NextFin News - Meta's new personal AI agent, Muse, has done something Wall Street did not expect: it turned the stock market's favorite business model into its biggest risk. Within two weeks of launching on Sept. 8, Muse climbed to the top of Apple's U.S. App Store, displacing OpenAI's ChatGPT, and sent the S&P 500 Financials Index down as much as 2.4% on Tuesday to its lowest level since July. The selloff spread far beyond banks. Insurers, telecom carriers, streaming services, online travel agencies, and even a gym chain came under pressure as investors priced in a simple, unsettling idea - that the "consumer inertia" many of these companies rely on is about to be automated away.
The move is not just about one app. It is a bet that when software agents can compare prices, switch providers, negotiate bills, and book travel on a consumer's behalf, the friction that has protected margins in banking, insurance, and subscription services for decades starts to disappear. The question now is whether the market is right to treat this as a structural regime change - or whether it is once again overreacting to a product that, for now, is still more curiosity than habit.
The Selloff: What Got Hit and Why
The damage was broad and, in a few cases, severe. The S&P 500 Financials Index closed nearly 2% lower on Tuesday, trailing a broader market that was roughly flat. JPMorgan Chase and Wells Fargo each fell more than 3%, Morgan Stanley declined 2.9%, and insurer Allstate dropped 5.5%. Planet Fitness, whose business depends on members paying monthly fees they rarely use, fell as much as 11%. Booking Holdings slid 3.9% and Expedia 3.7%. In Europe, telecom operators Orange and BT Group each dropped about 4%.
Goldman Sachs's trading desk gave the selloff its name. In a note to clients, the desk argued that as AI assistants like Muse and Instinct improve at price comparison, trip booking, and handling customer-service interactions, "industries that rely on recurring bills, negotiable pricing and add-ons could come under pressure." Its basket of at-risk "consumer inertia" stocks reads like a map of the modern subscription economy: AT&T and T-Mobile in telecom, Allstate and Progressive in insurance, Netflix and Paramount Skydance in streaming, and Expedia and Booking in travel.
Consumer inertia, in plain terms, is the tendency to keep buying something out of habit even when a better alternative exists. It is why people stay with a bank that charges them fees, an insurer that raises their premium, or a phone plan they no longer need. It is also why those companies can afford to spend heavily on acquiring customers once and then profit from them for years. The market's fear is that an AI agent removes the habit from the equation.
"Right now it's more of a curiosity, but I think two years from now we're all going to have agents," said Rhys Williams, chief strategist at Wayve Capital Management, who called the new tools "no doubt a negative for those kinds of companies."
The selloff echoed an earlier panic this year, when software-as-a-service stocks melted down after Anthropic launched agentic tools such as Claude Cowork. But this round differs in one respect: the agent is no longer a workplace tool aimed at knowledge workers. It is a consumer product backed by the distribution of one of the world's largest app ecosystems.
The Mechanism: Why Friction Is a Profit Center
To understand why banks and insurers are exposed, it helps to ask what their profits actually rest on. A large share of retail financial-services revenue does not come from winning a fresh price competition every month. It comes from customers who do not move. A bank earns fees from overdrafts, inactivity, and cross-selling to customers who find switching accounts tedious. An insurer profits when a policyholder renews automatically rather than re-shopping every year. A telecom carrier makes money when a customer stays past the promotional rate. A streaming service counts on subscribers who forget to cancel.
Each of these revenue streams depends on a form of friction: the time, effort, and annoyance required to change providers. Economists call it switching cost; managers call it a moat. An AI agent attacks the moat directly by absorbing the friction on the consumer's side. If a user can tell their agent, "Find me a cheaper phone plan with the same coverage and switch me," the carrier's retention advantage shrinks. If the agent can say, "My car insurance just renewed - shop it," the insurer's pricing power weakens. If the agent can book the same hotel for less without the customer ever opening a browser tab, the travel platform's take rate comes under pressure.
