Key takeaways
- Meta's Muse AI agent can now place phone calls to U.S. businesses for paid subscribers, with the feature expanding beyond its September testing phase.
- Muse offers two subscription plans starting at $20 monthly for 500 million tokens per week, or $100 monthly for 3 billion tokens per week.
- AI phone calling faces real-world obstacles, including trained receptionists who hang up immediately upon identifying artificial voices, though improving voice synthesis makes detection increasingly difficult.
- Muse targets customer frustration with traditional phone-based customer service, which typically involves long hold times, phone trees, and multiple department transfers.
Meta’s Muse AI agent has rolled out the ability to place phone calls to U.S. businesses on behalf of users, a capability that now reaches everyone with a paid Muse subscription. The feature, which moved into public availability after early testing in September, represents a significant expansion of what consumer AI agents can do in the real world.
Muse has grown to over 10 million users, according to Meta’s disclosures last month. The addition of calling functionality is part of a larger strategy to make AI agents useful for everyday tasks people would rather not handle themselves. The target use cases are specific and relatable: getting through to a cable company, scheduling a restaurant reservation with a restaurant that has no online booking system, renewing a registration at the DMV, or calling an insurance provider.
Alexandr Wang, Meta’s chief AI officer and head of Superintelligence Labs, emphasized that these are calls people actively wish they did not have to make. This focus on user frustration, rather than technological novelty, frames Muse’s calling feature as a response to genuine demand.
Paid-Only Access and Subscription Pricing
Moving Calling Out of Beta
During the September testing phase, Muse calling was restricted to early adopters who specifically requested the feature. Meta now offers it to all users with an active paid subscription in the United States. This shift from closed testing to public paid availability typically indicates confidence in the feature’s reliability and user satisfaction during early use.
Two-Tier Subscription Model
Muse subscriptions are priced at two levels. The standard plan costs $20 per month and grants users 500 million Muse tokens per week. The premium plan is $100 per month for 3 billion tokens weekly. Tokens act as a compute allocation system, determining how much AI processing a user can utilize across Muse’s various features, including calling. The disparity between the two tiers, a six-fold token increase for a five-fold price increase, is designed to appeal to heavy users of AI features.
Selective Feature Availability
Calling remains exclusive to paying subscribers for now, but Meta is expanding other Muse capabilities to the free tier. Wang noted this week that a new Muse email address feature would roll out to free users in the coming period. This suggests a deliberate strategy: complex, real-world-interaction features like calling stay premium, while simpler productivity features become more broadly available to encourage adoption and engagement.

AI Phone Calling Is Not New, But Muse Is Large
Google’s Gemini has offered calling capabilities, and a startup called Instinct has built its primary business model around placing calls on behalf of customers. Neither company is inventing the underlying technology. What distinguishes different implementations is voice quality, success rate on actual business calls, and integration depth with the core AI platform.
For Muse, which already integrates deeply into Meta’s messaging and social infrastructure, calling becomes a natural extension. Users who are accustomed to interacting with Muse through text can now delegate voice calls to the same agent without context switching.
The Reality of AI Voices on the Phone
Receptionists Know How to Hang Up on AI
The first problem any AI calling service encounters is immediate rejection. Business receptionists and customer service operators are increasingly trained to identify AI callers by voice or behavior and terminate the call. Many hang up as soon as they determine they are not speaking to a human being. This is not a technical limitation—it is a human decision made by businesses that do not want to interact with AI, or that have policies against doing so.
Better Voices Make Detection Harder
Recent advances in voice synthesis have narrowed the perceptual gap between human and AI speech. Modern language models can produce voices that sound natural, lack obvious artifacts, and maintain conversational flow convincingly. This development helps Muse: if a voice is indistinguishable from human speech, receptionists cannot rely on vocal cues alone to identify the caller as AI. However, this improvement raises its own questions around disclosure and transparency.
The Unsolved Problem of Scale
Even with better voices, the fact remains that some businesses will reject AI callers on principle or policy. Muse cannot solve this at the technology layer. The service only works if the receiving business either does not detect the AI or does not object to it. Widespread adoption would require a cultural shift in how businesses handle inbound calls.
Why Phone Calls Remain Frustrating to Make
Customer service by phone has been deliberately degraded over decades. Businesses deploy phone trees designed to discourage calls, hold times that stretch for hours, transfers to multiple departments, and representatives with limited authority to solve problems. For customers with phone anxiety, hearing disabilities, or simple time constraints, these barriers are formidable.
Muse’s value proposition is straightforward: let the AI absorb that friction instead. An AI agent can sit on hold indefinitely, can repeat information without frustration, and can follow a conversation through multiple transfers. It also never forgets the original request.
What Comes Next for Muse Calling
Meta’s rollout of calling to all paid subscribers signals confidence that the feature works in practice, not just in theory. The simultaneous expansion of email features to free users shows a clear tier strategy: premium features require premium pricing, while basic functionality encourages acquisition.
Whether calling remains permanently paid-only or eventually reaches free users will likely depend on adoption rates and technical performance in real-world calling scenarios. Meta is not typically in the business of leaving paying features on the table when the technology becomes stable, but the sensitivity around AI in customer service may change the calculation.
Source: TechCrunch
Frequently Asked Questions
Who can use Muse's calling feature?
Only paid subscribers in the United States can access the feature. Free users will soon receive Muse email address functionality instead.
How much does a Muse subscription cost?
Muse offers two subscription tiers: $20 per month for 500 million tokens per week, or $100 per month for 3 billion tokens per week. Tokens determine how much AI compute a user can access across Muse features, including calling.
Why do some businesses reject AI callers?
Many business receptionists hang up as soon as they identify that they are speaking to an AI agent rather than a human, whether because the business has a policy against AI interactions or the person prefers not to engage with automated calls.