We’ve been using Meta Muse to schedule meetings, manage inboxes, audit websites, and handle the other bits of agency work we already give AI agents. It’s even answered emails from other Meta Muse agents that found our articles about Muse and Shopify and asked us to test their sites. The agents are networking now. We assume someone will be organizing their holiday party soon.
There’s one thing we haven’t done with Muse: made a purchase.
That gap matters because Meta has laid out an ambitious business model for Muse. At Meta Connect on September 23, Mark Zuckerberg said the agent would be free for many users, with Meta expecting to make money over time by taking a small fee from transactions Muse completes. It’s an appealing plan. It also echoes an old bet from Amazon: give people a useful assistant for free, make it part of daily life, then earn money when it helps them shop.
That sounds simple until you ask how much of daily life ends at checkout.
A Free Agent With a Commerce-Shaped Business Model
Muse can do a lot more than answer questions. Meta says it can send emails, book travel, fill out forms, negotiate on a user’s behalf, and make purchases, with the user’s approval for sensitive actions. It can keep working after you close the app. Meta describes it as free for most of what people need, while offering paid subscriptions for people who want more use.
So “free” has some boundaries. Still, the strategy is clear: let a large number of people use Muse without paying a monthly fee, and make money from some of the transactions it handles. Zuckerberg has publicly described transaction fees as part of that plan. The company hasn’t published a fee schedule or shown how much revenue the model is generating.
That distinction matters. Meta has named the direction, but the size and mechanics of the revenue stream remain open questions. A fee on a completed purchase might pay for the work leading up to that purchase. It might not pay for the many tasks that never become purchases.
Amazon Made a Similar Bet With Echo
When Amazon introduced Echo, it put a voice assistant in the home and made Alexa available without a separate monthly charge. The smart speaker was an affordable way to reach Alexa, and Alexa could connect users to Amazon’s shopping, media, and other services. The larger idea was that if asking Alexa for help became a habit, asking Alexa to buy something might become one too.
That strategy didn’t require Echo hardware to be a huge standalone profit center. A speaker could be a doorway into Amazon’s commercial services. If Alexa helped generate additional purchases, Amazon had another reason to put the device in more homes.
There’s a useful parallel here, but it isn’t a perfect match. Muse is an app and online agent, not a speaker you buy for the kitchen counter. Meta’s costs and routes to revenue are different. The shared wager is about attention and habit: offer a helpful assistant at little or no direct cost, then find ways to earn from the activity it enables.
Alexa Became Useful. Shopping Was Another Matter.
Alexa found plenty of reasons to stick around. People used it for timers, weather, music, smart-home controls, questions, and jokes. Those are good jobs for a voice assistant. They’re also jobs that don’t necessarily send anyone to checkout.
That’s the key distinction for Meta. Frequent use is not the same as revenue-generating use. An assistant can become part of someone’s morning routine and still spend much of its time doing things that don’t earn a transaction fee.
It would be too neat to call Alexa a commercial failure, though. In its second-quarter 2026 results, Amazon said U.S. customers who use Alexa for Shopping spend more than 40% more per order on average than customers who don’t. That’s a significant claim about shopping behavior. It doesn’t tell us whether Alexa’s shopping revenue covers the costs of the service, or how much of the difference comes from Alexa itself versus the habits of people who already shop more.
The more careful lesson is that daily usefulness doesn’t automatically lead to shopping. Amazon’s recent results suggest Alexa can be associated with valuable orders. They don’t erase the basic business question: how much commerce does an assistant generate, and how much does it cost to support all the other things people ask it to do?
We’ve looked at what Alexa+’s expansion could mean for brand sellers. The broader story keeps shifting, but the connection between a helpful feature and a profitable one still needs to be earned.
Muse Has the Same Basic Math to Solve
Muse can research products, answer questions, compare services, make reservations, fill out forms, write and summarize, and communicate with businesses. Some of those tasks may end with a transaction. Plenty won’t.
Suppose you ask Muse to summarize a long email thread, find a meeting time, or help plan a week of dinners. It may save you time and make you more likely to use it again. But unless those tasks lead to purchases that Meta can earn a fee from, they don’t directly support a transaction-based revenue model.
Meta incurs operating costs across the whole service, not only at checkout. A business model based on transactions therefore has to account for the work that happens before a purchase, and for the work that never gets close to one.
Agent Tasks Cost More Than Setting a Timer
Asking Alexa to set a timer is a relatively simple request. Asking an agent to find the best carry-on for a two-week trip could involve interpreting your preferences, researching products, visiting multiple websites, comparing options, answering follow-up questions, and getting your approval before buying. That’s a much longer sequence of work.
Some agent tasks may require several model calls and outside services. Others will be quick. The point isn’t that every Muse request is expensive. It’s that the cost of supporting an agent includes much more than the final purchase screen.
A useful way to frame the business is:
Revenue from transactions ÷ cost of all agent activity
If the numerator grows more slowly than the denominator, a free tier needs another source of support, or tighter limits on how much people can use it.
