Instinct's assistant can now pick up the phone on a user's behalf — book a restaurant that takes no online reservations, get someone onto a dentist's cancellation list, argue with a cable company about a bill. Meta's Muse added outbound calling to US businesses at the same time. Instinct, based in San Francisco, is reported to be in talks to raise $1 billion at a $10 billion valuation. The capability meant to justify that number was matched by a rival within days of shipping.
Founder Noah Shin announced the feature, Instinct Concierge, on X, describing it as built for tasks that require a phone conversation or a high-end booking. It is reaching some users in early access, with a wider release planned later.
Meta has turned its own rollout into a mechanism. Muse's calling beta goes first to users who had already asked Muse to place a call, and the company said that when it ships a capability it will give it first to people who had already asked Muse about it. Calling, by Meta's account, was among the most requested. That is a neat piece of product management: the feature request queue doubles as a waitlist, and the people most likely to be satisfied get served first, which is also the fastest way to manufacture a good first impression.
Neither company got here first. Several of Instinct's competitors had already been promoting phone calls as their advantage, and Instinct Concierge is what closed the gap. With it live, every major rival in the category — Instinct, Wajo, Town, Ollie and Muse — now has the same headline feature. A market with at least five serious entrants reached parity on its most demanded capability in a matter of days.
The money moved faster than the product. Instinct raised $350 million at a $2.5 billion valuation last month. The reported $10 billion talks are a fourfold markup in roughly four weeks, over a period in which its defining new feature became table stakes. What investors are said to be buying is text assistants that can work across different services and interfaces on a user's behalf — a description that, on the evidence of this week, fits every company in the group equally well.
Muse's own numbers are the other half of the story, and they are doing more work in the coverage than they can bear. Sensor Tower has Muse growing faster in its first days than the Meta AI app did: more than 730,000 US downloads against 707,000 for Meta AI in its first five days, with Muse reaching second in the overall app charts at one point. Second overall is genuinely hard to hit. Beating Meta AI by about three percent is not. Meta AI is the app Meta is implicitly superseding, launched across the largest distribution network in consumer software, and clearing it by a few points is a floor rather than a signal. The figure that would settle the question — how many of those installs are still in use a month later — is not in the numbers Meta or Sensor Tower put forward, and for a product whose entire promise is running errands unsupervised, retention is the only metric that means anything.
Phone calls are also not the most consequential thing Instinct shipped. In the same week it gave assistants their own email addresses, so they can register and manage accounts, and it began letting users authorise their assistant to contact other people's assistants to handle specific tasks, by adding those people to a trusted network. An assistant that holds an account in its own right and negotiates with another person's assistant is a different category of software from one that dials a restaurant. That is the part of the roadmap that changes what a consumer agent is, and it arrived with no valuation attached and almost no attention.
The demo everyone copied is the easy half. What none of these companies has shown is what happens when the agent books the wrong table, registers an account it should not have, or settles something with a stranger's assistant on terms nobody reviewed. Those mistakes do not stay in a chat window, and the user finds out about them afterwards.