Since its public debut on September 8, Meta’s Muse AI assistant has been downloaded by over 10 million users across the United States and Canada. The platform’s early metrics reveal that roughly one-quarter of those who set up the agent open it on any given day, a retention rate that Mark Zuckerberg described as “more than twice the level that Threads or comparable AI apps enjoyed at launch.”
Beyond raw adoption, Muse’s engagement pattern points to a clear product-market fit: users are turning to the agent for everyday tasks such as email handling, calendar coordination, and, most prominently, online shopping. Data shared during Meta’s third-quarter earnings call show that over 25% of active users request assistance with purchases, and the average transaction facilitated by Muse exceeds $200. This commercial activity has already forced the company to work with payment processor Stripe to raise credit limits during the first week of operation.
Shopping as the primary revenue engine
The shopping use case is not accidental. Meta intends to monetize Muse by taking a modest percentage of each sale the AI helps complete. While the service remains free for consumers—except when they request more tokens or advanced model features—the company expects a growing “commerce-generated” revenue stream to supplement its traditional advertising earnings, which amounted to $63.1 billion in Q3. Zuckerberg noted that the average revenue per American family from existing Meta apps runs in the “several hundreds of dollars annually” range, and he anticipates Muse will lift that figure as purchase volumes expand.
To cement the shopping pipeline, Meta’s finance chief Susan Li revealed collaborations with major retail platforms such as Shopify and Walmart to craft a “Personal Agent Protocol.” This standard would allow a consumer’s Muse instance to communicate directly with a retailer’s AI counterpart, effectively bypassing traditional web navigation and granting Meta a pivotal role in the transaction flow.
Scaling compute and the economics of AI agents
Running a sophisticated, multi-step AI assistant is currently compute-intensive. Meta invested roughly $35 billion in capital expenditures during the quarter, largely to fuel AI-focused servers, data-center upgrades, and networking gear. The company projects total capex for the full year between $135 billion and $145 billion, more than doubling the previous year’s outlay. Consequently, Muse’s operating cost is a central concern.
Zuckerberg emphasized that Meta is targeting a ten-fold improvement in compute efficiency over the next twelve months. Achieving this milestone would transform Muse from a costly experimental service into a profitable line item, especially if the “compute efficiency continues improving quickly for another few years,” as he put it. The ultimate ambition is to support more than 1 billion Muse users globally, a scale that demands both hardware advances and software optimizations.
Privacy safeguards and retention trends
Amid rapid growth, Meta is also rolling out a “Muse Confidential VM” to address lingering privacy concerns that have dogged the platform. The move follows a recent legal judgment ordering Meta to pay a multibillion-dollar penalty for alleged data-privacy misrepresentations.
Retention data from third-party analytics shows a noticeable dip two weeks after initial download—a pattern shared by many AI chatbots. While the early drop is not unique, it underscores the need for ongoing feature enrichment to keep users engaged beyond the novelty phase.
Overall, Muse represents Meta’s most ambitious attempt to translate its massive AI infrastructure spend into a direct consumer-facing revenue source. By leveraging high-value shopping interactions, pursuing aggressive compute-cost reductions, and laying groundwork for a standardized agent-to-agent commerce protocol, the company hopes to turn the AI agent from a curiosity into a cornerstone of its financial future.
