Delegated phone calls in Muse
Meta has confirmed that its artificial intelligence assistant, Muse, can make phone calls on behalf of the user to manage specific tasks. This capability allows the agent to contact businesses directly to make reservations, ask questions, or manage appointments, reducing the need for users to handle these interactions manually.
The function integrates into Muse's voice platform, which already supports real-time conversations. Unlike text chats, this modality adds a telephone interaction component where the system acts as an intermediary between the user and the establishment, executing the delegated task.
With a delegated call, Muse turns a request expressed in everyday language into a telephone conversation with an establishment. The user states what needs to be handled, and the assistant conveys that request to the business. The interaction therefore shifts from written exchange to a conversation shaped by the staff’s responses during the call. The purpose is to handle the task without requiring the user to conduct the conversation personally.
The ability to place a call does not mean every request will produce the intended outcome. A reservation, appointment, or inquiry also depends on the establishment’s response and the conditions it communicates. The feature describes a way to delegate contact with a business, not a guarantee that the business will accept the request or that the task will always be completed.
Availability and access requirements
The general availability of Muse's voice features, including the ability to make calls, is confirmed for the US public. Meta has deployed this tool for all users in that market, with no exclusive limitations to paid subscription plans in the current documentation.
The function operates on mobile devices, on the web, and on Mac computers. The company has indicated that the technology will later arrive on its AI glasses and a forthcoming hardware device called Muse Charm, although current availability focuses on standard digital platforms.
The stated availability is limited to the US market; it does not establish that calls are enabled in other countries. Access on mobile devices, the web, and Mac computers places the feature on several digital platforms, but does not by itself extend its geographic reach. The later arrival on AI glasses and Muse Charm is presented as future availability, not as access users can rely on now.
The distinction between available and planned platforms matters: mobile devices, the web, and Mac computers are listed as current access points, while AI glasses and Muse Charm are listed as upcoming. That plan provides no specific date and does not confirm that every voice feature will arrive on each device under the same conditions. Present availability refers to the named digital platforms and the US public.
Supported platforms for Muse Voice
Summary of the devices where the call and voice feature is currently available.
| Platform | Availability |
|---|---|
| Mobile devices | Available |
| Web | Available |
| Mac computers | Available |
| AI glasses | Coming soon |
| Muse Charm | Coming soon |
Permitted tasks and operational limits
Examples of tasks that Muse can execute include restaurant reservations, management of medical or commercial appointments, and basic information inquiries at local businesses. The system is designed to interact with establishment staff to complete the user's request.
There is a relevant technical limitation: Muse does not 'see' the user during the interaction, but listens and reacts based on the conversation. Its gestural responses, when the visual avatar is used, are based exclusively on the verbal content of the exchange, not on the image captured by the camera.
The described tasks involve specific dealings with businesses: requesting a restaurant reservation, arranging an appointment, or asking for basic information. Muse joins the conversation to communicate what the user requested; the outcome may depend on details the establishment asks for or options it has available. The call is a way to handle the request, but the feature description does not establish that Muse can resolve every kind of task or act without limits in its dealings with a business.
The perceptual limitation also defines how the interaction works. Muse listens to the exchange and responds to its verbal content, but does not visually observe the user during the call. If the avatar gestures, those gestures are based on the conversation rather than on a camera image; gestural responses therefore do not amount to interpreting visual cues from the user.
The reservation example illustrates one possible sequence: the user specifies the restaurant, party size, and time; Muse conveys those details and reports the result. The confirmation in that example should not be read as a general guarantee of availability: the establishment may or may not accept the request. For the task to match what the user asked for, the details in the request and the business’s response both matter.
User participation and approval
Interaction with Muse requires the user to formulate the initial request. The agent acts as a representative of the user in the call, but the decision to delegate the task lies with the individual. Meta has not detailed whether each call requires an explicit prior confirmation or if the system operates autonomously once the task is initiated.
The customization of the avatar, nicknamed Jolly, allows users to modify its appearance, but this does not alter the approval logic of the tasks. The focus of the interaction is on the efficiency of call management, not on real-time supervision of every word exchanged.
The initial request establishes what task the user wants to delegate, but does not by itself explain what controls apply afterward. It is not specified whether Muse asks for additional approval before each call, lets the user review the details before contacting the business, or how it responds if the conversation departs from the request. Delegation begins with an instruction from the user, while the authorization process during the call remains unspecified.
Changing Jolly’s appearance affects the avatar’s presentation, not the task’s scope or the decisions the system may make. Visual customization should be distinguished from authorization: it does not establish that the user supervises the conversation live or approves every response. The described interaction emphasizes having the assistant handle the call, but does not detail a real-time review process for what is communicated.
Privacy risks and impersonation
The ability to make phone calls raises risks of identity impersonation if the system does not clearly identify itself as an AI agent. Meta states that the goal is to improve the user's connection with their assistant, but interaction with third parties (businesses) requires the system to be transparent about its artificial nature.
Errors in telephone conversations, such as misunderstandings or incorrect information, could cause inconveniences for the user. The company has not specified specific technical safeguards to prevent impersonation or transcription errors in these delegated calls.
Transparency matters because the establishment is speaking with a system acting on a person’s behalf. If the interlocutor’s artificial nature is not clear, the call could be mistaken for a conversation conducted directly by the user. That possibility raises a risk of impersonation, without establishing that it occurs in every call. Meta’s stated goal does not replace the need to distinguish the intended purpose from the specific safeguards that have been implemented.
Listening or interpretation errors could alter important details in a request and lead to an incorrect reservation, a mishandled appointment, or an inaccurate answer. The description does not detail specific technical measures to prevent those failures, so the ability to converse does not by itself establish accuracy. The result depends on Muse understanding the request and the business understanding and responding to it correctly.
A call also involves sharing the user’s intention with the assistant and bringing that request into contact with a business. The description does not specify whether recordings or transcripts are stored, how long they might be retained, or how information about the establishments is handled. This lack of detail does not demonstrate that records are stored; it means the handling of that data is not specified in the feature description.
Keep exploring
Corrections and transparency
If you spot incorrect or outdated information, send us a correction with the page and source we should review.
Submit a correction