The Muse Effect: How AI Agents Are Shaking Travel and Shopping Giants

In September 2026, Meta introduced Muse, a personal AI agent that does not merely answer questions but does the work, claiming everything from travel to shopping as its territory. The speaker's thesis is blunt: these agents are not a theoretical future but a practical break that already threatens aggregator platforms. According to about.fb.com, Muse can send emails, make payments and complete transactions such as travel bookings on its user's behalf.

Muse's architecture rests on a virtualized environment called Muse Secure VM , a dedicated secure computer for the agent, powered by Muse Spark, which Meta describes as its most capable model to date. Users talk to the agent like messaging another person, in a dedicated app or over WhatsApp, and decide themselves how much access it gets. According to the technical account on about.fb.com, the system learns from conversations, sharpens over time and turns long-term goals into action plans.

Across the table sits Mostly Borrowed Ideas, introduced as an independent equity analyst with more than 150,000 followers on X, closely read by professional investors. His core claim is that the age of AI agents will leave many businesses worse off by squeezing the commission layers of the value chain. That frame becomes the backbone of a wide portfolio debate stretching from Airbnb and Booking to Uber and Amazon.

Airbnb on the defensive

The most striking moment of the show is the analyst openly admitting he has changed his mind on Airbnb. A view that once trusted the company's moats has shifted to a cautious stance now that agents might skip the booking flow. That change of mind forms the emotional center of the episode and pulls the investing listener into the debate.

Airbnb's cards against agents include unique lodging inventory, a two-sided trust system and years of accumulated review data. In the speaker's account these assets cannot be copied easily, and the value the platform provides has not disappeared. The problem is framed not as vanishing value but as monetization that agents can bypass.

How agents chew through the direct traffic of online travel agencies makes for one of the episode's most technical debates. Once users state requests to an agent instead of visiting a search page, the intermediary's storefront value fades and its commission layer comes under question. In that scenario, platforms face the option of letting agent traffic through a dedicated interface or a paid-access model.

The analyst's general investing principle kicks in here: a business that produces genuine value others cannot easily replicate usually finds a way to monetize it. Amazon's closure of its catalog to outside crawling is cited as an early example of that defensive reflex. Even so, it is openly accepted that monetization terms could end up worse than before and that the business model may have to change.

Booking hits back

Booking Holdings chief executive Glenn Fogel pushed back against that fear onstage at the Skift Global Forum, saying that coding the travel-agent function is not the hard part. What matters, in his view, is building enough trust for travelers to hand over card details and know where their money goes. According to Skift's forum report, Fogel argues that travelers want a real counterpart when sensitive information is at stake.

Booking is not standing still; through an OpenAI advertising deal, Google integrations and AI bets gathered under its Connected Trip vision, it aims to keep the transaction itself. The strategy banks on travelers starting discovery in AI yet finishing payment in Booking's own till. According to Skift's analysis, the company is running five AI bets at once under pressure from AI startups, Airbnb rivalry and regulatory scrutiny.

Airbnb chief executive Brian Chesky takes the opposite line toward outside agents, arguing that agents need their own operating system. Chatbots were designed as single-player interfaces, he says, while trip planning is multiplayer work done with friends and family. According to TechCrunch's early-October interview, Chesky expects agents to need robust software infrastructure on the company side and wants to make Airbnb itself more agent-friendly.

Airbnb is building its own agent too; after shipping AI-powered search in its fall update, the company targets a full agent launch in 2027. That calendar suggests its reluctance to grant booking access to outside agents is a temporary defensive move. According to Skift's September report, Chesky wants to control the transaction and the experience in the agent race rather than merely supply the storefront.

Whether hotels still need platforms like Booking stands out as one of the episode's fiercest questions. If agents can plug directly into hotel inventory, the intermediary's storefront and marketing value looks even more debatable. Payment assurance, cancellation handling and customer service remain listed as the concrete cards platforms still hold.

