A butler in your pocket: how Zuckerberg wants to win the AI race with Muse

One race raises intelligence; the other seeks the mass market

The race towards superintelligence has a powerful image: a model able to tackle problems in quantum physics. OpenAI presents its advanced models through their scientific and professional capabilities; Anthropic highlights coding, research and knowledge work. Competition pushes towards ever harder and more valuable tasks.[1][2]

Muse offers a view of another race: getting people who barely engage with AI to start trusting it with errands. Meta also develops models, and its rivals offer consumer assistants. What sets this approach apart is the combination of easy access, connected services and everyday devices around an understandable promise: “you take care of it”.

The market remains open. The Reuters Institute’s October report puts daily use of generative AI at 28% of respondents in the six countries with comparable data. Across the study’s nine countries, only 18% have paid access, whether their own or provided through work or education.[3]

Most users still use free versions, according to the report. There is an audience to win over: people who ask the occasional question, try a tool now and then or still do everything as before. To enter their lives, the assistant needs to prove useful with something familiar, such as organising shopping or planning a trip.

The butler starts on the phone and finishes the errand

Meta launched Muse on 8 September 2026. People can talk to it through its app or WhatsApp. The company presents it as an agent that plans trips, manages email and keeps working after the app closes, asking for approval for purchases and messages it sends.[4]

Its virtual computer has a browser and connections to other services. That makes it possible to move from a conversation to a task: looking up information, filling in forms and coordinating steps across accounts. Connectors make it easier to access the services each person uses.[5]

A recipe video can become a menu and a shopping list. For someone new to AI, the appeal is getting the work done without learning to use several tools. Muse offers free access with limits and subscriptions for more use.[6] Trying an errand thus requires less commitment than signing up for another monthly service from the outset.

Shopping also needs a way to pay. Stripe introduced Link integrated with Muse for US consumers: the user approves the amount and the system can use a virtual card linked to that purchase.[7] Meta has announced more commerce and payment connections, as well as Expedia further down the line.[8]

Next, it wants to come along without taking out the phone

Meta has presented Muse Charm, a pocket device for talking to the agent through real-time voice.[8] Axios puts its release in December.[9]

The company has also announced Muse on its glasses in the coming months. The assistant will be able to act on what someone is looking at: a product, a sign or a school supplies list.[8] The camera provides the context that would now need to be described or photographed with a phone.

The proposed change in habits is simple. While walking or cooking, people would just say what they need and carry on with their day. The phone serves as the entry point; voice and glasses aim to make asking for help an everyday gesture, available as soon as the need arises.

Amazon has already shut the door on Muse

Amazon blocked Muse’s access to its store in September. The company confirmed this to GeekWire after asking Meta to remove Amazon from the experience. It cited its terms of use and raised concerns about authorisation, agent identification and security.[10]

The dispute began before Muse: Amazon sued Perplexity over its Comet browser and agent in November 2025.[11]

There is also an economic scale to it. Amazon earned $19.637 billion from advertising services in the third quarter of 2026, according to its latest results. That line includes sponsored ads, display advertising and video.[12]

The butler can spare the shopper a journey through sponsored results, recommendations and product pages. The software looks up what it needs; the person receives a selection and authorises payment. For the advertising business, what matters is the human attention that no longer passes through the shop window.

Amazon can keep a sale and lose opportunities to influence it: showing another brand, suggesting an add-on or charging to put a product in front of the customer. That is the commercial tension. The assistant stands between the person who wants to buy and the platform that used to organise the whole journey.

Google may lose the question before showing the results

In “Is Google’s business at risk?”, Juss og KI reported a Pew finding: users clicked on traditional results in 8% of visits with an AI summary, compared with 15% without one. The study analysed US adults’ browsing.[13][14]

A butler takes the change further. If asked to find and buy a washing machine that meets certain conditions, it can handle the selection without the customer opening a results page. The commercial question starts in another interface, and the decision may end there.

Google needs to keep those questions and find ways to monetise its AI answers. A search that ends in a purchase is valuable for the intent it contains. If that intent moves to someone else’s agent, so does the opportunity to display ads, direct traffic and learn what the consumer is looking for.

Shops will also have to convince the assistant

For a shop, the storefront may cease to be where the sale is decided. The conditions the agent can compare gain weight: final price, availability, delivery, returns and fit with the request. Competition for visibility moves into the assistant’s selection process.

Advertising will also seek a place in that journey. Brands will want to influence recommendations, and consumers will need to understand the criteria that rank them. Whoever controls the butler will hold a valuable position: they will hear the need before it reaches a shop or a search engine.

Switching butlers may be costly; Brussels is already watching the doors

Over time, the assistant accumulates preferences, connected accounts and routines. Switching requires rebuilding part of that relationship and perhaps replacing devices. Meta allows people to inspect, edit and download Muse’s memory.[5] How easy it is to move in practice will also depend on the services and habits organised around it.

WhatsApp offers a precedent on rivals’ access. In June 2026, the European Commission ordered Meta, as an interim measure, to restore competing assistants’ free access to WhatsApp Business during the competition investigation.[15] Distribution is already part of Europe’s dispute over this market.

The winner may be whoever gets the errand

My reading is that Meta has chosen a promise that can attract many people: having someone take care of small everyday tasks. An extraordinary model matters; turning it into a habit may decide who keeps the user.

That is why I see Muse as a serious threat to Amazon and Google. If the errand, the search and the purchase move to the butler, who listens first, who recommends and who gets paid to bring a product closer to the buyer all change. Control of that journey is at stake. It is a complete change in the rules of the game.