Key point
Meta’s consumer AI agent ‘Muse’ topped the U.S. App Store in just six days after its launch, recording over 900,000 downloads
Research firm Citrini estimates that businesses relying on ‘inertia’—customers who don’t switch, don’t cancel, and don’t compare options—will be hit the hardest
About 60% of U.S. subscribers are making at least one paid subscription purchase for a service they don’t even use, wasting an average of $26.79 per month
In a separate survey, only 7% of consumers said they would allow “AI agents’ purchases” that skip prior approval
A substantial portion of the U.S. consumer economy is funded by “people just leaving it alone because it’s annoying.”
Forgotten paid subscriptions, automatically renewing insurance, cash left in accounts that effectively earn no interest, and airline miles and points that eventually disappear.
Research firm **Citrini** said in a new report released on Tuesday that Meta’s AI agent **Muse** views the pivot point of these inertia-based revenue models as no longer a “safe bet,” because agents don’t get bored, don’t get tired, and don’t give up.
The economic scale of “inertia”
Numbers aren’t small. According to a Self Financial survey conducted in March 2026 among 1,272 U.S. adults, 59.9% of respondents had at least one or more paid subscriptions they did not use at all, and on average they were paying $26.79 per month while holding 2.6 “sleeping services.” That’s more than $320 per person per year—a figure that surged from $10.57 per month a year ago.
Subscriptions are only the tip of the iceberg. Citrini says this inertia economy includes everything from insurance re-enrollment (assuming the price is set without comparing quotes again), deposits tied up in low-interest accounts, unused airline credits, and all services that “handle complexity for the customer that they don’t dig into themselves because it’s a hassle.”
Citrini writes that, “the agent removes friction,” and points out that, “corporate value in the trillions of dollars is built on the assumption that human limits will continue.”
Why now?
Related article: Trump renames AI as “superintelligence” in U.S. government documents
The key isn’t the level of the technology, but the “rate of adoption.” According to Bloomberg, Muse surpassed 900,000 downloads within six days of launch to take first place on the U.S. app store. It’s the first agent example to reach a scale like this in the mainstream consumer market.
Citrini concedes that Muse “feels less robust than other solutions,” but sees that as the point—because Meta is targeting ordinary consumers rather than enterprises. Citrini likens it to a “Studio Ghibli moment” for agent-style AI: not a toy for early adopters anymore, but a point where people simply use it in everyday life, even without knowing the name accurately.
How far is it a “scenario”?
This argument is packaged in part like “history-writing.” The February report that Citrini quotes at length describes, in the past tense, a scenario set in 2027 in which real-estate brokerage fees collapse to below 1% and daily AI token consumption by Americans reaches 400,000. Of course, none of it has happened yet.
Citrini doesn’t hide this. “Some things will pan out,” it says, “but much more will be wrong.” In other words, these are scenario assumptions that capture direction—not what’s actually underway.
In the end, speed is determined by “trust.”
The tougher constraint is consumers’ psychology. In a YouGov survey this year commissioned by ACI Worldwide, only 7% of consumers said they would allow “AI agents to proceed all the way to a purchase without prior approval.” Meanwhile, 53% of respondents said they felt uncomfortable with the idea itself.
That said, Citrini’s hypothesis isn’t invalidated. The timetable just changes. To find and cancel long-forgotten paid subscriptions on a user’s behalf doesn’t require the same level of trust as replacing the purchase of an expensive item. As a result, the first layer likely to collapse is the “friction removal” area, while other areas may change far more slowly than any research note suggests.
Reading next: Bitcoin ETFs see biggest “net inflow of $999 million” since October