Meta brings Muse to Apple and Android watches with three ways to interact

You're walking to a meeting and remember that you need to book a table for the evening. Pulling out your phone, finding an app and typing a request is inconvenient. Meta is offering a way to hand the task to Muse directly from your watch: the agent has arrived on Apple Watch and Wear OS devices. You can interact with it through voice, text or dictation. TechCrunch reports on the launch.

For developers and product managers, what matters here is the moment someone turns to the assistant. If a task comes to mind while you're on the move, being able to say it immediately may prove more useful than another feature inside a mobile app. But a brief request from the wrist sets demanding requirements: the agent must understand the task, clarify important details and communicate the result clearly.

What Muse brings to the wrist

Muse lets users connect their accounts and handle email, meetings, bills and other everyday tasks. The service's advertised capabilities include bookings, grocery orders and shopping. Watches now provide another way to access this assistant.

It is important to distinguish access to the agent from the execution of a specific action. The arrival of a watch app does not mean every integration is available to every user, or that a purchase can be completed without additional steps. The original report does not describe these conditions in enough detail. It is too early to rely on the watch as a fully independent task executor.

Take booking a table as an example scenario, rather than a verified demonstration. The user says: “Find a table for two tonight near the office.” Questions then arise: what time, what budget, and which office? The convenience of a product like this depends on how quickly it resolves that uncertainty. If every clarification requires opening the phone, the advantage of a brief voice request diminishes.

For product teams, completing the task is what counts

Our editorial assessment: the launch shows why teams building agent products should test scenarios beyond the familiar chat interface. Long explanations are difficult to read on a small screen. Users need concise, clearly distinguishable states: request received, clarification needed, action awaiting confirmation, task completed or an error occurred.

This is especially apparent with actions that have external consequences. There is an important difference between “found a suitable restaurant” and “made a reservation.” The interface must make that distinction clear. Otherwise, someone may walk away believing the matter is settled when the agent has only prepared an option.

For your own product, it is useful to measure the entire journey: how much time passes between the initial request and the result, how many clarifications are needed, and how often users continue the task on their phones. The number of voice requests alone says little about convenience. People can talk to an assistant frequently and still finish the work manually.

You can begin testing with one common scenario the team already understands well. Preparing for a meeting, for example: a brief instruction, one clarification, then a clear status update. This will reveal where voice saves steps and where it merely shifts text input into a conversation.

What the launch does not yet prove

The report contains no measurements of latency, task completion rates or speech recognition quality in noisy environments. It also does not disclose connectivity requirements, compatible watch models or the regional availability of all features. The announcement alone is therefore insufficient to judge how convenient Muse is in practice.

The service has also received other recent updates: themed Spaces and conversations with an animated avatar. But these do not answer the central question about the new interface: can someone quickly delegate a task and trust the status shown on screen?

The practical takeaway for a founder or product manager is to check whether your agent has at least one task that is truly convenient to delegate on the go. If it does, consider voice input and a brief confirmation of the outcome. The value comes when users complete a task in fewer steps.