Google Tests Gemini Calls That Act for Users
The latest Google is testing a Gemini feature that can call businesses from a user’s own phone number, navigate automated menus, wait on hold and speak with human representatives to complete tasks.
· Originally published by ontime+

Key Points
- Google is testing “Call for Me,” enabling Gemini to phone businesses and complete user-authorized tasks.
- Initial access is restricted to US Pixel 11 owners with Gemini subscriptions and the Android Phone app beta.
- The feature moves Gemini beyond answers toward direct interaction with businesses lacking dedicated automated interfaces.
The latest
Google is testing a Gemini feature that can call businesses from a user’s own phone number, navigate automated menus, wait on hold and speak with human representatives to complete tasks. “Call for Me” is designed for requests including checking product availability, reserving restaurant tables, changing appointments and asking businesses to hold items. Users can follow each conversation through a live transcript, approve personal information for sharing and take control of the call at any point.
Details
- Restricted launch: Initial testing will be limited to Pixel 11 owners in the United States who have a Gemini subscription and use the beta version of Google’s Phone app for Android. Google is beginning at a small scale as it refines the system for nuanced real-world conversations.
- How it works: Calls are dialed directly from the device, allowing Gemini to introduce itself and manage the sequence from phone tree to employee conversation. The setup integrates the assistant with the operating system, Phone app and approved user information rather than confining it to a browser chatbot.
- Human oversight: The assistant is intended to handle repetitive portions of a transaction while leaving the user able to intervene. The live transcript shows the exchange as it happens, while controls let the user take over when unexpected questions, policy differences or misunderstandings arise.
- Earlier calling tools: Google first demonstrated AI-assisted calling at its 2018 I/O developer conference, when Google Assistant phoned a salon to arrange an appointment. It later introduced “Ask for Me” for inquiries about services and prices, alongside “Hold for Me,” “Talk to a Live Rep” and “Direct My Call” for navigating phone systems or reaching staff.
- Broader business reach: Phone calling gives Gemini a route into restaurants, retailers, clinics and service providers that may not offer APIs or other digital connections for automated agents. A business does not need to build a dedicated Gemini integration if the assistant can reach it through the same telephone channel used by customers.
- Agent competition: Meta is also developing AI systems that can act for users, as technology companies focus on agents able to execute multi-step tasks. The approach extends generative AI beyond writing, images, code and summaries into searching, navigating software, communicating with people and completing transactions within predefined limits.
Between the lines
The experiment makes the smartphone a bridge between Gemini and the physical economy: users describe an outcome, authorize information and let the assistant handle intermediate steps. Its practical test is not answer quality alone, but dependable performance amid changing menus, background noise, hold times, unexpected questions and variable business policies.
What’s next
The first indicator will be performance during the restricted US beta, particularly whether Gemini can complete calls accurately while preserving user control. Any decision to expand beyond subscribed Pixel 11 owners would mark the next concrete stage of Google’s rollout.