Google Lets Gemini Wait on Hold So You Don't Have To

The most hated ten minutes in consumer life are the ones spent listening to hold music. Google now wants Gemini to spend them for you. On September 24, the company began testing Call for Me, a feature that lets Gemini place a phone call to a business on a user's behalf, introduce itself, work through the automated menu, wait on hold, and then handle the conversation with whoever finally picks up.

It is an early experiment, and a narrow one. Call for Me is limited to Pixel 11 owners in the United States who pay for a Gemini subscription, and it requires the beta version of Google's Phone app, per TechCrunch. 9to5Google reports that the preview is rolling out to people enrolled in the Phone by Google beta on the Pixel 11, 11 Pro, 11 Pro XL and 11 Pro Fold.

What the feature actually does

The pitch is routine calls, the kind nobody enjoys and few people need to make personally. Google's examples include reserving patio seating at a restaurant, checking whether a hardware store has something in stock, rescheduling a haircut, and placing an item on hold. The user starts the request in the Gemini app. The call then goes out from the user's own phone number, dialed directly from the device.

Two design choices carry most of the weight. First, the user watches a live transcript of the call and can take over at any moment. Second, Gemini can share personal information the user approves as part of the call. That second piece is what separates Call for Me from earlier Google calling features. An assistant that can say your name, your booking reference or your callback number can finish a task, not just ask a question about one.

Google framed the launch around consent. "Asking AI to speak on your behalf is a big step that requires genuine trust," the company said, according to 9to5Google, which also reports that Gemini introduces itself as an AI handling the call for the user. Google told TechCrunch it is starting small because "real-world conversations are nuanced."

Eight years to get here

Google has been circling this idea since 2018, when it demonstrated Google Assistant, under the Duplex name, booking a salon appointment in a voice that mimicked human pauses and filler words. The demo drew as much unease as applause, largely because the person on the other end did not know they were talking to software.

The intervening years produced narrower tools. Hold for Me waited on the line and alerted you when a human arrived. Direct My Call transcribed phone menus so you could tap through them. Last year's Ask for Me let Gemini call businesses to ask about pricing and services. Call for Me stitches those pieces together and adds the ability to act: navigate the menu, wait, talk, and carry approved personal details into the conversation.

Google is not alone. The Decoder notes that Meta's AI agents, Muse and Instinct, can also make phone calls. Outbound calling is quickly becoming a standard capability for consumer assistants, and the differentiator is shifting from whether an assistant can call to how much it can be trusted to say once it does.

The enterprise side of the same bet

The day Call for Me appeared, Google made Gemini 3.8 Live with Live Avatar generally available in Gemini Enterprise. It is the other half of the same idea, aimed at the businesses on the receiving end of these calls.

Live Avatar pairs Google's native live-dialogue model with near real-time video generation, producing an animated face that lip-syncs as it talks. According to Google's announcement, the system can move across 97 languages with automatic language detection, adjusting lip-sync as it switches. It can watch a user's camera feed or screen share, and it calls tools asynchronously, so it keeps talking while it fetches data in the background. Google's demonstration shows the avatar checking a guest into a hotel without pausing the conversation.

Businesses can pick from prebuilt avatars, or generate a custom one from a reference image through an allowlisted program. Google says all audio and video output is marked with SynthID, its imperceptible watermark. The Google Cloud announcement lists the Gemini Enterprise console and the Gemini Live API as access points, with US and EU endpoints.

The early customer list suggests where this goes first. Cox Automotive is using it for an Autotrader shopping assistant. Equal AI, which Google says handles over a million live calls daily across nine Indian languages, credits the new model with better interruption handling and tool-call reliability. Salesforce is exploring an integration with Agentforce.

Machines on both ends of the line

Put the two launches side by side and a picture forms. On one end of a call, a consumer's Gemini navigates the menu and waits on hold. On the other, a business deploys a Gemini-powered agent, with or without a face, to answer. Google is building both halves of a conversation that, in a growing share of cases, may not need a human on either side.

That raises questions Google has not fully answered in public. Disclosure is the obvious one: Gemini announces itself, but the business receiving the call still has to decide how to treat an AI caller holding a customer's personal details. Error handling is another. A misheard date on a restaurant booking is an annoyance; a misheard account detail on a medical or financial call is a real problem. Limiting the test to paid subscribers on one phone line, with a live transcript and a take-over button, is how Google is buying room to learn.

Why it matters

Voice is where AI agents stop being abstract. A chatbot that drafts an email leaves the user to send it. An agent that places a call from your number, speaks for you and shares your information is acting in the world under your name, which is a different level of delegation.

Google's decision to ship this as a gated experiment on its own hardware is a signal that it knows the stakes, and that the Pixel remains its preferred proving ground for features too sensitive to launch everywhere at once. The simultaneous enterprise release matters just as much. If businesses answer the phone with agents and consumers call with agents, the phone tree, the hold queue and the call center all start to look like interfaces designed for a user who is no longer there. The companies that run both ends of that exchange will set its rules.