Meta's Muse Can Now Call Businesses For You — If You Pay

I've spent years dreading a very specific kind of phone call: the one where you're on hold for forty minutes just to ask a simple question. Meta thinks its AI should make that call instead of you. As of this week, Muse — Meta's consumer AI agent — can dial US businesses on your behalf, but only if you're paying for it.

The feature isn't brand new. Meta quietly tested it back in September with a small group of early adopters, and you had to ask the AI for access before it would do anything. Now it's open to anyone on a paid plan, which turns a lab experiment into an actual product feature. According to TechCrunch, Meta says Muse has already passed 10 million users.

What It Actually Does

The pitch is simple: hand off the errands you keep postponing. Alexandr Wang, Meta's chief AI officer and the head of its Superintelligence Labs division, framed it around the calls nobody wants to make — your cable provider, the DMV, a restaurant that refuses to take online reservations. You tell Muse what you need, it makes the call, and you get the outcome without the hold music.

Meta's own messaging leans heavily into that relief. Wang announced it publicly with a short post telling users that insurance, cable, the DMV, and dinner reservations are "not your problem anymore." It's a good line, and honestly it's the right one — the pain point is real and universal.

The Price of Never Calling Anyone Again

Here's the catch: free users don't get this. Calls sit behind the paywall. Muse subscriptions start at $20 a month, which gets you 500 million tokens per week, and scale up to $100 a month for 3 billion tokens if you're a heavy user. So the convenience is real, but it's priced as a premium tier, not a default.

That's a deliberate choice. Voice calling is expensive to run, and Meta is clearly using it as a reason to convert curious free users into subscribers. One small consolation: the Muse email address that Meta added recently will roll out to free users soon, per Wang.

The Other End of the Line Isn't Thrilled

This is the part I find most interesting, because it's where the product meets reality. Plenty of businesses and receptionists hang up the moment they realize they're talking to an AI. And I get it — a receptionist's job is to talk to people, and an agent on the other end changes the nature of the interaction, especially around consent and recording.

But that resistance has a shelf life. Voice models are getting unnervingly good at sounding human, which means the "are you a robot?" test is going to fail more often. At some point the question flips from can businesses detect AI callers to should they have to. Expect policy, not just technology, to decide that.

Meta Isn't Alone Here

Worth noting that Meta is not first. Google's Gemini can place calls, and the startup Instinct has been offering something similar. What Meta brings is distribution — 10 million Muse users is a very different scale from a scrappy startup's beta list. When a feature like this ships to that audience, it stops being a novelty and starts shaping how ordinary people interact with businesses.

My take: the technology is the easy part. The hard part is the social contract. We've all been trained to accept terrible customer service phone trees as normal, and an AI with infinite patience is a genuinely good answer to that. But the businesses on the other end need a way to handle it honestly — not just a race between better voice cloning and better detection.

If you build products, watch this one. Agentic voice is about to become a default expectation, and the companies that figure out how to receive those calls gracefully will have a real edge over the ones that just hang up.