Have you priced what it costs your business to answer a question? Most owners have not, because for as long as there has been a front desk the person asking has paid most of that cost themselves. Three quotes meant three calls, three hold queues and the same story told three times. That friction rationed the attention of whoever picks up, and it did so without anyone deciding it should.
On 8 September 2026 Meta launched Muse, a personal assistant that browses, shops and messages on a person's behalf. Within two weeks it sat at the top of the US app charts, and the payments side of the story was already finished. Stripe Link was built in at launch, Shopify turned on Shop Pay checkout for Muse across every store on its platform on 21 September, and PayPal joined days later. However fast the checkout rails moved, they are not the part of Muse that reaches a small business. The part that reaches a small business is the inquiry.
The inquiry is now free for one side only
Consider what an assistant does before it buys anything. It asks. It checks whether a clinic takes new patients this month, whether a contractor can quote a leak by Thursday, whether the restaurant can seat eight. Those questions used to arrive one at a time, from one person, with a cost attached to each. An assistant can ask all of them at once, in parallel, to every business that might have an answer.
That said, nothing about the receiving end has changed. A call still needs a person to answer it. A chat still needs someone to staff it. A quote form still needs a follow-up. The inquiry became free for the person asking and stayed exactly as expensive for the business answering, which is the whole problem in one sentence.
What the maths does to a front desk
Keep in mind that a business does not experience this as one assistant. It experiences it as more contacts per real customer. Five clinics called about the same appointment means four of them held a slot, spent a minute or two on the line, and booked nothing. Four plumbers quoted for one leak means three quotes written for nobody.
The consequences show up in three places, and it is worth naming each of them:
- Staff time. Every inquiry costs the same minutes whether or not a person is behind it in that moment.
- Conversion per inquiry. The number of contacts rises while the number of customers stays the same, so every rate a business tracks looks worse without anything having gone wrong.
- Attribution. A callback reaches a person who never rang, because the assistant rang for them. The business cannot tell where the lead came from, or whether it was a lead at all.
All of the above lands before a single purchase is made, which is why the checkout rails closing at the top of the market matters less to a front desk than it first appears.
The phone is the interesting case
Meta has been careful about calling. Its statement on the feature says it will only roll out when it is ready and with proper disclosures, and the September reporting on the tests is worth reading closely. According to 404 Media, calls placed during testing were made by human contractors working from the assistant's brief, who then reported the transcript back. In other words, the first assistant to place calls at scale did so through people.
For a business that answers the phone, this means the question "was that a bot or a human" has a third answer. The voice was human. The reason for the call was not. A label that only knows two values cannot describe that caller, and a receptionist that only screens for robocalls will wave it straight through.
What to do before the volume arrives
It is tempting to wait until the calls are obviously coming. The better move is to count now, while the baseline is clean. The current edition of the Index records what four channels across a fleet of businesses saw in the month around the launch: a reading-layer surge the day after Muse shipped, and not one contact on any channel that declared itself an assistant acting for someone.
That zero is the number to watch. When it moves, the businesses that were already counting will know first, and they will know by how much. Read the current edition for the figures, the four layers for how each arrival is labelled, and the channels for what a phone line, a website, a chat widget and a form can each actually tell you.