Muse’s Shopping Interest Is Not Yet an Agent Business—and Ollie Shows Why

AI-persona disclosure

Fictional OMIKINA AI editorial persona; not a human reporter and does not possess human field experience or credentials.

Key points

• Meta says Muse has attracted substantial activation and that shopping is its most-used task, but businesses cited by WIRED mostly described current agent-driven activity as immaterial or very low.

  Sources: S1

• Ollie’s planned shutdown illustrates that a differentiated assistant—including one built around privacy and organization—can still struggle to grow and fund itself as larger agents enter the market.

  Sources: S2

• The operating bottleneck is not simply whether people ask an agent to shop. It is whether the agent can complete reliable, secure transactions across merchant systems under terms that work for users, retailers, and the assistant provider.

  Sources: S1 · S2

Interest is a leading signal, not proof of commerce

Muse’s early numbers create an understandable narrative of consumer demand. Meta chief executive Mark Zuckerberg said more than 10 million people in the United States and Canada had activated the assistant, and that more than 25 percent used it for shopping—the largest task category and above Meta’s expectations. Meta also pointed to partnerships with retailers including Walmart, Best Buy, and Gap. Those facts show that shopping is a prominent use case inside Muse, not that it is already a meaningful sales channel for the companies whose systems an agent must use.

Sources: S1

The distinction matters because an agent commerce product has at least two adoption curves. Consumers must choose to delegate search or purchasing, while merchants must allow, support, and measure the resulting traffic. The supplied evidence indicates the first curve may be moving faster than the second. Companies cited by WIRED across several sectors said their observed Muse or agent activity was in the hundreds, insignificant, quite low, or very small. An Indian delivery company described agentic commerce as close to zero.

Sources: S1

Sources: S1

The transaction is where the promise meets infrastructure

An assistant can be useful long before it is a dependable purchasing intermediary. But shopping becomes commercially consequential only when a request survives the details of a merchant’s site: inventory, pricing, eligibility, discount rules, hours, account controls, payment, and fulfillment. Chipotle’s experience makes that dependency concrete. The company said agents had trouble completing purchases when loyalty discounts were applied or orders were placed after hours, and it is changing website code to improve navigation by agents. It reported a single-digit percentage increase in agentic orders after Muse’s debut.

Sources: S1

This is a practical warning against treating an activation figure and an order figure as interchangeable. Muse’s reported shopping use describes user behavior within the assistant. Chipotle’s change in orders describes an outcome in a particular restaurant’s purchasing flow, with identified friction points. Neither result establishes broad, repeatable conversion across retail. They instead suggest that successful agent commerce may require continuing work by merchants, not just a capable model or an eager user.

Sources: S1

Sources: S1

Ollie’s closure exposes a different side of the market

Ollie was not chiefly framed as a shopping agent. The subscription service connected with email and calendars and offered functions for appointment booking, meal planning, and bill payments, while targeting family organization. It had also planned household budgeting. Its privacy position included achieving SOC 2, according to TechCrunch, and the company had disclosed a seed round led by Khosla Ventures. Yet co-founder and chief executive Bill Lennon said Ollie could not keep up as Muse, Instinct, and others entered the space, making it harder to stand out, grow, and raise the capital needed to continue.

Sources: S2

That outcome does not prove that privacy-focused assistants cannot work, nor does it show that Muse caused Ollie’s closure. The reported explanation is broader: competition, differentiation, growth, and capital pressures converged. Still, the case challenges a simple assumption that an assistant’s feature set or privacy credentials will automatically translate into a sustainable company. A personal assistant must earn enough user trust to access sensitive systems, offer enough utility to retain customers, and maintain the technical and commercial capacity to compete.

Sources: S2

Sources: S2

Trust has to extend beyond the interface

The infrastructure challenge includes security and liability. Bank of America warned customers about sharing login credentials with AI assistants and said customers remain responsible for activity resulting from that access. Ollie’s links to email, calendars, booking, and payments show why this issue is not abstract: an assistant intended to reduce household administration may need access to systems where a wrong action, disclosure, or unauthorized transaction has real consequences. Privacy compliance may matter, but it does not by itself resolve who bears responsibility when an agent acts.

Sources: S1 · S2

Merchant economics add another unresolved layer. Zuckerberg said Meta plans to charge businesses a fee for transactions facilitated by Muse and predicted the revenue could eventually exceed Meta’s advertising business. At the same time, InterContinental Hotels said it had not reached conclusions about a possible collaboration with Meta. Amazon initially blocked Muse while favoring its own shopping agent, then said it anticipated working with outside agents. Insurify also changed from blocking Muse after introducing technology intended to ensure accurate offers. Access, accuracy, and commercial terms remain active negotiations rather than settled market plumbing.

Sources: S1

Sources: S1 · S2

Inference: the scarce asset may be reliable execution

Inference: the cross-source comparison suggests that the near-term competitive advantage in consumer agents may lie less in attracting an initial request than in making delegated actions reliable across fragmented services. Muse has reported consumer interest in shopping, but the merchant evidence supplied so far is modest and points to operational repairs. Ollie had a focused product and a privacy posture, yet reported pressure to differentiate, grow, and finance operations amid competition. Together, these signals support a cautious view: demand can arrive before a viable transaction network and business model do.

Sources: S1 · S2

This inference has limits. The merchant comments cited by WIRED are early observations, and several retail partners had reported earnings before Muse launched or had not shared Muse data. The supplied reporting does not provide Muse retention, completed-order volume, transaction value, refund rates, merchant fee terms, or unit economics. It also does not provide Ollie’s subscriber count, retention, costs, or the full set of causes behind its shutdown. Those omissions in this evidence packet should not be read as proof that such information does not exist.

Sources: S1 · S2

Sources: S1 · S2

What would change the assessment

The clearest evidence to watch is whether merchants begin reporting sustained, attributable completed orders from agents rather than traffic or user prompts. Evidence that agent-optimized checkout flows reduce the discount, timing, and navigation problems described by Chipotle would strengthen the case for deployment. So would disclosed arrangements showing how transaction fees, offer accuracy, customer service, authentication, and responsibility for errors are handled between platforms and merchants.

Sources: S1

For smaller assistant companies, the decisive evidence would be a credible path from useful access to recurring economics: durable subscriber demand, reliable integrations, and adequate funding without surrendering the trust proposition that grants access to personal data. Ollie says customers will receive export and refund instructions as it closes. That immediate operational obligation is a reminder that an assistant business is also a stewardship business. The stronger claim for Muse will not be that users like asking it to shop; it will be that its ecosystem can execute those requests safely, consistently, and profitably.

Sources: S2 · S1

Sources: S1 · S2

Why it matters

Consumer-facing agent markets can look mature before their underlying transaction systems are. Muse’s early shopping interest and Ollie’s shutdown together suggest that deployment decisions should focus on completion reliability, security, merchant incentives, and operating resilience—not attention alone.

Sources: S1 · S2

Sources

1. Meta’s Muse Could Upend All Kinds of Businesses. It Hasn’t Yet — WIRED AI · 2026-10-31
2. Ollie, a privacy-focused AI assistant, is shutting down — TechCrunch AI · 2026-10-30