Meta's consumer AI agent Muse handled a morning's worth of deferred household chores in a hands-on test published September 23, 2026, completing an email follow-up, a $7 espresso-part order, and a fallen-tree-branch removal request — then talked its user into $500 of camping gear. The agent, which users personify with a suggested avatar and name (the reviewer picked Marley), runs tasks in parallel by spinning up virtual browser windows and executing web actions on the user's behalf.
The test, run by Allison Johnson, a senior reviewer with over a decade of experience covering consumer tech, is one of the first real-world workflows published for Muse since Meta positioned it as a mass-market agent. The results map cleanly against where agentic AI is genuinely working today and where it still hits walls — namely, phone calls and any site that has decided to block agents at the door.
Muse cleared the easy stuff quickly. Given access to Gmail, it identified an overdue landscaper bid the user had been sitting on for two weeks, drafted a follow-up email, and shortened it on command. It then found a replacement steam-wand assembly for a six-year-old Breville Barista Express, added the $7 part to a cart, and left the checkout for the user to confirm. Delivery was quoted at five to seven business days.
“It was dizzying watching Marley spin up virtual browser windows chasing down my requests”— Allison Johnson, Senior reviewer, The Verge
Key facts
- 01Meta's Muse agent completed a landscaper follow-up email, a $7 espresso-machine part order, and a fallen-tree-branch hauling quote in a single morning.
- 02A camping-kitchen and LavaBox propane fire pit build-out via Muse totaled $500, split between OXO's site and Walmart after Amazon blocked the agent.
- 03Muse cannot yet make phone calls — a security-system plan downgrade the reviewer had put off for one to two years still required a human call.
- 04Amazon actively blocked Muse with a message flagging AI agents as a terms-of-service violation, forcing the agent onto retailer sites that allow it.
The junk-removal task was where Muse did its most useful work. Asked to deal with a fallen tree branch the user had ignored for a week, the agent pulled listings for local hauling services, requested the user's address, phone number, and a photo of the branch, and submitted quote-request forms to multiple companies. One replied by text, and a pickup was scheduled for the following day. That is the agentic loop as advertised: find, contact, negotiate, book.
Shopping is where Muse shifted from useful to expensive. Tasked with assembling a camping-kitchen setup, it tried to access Amazon and was blocked with a message flagging AI agents as a terms-of-service violation — a live signal that major retailers are already treating agentic traffic as adversarial. Muse rerouted to OXO's own site and Walmart and completed the build. A LavaBox propane fire pit and carrying case were added for ambience. Grand total: $500.
Payments were handled through a Stripe connection. Once the reviewer authorized her card the first time, the friction dropped to a single 'Go for it' per cart — which is precisely the outcome merchants have been engineering toward for a decade and the outcome consumer advocates have been warning about for just as long.
“If our wallets were already defenseless in the era of one-click checkout, we're so cooked when AI agent shopping truly arrives.”— Allison Johnson, Senior reviewer, The Verge
The clearest capability gap was voice. A security-system plan downgrade the user had been putting off for one to two years required a phone call, and Muse could not place one. Meta has said phone calling is in the works, and the reviewer speculated openly about the near-term future where her agent negotiates with a call center's agent before either speaks to a human — a scenario every major agent vendor is now building toward.
Meta's positioning of Muse as a mainstream utility is a deliberate contrast to the enterprise-agent framing dominant at OpenAI and Anthropic. The company conceded last week that Muse's design was heavily inspired by rival OpenClaw, which AI Chat Daily covered at the time, but the go-to-market bet is scale: Meta has more consumer distribution than any AI-native competitor and is aiming Muse at the everyday to-do list rather than the developer workflow.
“It became dangerously easy to tell Marley "Go for it" and approve whatever it had just put in my shopping cart.”— Allison Johnson, Senior reviewer, The Verge
The counterweight is real. Muse's email tone required correction, its account-management ceiling is whatever a human phone tree demands, and the Amazon block is a preview of a much larger fight — every retailer with a margin structure to defend will have to decide whether agent traffic is a customer or a scraper. Anti-bot enforcement, checkout-flow changes, and pricing discrimination against agent-originated sessions are all on the table in the next 12 months.
The commercial read is straightforward. Agents that can spend money frictionlessly on behalf of users will move volume — and the platforms that own the agent, the payment token, and the retailer relationship will capture the take rate. Meta has the users and, via Stripe integration, a plausible path to the payment layer. Whether Muse becomes the default consumer agent depends less on its email prose and more on how many retailers decide, like Amazon, that the agent at the door is not welcome.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
