Google's test gives paying users an AI that makes calls and handles tasks on their behalf.
Outlets are counted by registrable domain, so a broadcaster’s station subdomains count once. 1 of the 3 items repeat an outlet already counted.
Google's test gives paying users an AI that makes calls and handles tasks on their behalf.
Google is testing a feature called "Call for Me" that lets its Gemini AI place calls to businesses on a user's behalf. The AI can navigate automated phone menus, wait on hold, and conduct conversations to complete tasks like checking if a product is in stock, making restaurant reservations, or rescheduling appointments. Users can follow the call through a live transcript and take over the conversation at any moment.
The feature requires specific conditions: it is initially available only to U.S. users who are paying Gemini subscribers enrolled in a beta version of Google's Phone app, and they must own a Pixel 11 phone. According to Google, the AI will use the user's own phone number to make calls directly from their device.
This launch is part of years of Google experiments with automated calling technology. The company previously demonstrated an AI assistant making natural-sounding calls with verbal quirks, and last year introduced "Ask for Me," which let Gemini inquire about pricing or services for users. Other existing features for Pixel owners include "Hold for Me" and "Direct My Call," which manage hold times and navigate phone trees.
Google described this new test as a more capable evolution. While TechCrunch reported that the earlier "Ask for Me" feature did not allow user takeover during calls, the new version adds that control. Google told TechCrunch it is starting the experiment at a small scale because real-world conversations have many nuances and the feature needs refinement before a wider release.
This development places Google amidst several companies working on AI agents that can act for users. TechCrunch noted that Meta has AI agents named Muse and Instinct capable of making calls. The Verge reported that Meta is also working on an automated phone call feature, though it noted that some of those calls might be handled by a person rather than fully by AI.
2 independent outlets filed 3 reports within the same hour.
This story is moving forward entirely on the force of new information and intent. The engine’s deeper technical analysis shows Google is contending with several internal pressures. High 'Adaptive Decay' indicates the company must overcome legacy systems—likely technical debt or older product lines like Google Assistant—that will slow this feature's progress.
Pulling in the other direction, strong 'Interaction Field' and 'Feedback Momentum' illustrate the strategy is clear and that successful user adoption could create its own growth. However, significant 'Cost & Friction' and a steep 'Ethical Gradient' create headwinds, making failure or backlash as likely as success.
Google’s test of Gemini calling businesses is a high-stakes probe into human-machine interaction, but its future is genuinely undecided. The engine finds high uncertainty, with inertia leaning toward a niche feature, but powerful forces could push it toward widespread adoption or a regulatory halt.
Watch for signals by late 2026: whether the feature expands beyond Pixel exclusivity, integrates into core Google services, or attracts formal regulatory attention will determine which of the three main projected futures unfolds.
The AI-powered feature builds a virtual wardrobe from your photos, and is now broadly available after first rolling out to Android users in June.
Google's launching an "early experiment" feature on Pixel 11 that lets users delegate local business calls to Gemini, like making a reservation, checking if a product is in stock, or rescheduling an appointment. According to Google, you don't even need to start the call to have Gemini handle it for you: Instead of dialing
Google says the AI-calling feature will first be available to Pixel 11 owners in the U.S. who pay for a Gemini subscription.
Sources are evidence, not content. Each keeps its own name, its own link and an extract capped at 400 characters; none of it is rewritten into the copy above.
