Personal agents like Meta’s Muse could fuel AI’s growth — and challenge Amazon and Apple
Hopes are high on Wall Street that autonomous AI agents will drive a new revenue phase for the AI build-out. But these bots are still a show-me story while concerns mount about online privacy and cybersecurity. Meta Platforms’ Muse epitomizes the excitement with the app off to a fast start. Data from Sensor Tower provided to CNBC shows that while it took Anthropic’s Claude 670 days to hit 9 million U.S. downloads, it took Muse just 36 days to hit the same number. Despite a recent pullback, Meta’s stock has gained 5.7% since the launch as analysts pencil in billions of additional revenue. Meta CEO Mark Zuckerberg highlighted Muse on Wednesday’s earnings call, calling it “the centerpiece of our vision for what we’re building.” He said the bot’s adoption by users was so fast that the company had to coordinate with payment processor Stripe to raise credit limits. Muse is potentially just the vanguard for a new age of personal AI agents. Apps like OpenAI’s Dots and SpaceXAI’s Grok Bot could also take off. What’s more, they’re all supposed to get smarter as they learn about your lifestyle habits and personal preferences. The stakes for these personal bots are high. As open-source and freemium AI products proliferate, they could erode pricing for enterprise-scale AI and token pricing for frontier labs, endangering returns on invested capital. Personal agents could generate new revenue streams and another large source of token demand. “Agentic AI tools are what is driving this next era of AI infrastructure build-out because they are driving usage in a way that expands the amount of AI compute that’s necessary,” said Bob O’Donnell, president of Technalysis Research. “They’re doing tasks that are being repeated over and over, and they’re driving more and more demand for it.” Meta and others selling personal agents also aim to gain e-commerce power, capturing more sales and fees from online sellers and exchanges. The goal: controlling e-commerce through robot operators that can act as brokers, shoppers and task managers. “Muse is reshaping how we surf, search, compare, and transact across the Internet, saving users time, money, and improving productivity as it becomes that always on, daily habit,” Citi analysts said in a recent note. Muse could bring in $27 billion of revenue for Meta over the next several years through subscription and transaction revenue, Citi estimates. Personal agents start a new AI knife fight While the apps show promise for online shopping, they’re a long way from ubiquity. While they may be welcome by small businesses and mom-and-pop sellers, they also pose threats to large e-commerce sites and proprietary application systems. Amazon.com has taken steps to limit the access of agents to their online stores and the user-oriented recommendation systems that generate additional sales. “We think it’s fairly straightforward that third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions,” an Amazon spokesperson told CNBC in September when the company blocked Meta’s Muse. Wall Street also sees a threat to Apple ‘s App Store. “It is not a stretch to imagine that AI apps will route outside of in-app purchases to avoid paying Apple’s toll,” Craig Moffett at MoffettNathanson wrote in a Tuesday note. Agents could also leave application interfaces behind entirely, by being routed through text-messaging services, as Spear Street Technology’s Instinct does. “Users interact with Instinct purely via text messages, WhatsApp and phone calls without a traditional app or web dashboard. Apple has no way to monetize this digital interaction,” Moffett said. The trust factor Another major hurdle is privacy, data security, and protection from AI hacking. Since personal agents act like AI butlers — getting to know your personal habits, shopping preferences, and details of your private life — they need vast access to your personal data. “There are a huge range of risks,” said Calli Schroeder, senior counsel at Electronic Privacy Information Center, an advocacy group. “You could have the AI agent collecting its own information and also combining what it learns with information it buys from data brokers. So it can get extremely, extremely specific about you in ways that I don’t think users are aware of.” Some potential privacy breaches may fall under product safety laws that attorneys general can prosecute under UDAP statutes , Schroeder said, acknowledging the possibility of broader AI legislation coming out of the next Congress. Meta says it built a dedicated cloud-based “virtual machine” to run Muse, including “state of the art privacy and security,” as Zuckerberg put it Wednesday. Still, there may be holes. Researchers recently discovered some vulnerabilities in Muse; security firm Claroc found that user-shared Muse pages could be discovered easily through searching domain names, revealing “people’s personal files and companies’ internal documents.” “If you had asked us beforehand, we would have guessed that gathering these would take real effort. It took one query,” Claroc researchers wrote. A Meta spokesperson told CNBC that the company has “paused this functionality while we review this report and consider changes to how these links can be discovered.” Meta’s data protection practices, more broadly, have generated large fines in legal cases. The company was ordered by a New Mexico judge to pay $25 billion in penalties on Wednesday over some of its data practices. The company said it would appeal the order. There’s no guarantee, moreover, that personal agents will add economic value — if they just take from one source and shift it to another. And until they start generating material revenue growth, they aren’t likely to ease the market’s rising anxiety about AI capex and debt. Wall Street remains bullish, though, viewing Muse in particular as a potentially new killer app. Analysts expect Muse and related revenue to help fuel revenue growth for Meta from $255 billion this year to $430 billion in 2029, according to consensus forecasts. Deutsche Bank’s Benjamin Black said Muse has put Meta “back in the frontier discussion,” according to a recent note, maintaining a buy on the stock and $820 target. For now, markets appear more concerned about rising AI capex, debt levels and equity issuance — reasons why Meta’s stock slumped 6% Wednesday even after a solid earnings report.
About the Author