Citrini Research, the firm whose bearish February report dragged down delivery, payments, and software stocks earlier this year, put the point bluntly this week: "Tactics that worked when consumer behavior was dictated by human psychology will fall by the wayside." The firm suggested that health insurers, in particular, could see their margins questioned once consumers stop tolerating hours on the phone waiting for coverage approvals.
There is a second, less obvious channel at work - the platform layer. Analysts Mandeep Singh and William Tong argued that personal AI agents could shift customers away from established online platforms, with Muse and Instinct acting as "toll collectors" that generate revenue from transactions flowing through AI apps. In this version of the future, the agent - not the bank's app, the airline's website, or the insurer's portal - becomes the first point of contact. The company that owns the customer relationship captures the economics; everyone downstream becomes a commodity supplier.
That is why the market reaction extended to wealth managers and brokerages as well. Charles Schwab fell more than 5% on Tuesday. If an agent becomes the default interface for financial information and routine transactions, the customer relationship that brokerages have spent decades building could be disintermediated - or, at minimum, forced to compete harder on fees.
Why This Time Could Be Different
Skeptics have heard this story before. Personal AI agents have been "kicking around for a while," as Citrini itself noted. Yet the firm called Muse "a watershed moment, not necessarily because of its technical abilities but because of its reach." That distinction matters.
Muse is not launching into a vacuum. It arrives inside an ecosystem that spans Facebook, Instagram, and WhatsApp - platforms with billions of users - and it is designed to work the way people already communicate. Users talk to it in the Muse app or directly in WhatsApp as if messaging another person. It runs on a dedicated secure virtual machine, with a separate "Sentinel" agent that approves any internet-bound action, and it can connect to third-party services such as Gmail and OpenTable to complete tasks on a user's behalf. It keeps working after the app is closed and comes back when it needs approval - before sending an email or making a purchase.
The payments plumbing is already being laid. At launch, Muse checked out using Link built by Stripe, becoming the first AI agent covered by Link's purchase protections. On Sept. 22, PayPal announced a partnership that lets its customers shop and check out through Muse agents across PayPal's global merchant network. Plaid has also integrated with the agent for personalized financial management. The pattern is clear: the infrastructure for agent-driven commerce is being built faster than the business models it threatens.
Meta's own stock told the other side of the story. Shares rose 10.7% on Monday, putting the company on track for its best month in more than a decade, as Muse topped the App Store charts. For investors who have watched Meta spend heavily on AI infrastructure for years, Muse offered a tangible product to weigh against the spending. The market, in other words, is not just pricing a threat to banks - it is pricing a potential winner, too.
There is also a competitive race underway. Instinct, an invite-only personal AI assistant from Spear Street Technology founded by 23-year-old Noah Shinn, went viral in August and drew both early praise and privacy questions. Meta launched Muse weeks later. The existence of two credible consumer agents in as many months suggests the category is moving from prototype to product faster than most industries have planned for.
The Counter-Case: Why the Selloff May Be Premature
The strongest argument against the bears is the simplest: adoption is not the same as behavior change. Just because a consumer has an agent does not mean they will ask it to switch their bank, renegotiate their insurance, or cancel a subscription. Trust is the real bottleneck. Handing an AI agent access to your finances, your email, and your purchase history requires a level of trust that most consumers do not yet grant to any software company - least of all one with a history of monetizing personal data through advertising.
Wealth advisor Mark White framed the counter-thesis directly: "Clients still need someone accountable for understanding their full situation, exercising judgment and helping them make decisions when the stakes are high." For complex, high-stakes decisions - a mortgage, retirement planning, a serious insurance claim - an agent that compares prices is not a substitute for accountability. That is why some investors believe financial stocks could reverse course quickly if the industry learns to use the technology rather than merely defend against it. "The question is which companies adapt quickly enough to benefit," White said, noting that AI could help banks and financial firms lower costs and improve customer service.
