Instinct and Meta’s Muse race to make AI agents that can place phone calls

Updated September 17, 2026 5:25 pm

Instinct and Meta’s Muse have both added the ability to make phone calls, a move that removes one of the last clear differences between the fast-growing AI agents competing to act as digital concierges for consumers. The updates matter because calling businesses remains one of the most common real-world chores users want delegated to AI.

The change underscores how quickly the market for text-based assistants is converging around the same core promise: not just answering questions, but completing tasks across apps, websites and now telephone lines. With both companies adding outbound calling, the competitive edge is shifting from feature checklists to reliability, convenience and trust.

What changed this week?

Two of the best-known consumer AI agents announced calling features almost back-to-back, signaling that phone-based task completion is becoming a standard expectation rather than a novelty.

San Francisco startup Instinct introduced Instinct Concierge, a new capability designed to let the assistant place calls on a user’s behalf. Meta said its Muse assistant can also now make outbound calls to U.S. businesses. Together, the launches put the rivals on similar footing in a category that has been moving at breakneck speed.

These products are part of a crowded race that includes Instinct, Wajo, Town, Ollie and Meta’s Muse, all of which are trying to become the go-to assistant for everyday errands and administrative work.

Why phone calls matter for AI agents

Phone support is important because many businesses still handle reservations, service requests and cancellations by voice rather than through websites or apps. That makes calling one of the most practical tests for whether an AI agent can move beyond chat and actually do useful work.

For users, the appeal is straightforward: an assistant that can negotiate a dinner reservation, wait on hold with a utility company or confirm a spot on a cancellation list could save time and remove friction from repetitive tasks. For companies building AI agents, calling also opens a broader promise of “high-touch” support, where a human-like interaction is still required to finish the job.

How Instinct’s Concierge feature works

Instinct’s founder, Noah Shinn, said the new feature is meant to handle situations that require a more personal or persistent interaction. In practice, that includes booking restaurants that do not accept online reservations, joining a dentist’s cancellation list and dealing with problems such as cable billing disputes.

The company is rolling out the feature in early access first, with a wider release planned later. Instinct described the tool as a “white glove” service, indicating that it is intended for more complex or higher-value requests rather than only routine automation.

Instinct said its new concierge mode is meant for high-touch requests, such as making phone calls, booking premium services and handling tasks that usually require a person to stay involved.

The calling update follows another recent expansion: Instinct also introduced email addresses for assistants, allowing the product to manage sign-ups and account administration. The company further said users can place assistants into a trusted network, enabling them to contact other people’s assistants for certain tasks.

Why this is a notable product step

Instinct has built a reputation for moving quickly, and the calling feature strengthens its pitch as an assistant that can operate across multiple communication channels instead of staying trapped inside a chat window. That broader utility is increasingly important as rivals add the same capabilities.

In other words, the update is less about inventing a new category and more about proving that Instinct can keep shipping the features users expect from a serious digital helper.

What Meta’s Muse added

Meta’s Muse also now supports outbound calls to U.S. businesses, according to the company. The rollout is starting with users who explicitly asked Muse to make a call, which Meta says effectively acts as a feature-request signal for future launches as well.

That detail is important because it gives Meta a built-in way to prioritize the most eager users while gathering feedback from people who already want the feature. The company said new capabilities will first go to users who have requested them through Muse, a strategy that helps manage demand and test usage patterns.

Meta said the calling feature is being expanded in beta first to users who had previously told Muse they wanted phone calling, and the company framed that feedback as a major factor in shaping the rollout.

Meta’s announcement suggests that Muse is not just keeping pace with Instinct, but also using user signals to guide which features surface first. That approach could help the assistant build momentum without relying on broad, simultaneous releases for every upgrade.

How the competitive landscape is shifting

The latest launches reduce one of the marketing advantages competitors had been using against Instinct. Until now, some rivals emphasized their own calling abilities as a differentiator. With Instinct Concierge now rolling out, that advantage has narrowed sharply.

The result is a market that is beginning to look more like a feature race than a philosophy race. Each assistant is still trying to position itself as the most capable personal operator, but the core capabilities are becoming increasingly interchangeable.

That creates a new challenge: if everyone can send emails, make calls, manage reservations and coordinate tasks, then product quality will likely come down to execution. Factors such as accuracy, speed, interface design, memory, trust and user onboarding may matter more than headline features.

