Meta’s Muse can now call U.S. businesses on behalf of paid subscribers, expanding a feature previously available only to selected early adopters. Meta chief AI officer Alexandr Wang announced the wider availability on November 12, 2026, citing calls to insurers, cable companies, restaurants and the DMV.
The change gives paying users a concrete task to delegate: a phone conversation with someone outside the app. According to TechCrunch’s report on the rollout, Muse subscriptions start at $20 per month.
Access and completion are different things here. Meta is opening the ability to request calls to all paid U.S. subscribers. Whether Muse can finish a particular task still depends on what the business needs, what the user has authorized and whether the person answering accepts an AI caller.
From Early Access to Paid Availability
Meta began testing Muse’s calling feature in September, according to TechCrunch. During that period, selected early adopters could ask the agent for access. The November announcement removes that selection step for paying subscribers in the United States.
Wang described the feature as “out now to paid subs in the US,” in the announcement reproduced by TechCrunch. Free users are outside the announced rollout.
The product change matters even without a newly demonstrated technical capability. A limited test can show that an agent sometimes handles a call; general paid availability makes calling something subscribers can expect to request as part of the service.
The scope remains specific: paid U.S. users calling U.S. businesses. The cited rollout coverage does not resolve geographic edge cases, such as using the service while traveling abroad or calling a business outside the country. Readers should not interpret the announcement as worldwide calling access.
I’d treat this as a broader access announcement, not evidence of a new level of reliability. The reports establish who can use the feature, but they do not provide independent measurements of successful calls, failed requests or completed transactions.
What Calls Meta Says Muse Can Make
Meta’s examples cover both straightforward requests and potentially complicated customer-service conversations. TechCrunch lists restaurants, insurance providers, cable companies and the DMV, while Engadget also describes scheduling appointments and waiting on hold.
The advertised uses include:
- Restaurant reservations: Contacting a restaurant, including one without an online reservation option.
- Appointment scheduling: Asking about availability and arranging a booking.
- Insurance inquiries: Calling a provider on the user’s behalf.
- Cable-company calls: Delegating a customer-service conversation.
- DMV inquiries: Contacting a government office about a user’s task.
- Waiting on hold: Keeping a request moving through a phone-based process.
Those are vendor-promoted use cases, not interchangeable measures of difficulty. Asking when an appointment is available requires less authority than changing an insurance policy or accepting a new cable contract.
That distinction matters when setting a task. A useful request would specify the business, the desired outcome and the limits of the agent’s permission. For a reservation, that could mean a date, party size, acceptable time range and whether the agent may accept a deposit requirement. These are practical instructions a user could prepare, not confirmed Muse interface fields.
My view is that information gathering is the clearest starting point. Asking a business about availability gives the agent a bounded objective. Letting it commit money or change an account introduces decisions that deserve explicit instructions.
What the $20 and $100 Plans Include
Muse’s calling feature is tied to a separate paid Muse subscription. Digital Trends identifies the plans as Power and Maximum, while TechCrunch reports the same prices and weekly token allowances.
| Muse plan | Monthly price | Weekly Muse token allowance | Calling access announced |
|---|---|---|---|
| Power | $20 | 500 million | Included for U.S. subscribers |
| Maximum | $100 | 3 billion | Included for U.S. subscribers |
The $100 plan costs five times as much and provides six times the weekly token allowance. That comparison describes the published subscription quotas; it does not establish that Maximum supports six times as many calls.
Neither allowance is a published bundle of call minutes. The cited reports do not explain how a call consumes Muse tokens, whether waiting on hold affects usage, or whether separate calling limits apply.
That missing conversion is important for anyone considering a subscription mainly to avoid phone queues. A large token number cannot tell you how many restaurant bookings or lengthy customer-service calls you can complete.
I would not choose the $100 plan solely for calling on this evidence. Both tiers are reported to include access, and the documented difference is the token allowance. The higher price may make sense for heavier overall Muse use, but the rollout coverage supplies no calling-specific performance or capacity advantage to evaluate.
AI Disclosure Is Separate From Calling on Someone’s Behalf
Engadget highlights a revealing example from Matt Schlicht, a Meta Superintelligence employee and Moltbook creator. Schlicht shared a recreation of a conversation in which his agent contacted a Volvo dealership about appointment availability. He said it waited on the line for 15 minutes.
According to Engadget’s account, Muse said it was calling on Schlicht’s behalf and that it was transcribing the conversation, but did not explicitly identify itself as an AI agent.
The example is an employee-shared recreation, not an independent test or a complete record of Muse’s standard behavior. It still exposes a consequential distinction between three kinds of disclosure:
- Representation: Who is the caller acting for?
- AI identity: Is the caller a person or an automated agent?
- Transcription: Is the conversation being converted into text?
Answering one question does not answer the others. A human assistant can call on someone’s behalf and transcribe a conversation, too. A receptionist hearing those two statements would not necessarily know that the speaker was AI.
For businesses, the distinction affects how a call is handled. A general question about opening hours may need little follow-up. A request involving customer records, payments or an account change may require the business to establish the customer’s identity and the caller’s authority.
That is an operational issue, separate from any legal assessment. The rollout reporting does not provide enough detail to evaluate Muse’s disclosure practices across jurisdictions or call types.
My view is that a clear AI introduction would make these interactions easier to manage. It would let the recipient decide early whether to continue, transfer the call or ask the customer to contact the business directly. The Volvo example alone cannot establish whether Muse consistently provides that introduction.
Businesses Can Still Decline the Conversation
TechCrunch reports that some businesses and receptionists hang up when they recognize an AI caller. Its coverage does not quantify how often that happens, so it should not be read as a representative rejection rate.
The practical consequence is straightforward: the subscriber pays Meta, but the business controls its own end of the conversation. Broader consumer access does not guarantee that a restaurant, insurer or government office will accommodate the agent.
Frequently Asked Questions
3 questions
1Who Can Use Muse’s Calling Feature?
Meta says the feature is available to paid Muse subscribers in the United States. The announced capability covers calls to U.S. businesses; free users are not included.
