation engine wrapped in a lot of confirmation dialogs. If users end up doing the final clicking themselves, the justification for a transaction fee weakens.

Meta knows it has a trust deficit, and the Muse launch has been designed with that in mind. The company delayed the rollout by thirty days to ensure the product was encrypted and as bug-free as possible. Muse runs on a dedicated Muse Secure VM that isolates user data. A separate Sentinel agent monitors actions and controls network access, requiring user permission before anything sensitive happens. Meta has even launched a public bug bounty program offering up to $300,000 for valid vulnerability reports, including prompt injection exploits.

Those are serious technical safeguards. The problem is that trust goes beyond technical fixes. According to an Oppenheimer consumer survey cited in recent coverage, only 8% of US respondents said they would trust Meta with their passwords. That compares poorly to other tech platforms and reflects years of privacy scandals and data mishandling that have eroded confidence. The Wall Street Journal put it bluntly: Meta can ill afford any user-privacy hiccups in Muse’s rollout because the risk is high that users will lose trust and flee.

Concrete incidents are already feeding that fear. National Public Radio reported that one user declined iMessage access only to later discover that Muse had synced the data anyway. Meta insists that Muse has no visibility into passwords or payment methods and that credentials are stored securely, but stories like the iMessage sync make those assurances ring hollow for skeptical users. If the small business owners most likely to benefit from Muse refuse to hand over their accounts receivable or payroll passwords, the agent becomes a toy instead of a tool.

Muse will not earn durable revenue merely by becoming the assistant that can do everything; it must become the agent users can safely hold accountable when something goes wrong.

Transaction fees and subscriptions are easy to describe, but the real product is dependable delegated action. A single purchase, message, or data-handling mistake could undermine the access and confidence that the business model requires.

The freemium model Muse is using looks familiar because it mirrors the path taken by countless consumer apps. Give away value to build a user base, then flip the monetization switch once habits are locked in. Zuckerberg has been explicit that Muse will be free for a large number of tokens. BNP Paribas analysts and other market watchers have pointed out that for now Muse appears to be charging partners a small transaction fee, but the long-term revenue prospects remain unclear. One likely outcome, they suggest, is that advertising eventually subsidizes the product.

That would be a letdown for anyone hoping Meta had found a genuine alternative to its ad model, but it would also be the path of least resistance. Meta has 3.6 billion daily active users across its family of apps. It knows how to sell ads. If transaction fees fail to materialize at scale, the company can always monetize Muse through sponsored suggestions, promoted merchant listings, or data enrichment for its existing ad targeting. McCormick essentially admitted this when she said that artificial intelligence products coming out of Meta’s lab serve to make its core business stronger. Muse might simply serve as a very expensive feature designed to keep people inside the Meta ecosystem, with transaction revenue as a secondary benefit.

The subscription possibility faces its own headwinds. If power users are the only ones willing to pay, Muse becomes a niche productivity tool rather than a mass-market agent. And if the free tier is robust enough to handle basic tasks, casual users have little reason to upgrade. Meta is walking a tightrope between giving away too much and charging for something people can already get from ChatGPT, Copilot, or a simple spreadsheet.

Meta officially launched Muse for small businesses during the same “Mad Money” broadcast, and it’s easy to see why this segment is a priority. The company claims 200 million small business owners use its platforms. These are exactly the kind of time-starved operators who might benefit from automated back-office work. Muse promises to manage cash flow, inventory, website design, and direct customer communications by connecting to tools the business already uses.

The integration list is genuinely useful. A shop owner can link Shopify for storefront data, QuickBooks for bookkeeping, Stripe for payments, Slack for team chat, and Meta’s own ad accounts for marketing. Muse can pull data from these sources, draft materials, and prepare actions for review. Meta insists that nothing publishes, sends, or spends without approval, which addresses the legitimate fear of an AI accidentally ordering 10,000 units of the wrong product.

Yet small businesses are notoriously price sensitive. They adopt technology when it drives immediate, measurable profit. An AI agent that requires careful supervision before every purchase or post cuts some corners but leaves the heavy judgment with the owner. If the owner still has to review every action, the time savings shrink. And if Muse tacks on fees for inventory purchases or customer communications, the cost structure of a lean operation takes a hit. The value proposition has to be airtight: either Muse saves so much time that a subscription is obviously worth it, or it negotiates discounts so steep that transaction fees are invisible. Neither outcome is guaranteed.

There’s no question that Muse has momentum. It reached number one on the App Store and has logged millions of downloads. Those are enviable launch metrics. They are also vanity metrics if the users refuse to connect their wallets.

Converting a free downloader into a paying commerce user is one of the hardest jumps in tech. Meta is asking people to do three difficult things in sequence. First, grant an AI agent access to sensitive personal and financial accounts. Second, allow that agent to transact on their behalf. Third, pay a fee or subscription for the privilege. Each step filters out a huge chunk of the user base.

McCormick highlighted Meta’s reach of 3.6 billion daily active platform users as a built-in advantage. Reach is not revenue. Most of those users are on Instagram and Facebook to socialize, not to hand their credit cards to an AI. The subset willing to use an agentic shopping and business tool is far smaller than the total audience. Meta needs to identify and convert that subset quickly, because Wall Street has little patience for a monetization story that still relies on phrases like “thinking through” and “potential.”

Meta deserves credit for shipping a bold product and for talking openly about monetization rather than hiding behind vague promises about future innovation. The transaction fee model is at least a coherent theory. If Muse genuinely saves users money, those users might happily share a slice of the savings. If the small business tools work as advertised, Muse could become an indispensable layer between merchants and the software they already pay for.

But coherence is not cash flow. Meta hasn’t disclosed what percentage it intends to charge, when the freemium limits kick in, or how revenue will be split with Walmart, Shopify, Expedia, and the rest. It hasn’t explained how it will overcome a trust rating that puts it behind most of its rivals. And it hasn’t proven that users want an active agent enough to pay for one, especially when every sensitive action still requires manual approval.

Right now, Muse is an experiment funded by one of the largest advertising businesses on Earth. The transaction fees and subscriptions that McCormick and Zuckerberg have floated are possibilities, not plans. Until Meta shows real revenue numbers and real user retention among paying customers, the safest assumption is that Muse is a very impressive, very expensive bet. Whether it becomes a profitable new business line or just another feature keeping eyeballs on Meta’s apps is a question the company still needs to answer.