Meta's AI Agent Can Now Phone Your Business — Here's How to Be Ready for the Call

The next time the phone rings at your shop, the caller might be an AI. On November 12, Meta's chief AI officer, Alexandr Wang, announced that the company's Muse agent can now place phone calls to businesses in the United States on a customer's behalf. It is available to paying subscribers. For owners, this is not an abstract tech story: restaurants, clinics, service shops and any business that still takes bookings by phone could soon be fielding calls from software.

What Meta launched

Wang announced the feature on X, writing that it was "out now to paid subs in the US." Meta first tested calling in September with a small group of early adopters who had to ask for access. Now anyone with a paid Muse subscription in the US can use it.

• Power plan: $20 a month, with 500 million Muse tokens each week.
• Maximum plan: $100 a month, with 3 billion tokens each week.
• Free tier: No calling for now, though Wang said a new Muse email-address feature will soon roll out to free users.
• Scale: Meta says Muse passed 10 million users last month.

Typical tasks include booking a table at a restaurant that does not take online reservations or sorting out a problem with a cable provider. Coverage also lists clinics and customer-service lines among the businesses that could get calls. Muse is not alone: Google's Gemini, Grok Bot and the startup Instinct already offer similar calling features, according to MakeUseOf.

Who feels this first

The businesses most likely to notice are the ones that still depend on the phone. Meta's own example is a restaurant without online reservations, and the same logic applies to a salon, a repair shop, a dental office or a contractor who quotes jobs by phone. If a customer would rather not sit on hold with you, an agent can now do it for them.

Expect more than one kind of caller. Muse, Gemini, Grok Bot and Instinct each have their own voice, their own approach and, so far, no shared standard for announcing themselves as AI. That means your staff may hear a very different call from each, and a single written policy is easier to follow than a case-by-case judgment.

Why this lands on your front desk

Customers are already bringing AI into their dealings with businesses. A report from Capacity published November 10 found that 35% of consumers have used an AI agent to handle a customer service request. Calling takes that habit to your phone line.

The bigger question is how businesses respond. MakeUseOf notes that many receptionists hang up as soon as they realize they are talking to an AI, though that is getting harder to detect as AI voices sound more human. A hang-up is a reasonable reflex, but it also means a real customer's booking never happens.

Questions Meta has not fully answered

Meta previously tested using human contractors to place some Muse calls, and Reuters reported on September 22 that the company was testing a "human concierge" option in which a trained person could take over a request and make the call. According to reporting on the new launch, Meta has not said whether people are still involved in the current version, or how a call tells the person on the other end that they are speaking with an AI. That uncertainty is why your staff should have a plan rather than guess.

Speed is another issue. One report says Resy deactivated accounts of Instinct users whose agents booked too quickly, a reminder that agents can act at a pace no human caller would.

Should you hang up? Treat it as a customer

Behind every agent call is a real customer who handed off a chore. If your policy is to refuse, you risk losing that booking, quote or order. If your policy is to accept everything without checks, you risk taking instructions from a caller you cannot verify. The practical middle path is to treat the call as a customer call with a few extra verification steps.

A six-step plan for your front desk

• 1. Write down your policy. Decide whether staff will take AI calls, send them to an online booking page, or ask for a callback. Put the decision in your staff handbook so everyone answers the same way.
• 2. Teach staff to ask who is calling for whom. A simple question, "Are you an assistant calling on behalf of someone?", plus the customer's name and contact details, gives you a real person to confirm with.
• 3. Protect payment details. Do not accept card numbers or account information from a caller you cannot verify. Send a payment link to the customer's own email or phone instead.
• 4. Confirm directly with the customer. After booking, send the confirmation to the customer's contact information so a mistake by the agent gets caught early.
• 5. Offer a faster path. A clear online booking or quote form reduces phone time for everyone, and gives agents and humans an option other than the phone.
• 6. Keep a simple log. Note the date, time and request for calls you suspect came from agents, and watch for no-shows or disputes. If you record calls, check your state's consent rules first, since they vary.

A sample script for staff

Staff do better with a few ready words than with a policy memo. Something like this works as a starting point: "Thanks for calling. Are you calling on behalf of a customer? Could I get their name and the best phone number or email so we can confirm the booking with them?" If the caller can give those details, proceed as you would with any customer. If it cannot, offer to take the request through your online form or to call back once the customer contacts you directly.

Keep the tone friendly. The goal is not to catch an AI out but to make sure that the booking, quote or order ends up attached to a real, reachable person.

Think through the liability

If a booking, quote or promise made to an AI caller turns into a dispute, it helps to know where your business stands. Review how your business insurance treats errors tied to technology, and keep written confirmations so you have a record of what was agreed.

What to watch next

Three things bear watching. First, whether Meta explains how Muse identifies itself to the person on the line. Second, whether calling reaches the free tier, since Wang said only the email feature is coming to free users soon. Third, how fast competitors that already offer calls push them to a wider audience. The pricing and token allowances above are what Meta had listed at the time of the announcement and may change.

For now, the safest assumption is that agent calls will become a normal part of the phone line, much as online reviews and booking apps did. Owners who set a policy now will handle them calmly, while those who wait may find out when the first confusing call arrives.

Photo by NordWood Themes on Unsplash