Proactive commerce rises

The concept widening the debate is proactive commerce , where AI acts before the user even asks. Once an agent is given a goal, watching prices, catching the right moment and completing the purchase flips the classic e-commerce flow upside down. Meta's business-agent products and creator-led product-discovery announcements show this vision is fed from the enterprise side as well.

The strongest rebuttal to the AI-will-break-Booking thesis rests on travel being more than a transaction and trust being a non-transferable asset. In a relationship where credit cards, identities and travel documents change hands, brand and accountability cannot be swapped out easily. That objection stands as the main counterweight to agent optimism and the argument giving platform shareholders breathing room.

Uber and food delivery

The picture looks harsher for Uber; ride and delivery prices comparable through agents within seconds erode the platform's pricing power. That pressure arrives in the same era as driverless cars, forcing the company to defend on two fronts at once. In the speaker's account, autonomous-driving competition combined with agent-driven price transparency makes ride-hailing one of the most exposed links.

DoorDash refused to wait and pulled its own agent out of its pocket, announcing an AI helper that takes orders through Apple Messages. When users text their usual order, the agent remembers history, gathers mixed dietary preferences into one group cart and brings food photos into the chat. According to TechCrunch's late-September report, the company opened a US waitlist while starting drone-delivery tests with selected restaurants in Northern California.

The real moat in food delivery hides in the question of who stands behind the customer when an order goes wrong. Uber Eats' support experience, buried behind layers of menus after a useless bot chat, is told as the breaking moment that sends users to rivals. That story recalls that reliable customer service remains the core of platform loyalty even in the agent age.

Amazon holds the gate

Amazon started the agent era's first great gateway fight by blocking Muse from shopping on its behalf in its marketplace. The retailer's motive reads as a wish to keep the customer relationship and the checkout experience in its own hands. According to Forrester's October analysis, the matter is a collision between Meta's ambition to become the gateway to digital experiences and Amazon's ambition to remain the starting point of product discovery.

The numbers explain why the fight turned sharp; according to EMARKETER's October report, Muse reached 2.5 million downloads between its September 8 US launch and September 21, with Sensor Tower data showing roughly 448,000 daily users by day ten. The same report notes Meta announcing partnerships with Walmart, GameStop, Sephora, Expedia, OpenTable and Shopify while Amazon's ad revenue hit $19.8 billion in the second quarter of 2026. Backed by more than three billion monthly WhatsApp users, that reach means both opportunity and threat for retailers.

Commentary framing this as a break as large as the move from physical stores to the digital cart forms the episode's philosophical spine. Once agents take over everything from groceries to plane tickets, the value of the storefront and the brand gets redefined. According to the interview published on businessinsider.com, this shift carries a scale at least as disruptive as the brick-and-mortar to e-commerce transition.

Amazon's catalog, estimated between 350 million and 600 million listings, is already an ocean nobody can browse alone, which is why shopping starts at the search bar. Since search sits at the heart of Amazon's advertising income, the company hesitates to open that gate to outside agents. In EMARKETER's assessment, retailers should read this picture as the first round at the negotiating table between two giants rather than a definitive signal about the future.

The market reprices

Wall Street took the threat seriously; travel-booking stocks sold off hard on the Wednesday morning after the Muse announcement. According to 247wallst.com, Expedia fell 7 percent to $260.33, Airbnb dropped 6 percent to $151.66 and Booking Holdings sank 5 percent to $156.02, while Meta, the source of the threat, rose 2 percent to $751.12. The consumer sector fund lost 2 percent and the tech index 1 percent in the same session.

On the safety front, Meta says privacy sits at the center of the design through a private virtual-machine architecture carrying each user's data. Yet Muse remains open only to users over 18 and only in the US and Canada, so billion-scale reach claims sit far from today's reality. As EMARKETER reports, that limited rollout shows regulatory and trust questions remain unanswered.

For investors, the conclusion runs through terminal value; once computers grow far faster and cheaper within five years, the cost of agents browsing directly will collapse too. Bypass scenarios that look expensive today will get cheaper as compute cheapens and the guards at the gate erode. The question therefore changes from whether a platform creates value to how it reprices that value in the agents' world.