Why investors are paying attention

Investor interest in AI agents has intensified because the category is moving beyond novelty and toward practical workflow automation. Instinct is already one of the most closely watched startups in the space, with reported investor enthusiasm tied to a valuation that could rise significantly if funding talks progress.

The company recently raised $350 million at a $2.5 billion valuation, and reports have suggested it has discussed a possible $1 billion round at a $10 billion valuation. Even if those figures are only preliminary, they show how much capital is chasing startups that can prove consumers will rely on AI to handle everyday tasks.

That interest is not only about consumer convenience. Investors are also betting that agents with real-world capabilities could become a durable layer between users and the services they buy from, creating new distribution power and potentially new monetization paths.

How much traction is Meta’s Muse getting?

Muse appears to be drawing notable early interest of its own. According to analytics firm Sensor Tower, the assistant’s app downloads outpaced the launch of the Meta AI app in its first few days.

The firm said Muse recorded more than 730,000 U.S. downloads in the first five days, compared with 707,000 downloads for the Meta AI app over the same initial window. At one point, Muse also reached No. 2 in the app charts, showing that curiosity around the assistant extends beyond Meta’s core app ecosystem.

For Meta, that early performance is significant because it suggests users are willing to try a dedicated agent product if it offers a clearer use case than a general AI utility app. It also reinforces how crowded, and how opportunistic, the consumer AI assistant market has become.

What the download numbers suggest

The numbers do not prove long-term product-market fit, but they do indicate strong launch-day attention. In a market where many AI products struggle to stand out, exceeding an established app’s early download pace is a meaningful indicator of demand.

Still, downloads are only the first hurdle. Retention, repeat usage and task completion quality will matter much more if Muse wants to remain competitive once the novelty wears off.

What users can actually do with these assistants

At this stage, the most useful AI agent tasks are the ones that save time but do not require a fully autonomous decision-maker. That includes the kinds of jobs people repeatedly postpone because they are tedious rather than difficult.

• Book a table at a restaurant that only accepts phone reservations
• Ask to be added to a dentist’s or specialist’s cancellation list
• Follow up on a cable bill or account issue
• Place calls to U.S. businesses for routine information or requests
• Coordinate with other assistants in a trusted network

These are not flashy use cases, but they are the kind of friction-heavy errands that can turn a demo into a habit. If AI agents can prove themselves on ordinary administrative chores, they may earn a place in people’s daily routines.

How do these features fit into the broader AI assistant race?

They fit into a broader push to turn chatbots into action-oriented agents. The industry’s shift is no longer about merely generating text; it is about using that text capability to interact with services, apps and human gatekeepers in the real world.

That has pushed companies to compete on the completeness of their toolset. Some focus on web navigation, others on email, and now phone calling has become another baseline capability. The result is a more mature and more crowded market in which users may eventually compare assistants the way they compare productivity software: on speed, reliability and integration rather than on novelty.

For companies like Instinct and Meta, the opportunity is to become indispensable intermediaries. For users, the promise is simpler: fewer calls, fewer hold times and fewer annoying tasks to manage personally.

What could happen next?

The next phase of competition is likely to center on execution. As the calling gap closes, the differentiators may become less visible but more important: call success rates, latency, natural-sounding dialogue, escalation paths when an agent fails, and the ability to remember user preferences safely.

There is also a strong chance that more assistants will copy these capabilities quickly. Once outbound calling becomes standard, startups will need to show why their approach is better than the next app’s, rather than simply claiming a feature first.

That dynamic is familiar in consumer tech: once one breakthrough proves useful, it spreads quickly. In AI agents, the pace is even faster because users and investors alike are already hunting for signs of genuine utility. Phone calls may sound old-fashioned, but in this market they are becoming a crucial proving ground.

Timeline: how the latest rollout unfolded

The competition moved fast over a short stretch of time, with each company making separate but closely related announcements.

Bottom line for the AI agent market

The latest wave of updates shows that the consumer AI agent market is moving from experimentation to standardization. Phone calling, once a standout feature, is rapidly becoming table stakes for serious assistants.

That is good news for users who want AI to do more than chat, and it is a warning sign for startups that were relying on one or two headline features to stand apart. In a field this hot, the winners are likely to be the assistants that prove they can perform consistently, handle edge cases and earn user trust in ordinary but time-consuming tasks.

For now, Instinct and Meta’s Muse are both betting that the most persuasive AI feature is still the oldest one in the book: the ability to pick up the phone and get something done.