For a simple hypothetical, imagine 100 agent tasks lead to five purchases. Those five transactions have to help pay for the other 95 tasks, as well as the work involved in the shopping tasks themselves. That isn’t a Muse usage statistic or a forecast. It’s just the arithmetic the business model has to face.
Muse Can Join the Shopping Process Earlier
Alexa often entered the shopping process when a customer already knew what they wanted: “Alexa, order more paper towels.” That can be convenient, especially for a familiar household item. It leaves less room for product discovery because the decision is mostly made.
Muse can enter earlier. “I need a good carry-on for a two-week trip to Italy” gives it room to ask about budget, size, airline restrictions, and the features you care about. It can research the category, compare products, recommend one, and potentially complete checkout.
That earlier entry point gives Meta more chances to influence a purchase than a reorder command does. It may also make the assistant useful even when the shopper isn’t ready to buy. A recommendation could lead to a sale now, a saved option for later, or a long conversation that ends with “thanks, I’ll think about it.” Opportunity isn’t the same as conversion.
Meta’s own executives have talked about shopping as a discovery process, including product inspiration and comparisons. That’s a wider role than taking a voice command for a repeat purchase. For sellers, it also puts product information and recommendation quality closer to the start of the sale. Our coverage of the legal fight over AI agents shopping on Amazon gets at another part of the change: who gets to act as the shopper’s representative, and under what rules.
Recommendations Bring a Trust Question
The more influence an agent has over product discovery, the more important its incentives become. What happens if Retailer A pays Meta a fee and Retailer B doesn’t? What if Product A generates twice the revenue of Product B?
Those questions don’t prove that compensation will affect a recommendation. They show why users need to understand whether it could. A shopping assistant that earns money from purchases sits inside a familiar conflict from affiliate marketing. But the setting is different. People may treat Muse less like a product listing and more like a personal assistant that knows their preferences.
That raises practical questions for Meta. Will Muse disclose when a recommendation earns the company money? Can users tell why one product appeared ahead of another? Does a fee change how products are ranked, or which retailers the agent searches?
Clear answers matter even if Meta says payment has no effect on recommendations. Trust depends on what users can see and understand, not only on what the company says happens behind the scenes. For an agent that can both recommend and buy, disclosure and neutrality are part of the product experience, not fine print to leave for later.
More Usefulness Can Mean More Cost
The central question isn’t whether Muse can generate commerce. It can research products and, with the right setup, complete purchases. The harder question is whether that commerce can support the rest of the service.
Return to the hypothetical: five purchases generated from 100 tasks. Those five have to help cover the other 95, many of which may have made Muse valuable to the user but earned Meta nothing directly. The example is deliberately simple. The real economics would depend on usage levels, compute costs, transaction values, fees, and the number of purchases that Muse can influence.
There’s an odd tension here. The more useful Muse becomes, the more reasons people have to give it work. Some of that work will involve shopping. Much of it may not. More use can help the product become a habit while also increasing the amount of activity that transaction fees need to support.
That’s the Echo problem in a new form. A helpful assistant can win a place in someone’s routine without making every useful interaction pay for itself.
Shopping Commissions Aren’t Meta’s Only Option
Meta has more than one possible way to support Muse. It already offers paid subscriptions for heavier use, and it can set usage limits on the free tier. It could also earn transaction fees from merchants, use affiliate-style commissions, sell advertising or sponsored product placements, or offer tools and services to merchants.
Those are possible revenue routes, not a list of confirmed Muse features. Zuckerberg has discussed fees from transactions; Meta has not publicly established all the other options as part of Muse’s business model. Each would come with different trade-offs. Paid tiers could help cover expensive users. Merchant services could bring in revenue without waiting for a consumer purchase. Sponsored placements could raise money, but would make the recommendation and disclosure questions even sharper.
A mix of revenue could change the comparison with Alexa. Meta wouldn’t have to cover every free task with shopping commissions alone. But multiple revenue sources don’t make the trust question disappear. If an agent is paid to recommend or connect people with sellers, users need to know where that money enters the process.
From the Echo Problem to the Muse Problem
Amazon’s experience with Alexa shows that becoming indispensable isn’t the same as becoming profitable. Alexa became useful in millions of homes. That usefulness didn’t mean every timer, song, or weather check turned into a purchase. Amazon’s newer results point to shopping value among Alexa users, but the cost and revenue picture for the full assistant remains a different question.
Muse is making a broader version of the same wager. Meta wants to put an agent between people and online services, help with work and everyday tasks, and earn money from some of the transactions that flow through it. Its earlier role in product research gives it more opportunities to influence shopping. Its usefulness outside shopping gives it more activity to pay for.
That could become a very valuable business. The test is whether the transactions and other revenue can support the agent people use between purchases, too.
What happens when people love your free assistant, but mostly use it for things that don’t make you any money?
Tell Parker-Lambert about your brand and ask about a custom quote for your project.