There is also the matter of corporate resistance. Amazon has already blocked Muse from completing purchases on its website - a reminder that incumbent platforms have both the incentive and the tools to defend their turf. Banks, insurers, and telecoms are not passive targets; they have capital, regulatory relationships, and their own customer data. Many are already deploying agents of their own. A report from the Capgemini Research Institute found that nearly half of banks and insurers are creating roles to supervise AI agents, and McKinsey has estimated that AI adoption could drive up to 20% in net cost reductions for banks. The firms that weaponize agents to serve their own customers - proactive renewal alerts, automated claims handling, personalized product matching - may retain the relationship even in an agentic world.
Finally, the market has a history of overpricing disruption at the moment of maximum excitement. The SaaS selloff earlier this year on agentic-tool news is a recent example. In many cases, the initial panic gave way to a more nuanced read once actual usage data arrived. The same could happen here.
Cyclical Panic or Structural Shift?
This is the judgment the market is being asked to make, and it is worth stating plainly: the short-term move is cyclical panic, but the long-term direction is structural. The two should not be confused.
The cyclical leg is the 2.4% intraday drop in financials, the 11% plunge in Planet Fitness, the reflexive repricing of any stock whose revenue model depends on customers not shopping around. That leg is driven by sentiment and positioning, and it can - and probably will - revert as traders digest actual adoption data, as companies report earnings that show little near-term impact, and as the initial shock fades. History suggests the first reaction to a disruptive technology is almost always too violent.
The structural leg is different. It is the slow, durable erosion of friction as the unit economics of switching approach zero. Even if only a minority of consumers - the young, the tech-comfortable, the high-value shoppers - delegate price-sensitive decisions to agents, that minority is disproportionately the one that companies compete hardest to keep. Their defection compresses margins at the top of the customer pyramid first. Over time, as agents become more capable and more trusted, the behavior spreads. This is not a mean-reverting cycle; it is a change in the cost structure of consumer decision-making, and cost structures do not revert on their own.
The evidence for the structural read rests on three pillars. First, the distribution is real: Muse reached the top of the App Store within days, not years. Second, the infrastructure is being built in parallel - Stripe, PayPal, and Plaid are all integrating, which lowers the barrier for agent-driven commerce regardless of which agent wins. Third, the incentive is one-directional: once a consumer experiences an agent saving them money with a single request, the habit of manual shopping becomes harder to justify. The burden of proof now sits with the incumbents to show that their moats survive the removal of friction.
What Comes Next: Scenarios and Signals
The base case is a period of elevated volatility for "consumer inertia" names, with the sector underperforming until earnings reports provide concrete evidence of either resilience or damage. Banks and insurers that can demonstrate they are deploying agents to improve retention - rather than merely absorbing the cost of customers who leave - should begin to separate from the laggards. The beneficiaries in the short term are the agent platforms and the payments infrastructure sitting between agents and merchants: Meta, Stripe-linked checkout providers, and the networks that can prove they add value in an agentic flow.
The upside case for incumbents is straightforward: if adoption proves shallow and trust remains low, the selloff reverses. A single strong earnings quarter from a major bank or insurer, showing stable retention and fee income, would be enough to reprice the panic. The downside case is that adoption compounds faster than expected - if a visible share of consumers begin routinely delegating bill negotiation and insurance renewal to agents, the pressure spreads from the named basket to utilities, subscription software, and any business with recurring revenue and negotiable pricing.
There is one signal worth watching above the rest: the share of consumers who report using an AI agent to complete a financial or billing task in the past month. A sustained move above roughly 15% in a credible consumer survey would confirm the structural thesis is moving from theory to behavior; a figure stuck in the low single digits would support the counter-thesis that this is a curiosity, not a habit. Second, watch the agent platforms' own disclosures: transaction volume routed through Muse or Instinct, and the take rate those platforms can command. If agents become toll collectors, the toll will show up in their revenue before it shows up in banks' attrition data.
For now, the market has made its first call: friction is a profit center, and it is being automated. Whether that call is early or exactly on time will be decided not by what the agents can do, but by what consumers are willing to let them do.
Data as of the close of trading on Sept. 23, 2026.
Explore more exclusive insights at nextfin.ai.
