Gemini Tasks calls the dealership and checks out, free in six countries
Google's Gemini app gets a Task mode that phones businesses, analyzes bills and completes purchases, free in six countries on the new Gemini 4 Hydrogen model.

Google has added a Task mode to the Gemini app that can phone businesses, analyze bills and complete purchases for users. According to the official announcement, it is free in six countries and runs on the new Gemini 4 Hydrogen model.
The post, written by Josh Woodward, Google’s vice president for Labs, the Gemini app and AI Studio, is summed up in its own subhead: “Gemini can now make the calls, pay the bills, and finish your tasks.”

Free in six countries, with the advanced features U.S. only
Google says the update is rolling out at no charge to hundreds of millions of people in the U.S., Japan, Korea, Canada, Australia and India. A further set of advanced features starts in the U.S. only. The post does not say when, or whether, those reach the other five countries.
To use it, open Gemini and flip the home-screen toggle from Chat to Task. In the U.S., users can also forward emails to [email protected] or text Gemini. Setup runs through Settings, then Channels, then Set up under Text with Gemini, followed by phone number verification. Google also mentions an RCS integration for texting.

Woodward’s examples show the range. In one, Gemini called two car dealerships at once to find a Friday morning service slot, then added the booking to his calendar. In another, it analyzed past utility statements in his email after he granted access to a utility portal through a new Google Password Manager integration, and traced a $389 bill to a heat pump running at peak-price hours. In a third, it added grocery items to an Instacart order and handed him off to finish it.
Google says Gemini now connects to more than 60 apps, and developers can apply to its Connected Apps Platform.

How payments and phone calls are limited
Purchases use the Agent Payments Protocol (AP2). Google says it is working with Mastercard and Visa on digital tokens for supported cards, with American Express listed as coming soon. Cardholders keep points, cashback and purchase protections, according to the post. Users can set a hard spending cap, and receipts land in Google Wallet, which expands to iOS starting today. Google says it will add the Universal Commerce Protocol (UCP) next.
On calls, Google says Gemini tells the business at once that it is an AI, that businesses can opt out, and that it will not give out payment or private account details unless the user says so.
Each task runs in a hardware-isolated virtual machine on Google Cloud’s Agent Compute Engine, dedicated to one account and not shared. Google adds prompt-injection filters and Safe Browsing link limits. Gemini is designed to ask for confirmation or pre-approval before clicking “Buy”. Users can set routine actions to “Always allow”.
Google also admits the tool is not finished. It says it has not eliminated all mistakes, and that on some sites Gemini gets most of the way and then hands the screen back for the final click.
Rivals already sell pieces of this
Google’s pitch is that a user’s email, calendar, passwords, photos and cards already sit in its ecosystem, so little data has to leave it. That is a distribution argument more than a technical one, and it matters because the individual pieces already exist elsewhere.
Phone-calling agents for car dealerships are a funded business: Numa tripled its revenue and raised $45M to put an AI manager in dealerships, as we reported earlier today. Gemini’s minivan example puts a consumer-side agent against the dealer-side ones. On the orchestration side, Perplexity’s Multiagent runs Gemini, GPT and Opus in parallel, which is a bet that no single assistant should own the whole task.
Google’s differentiators are the Gemini app’s reach and the access it already holds to users’ accounts. The same access raises the stakes if something goes wrong. Google has already moved on privacy elsewhere this week with AI Mode threads that vanish when closed. Tasks pushes the other way, toward an assistant with broad, persistent access.
What the announcement does not establish
Every result in the post comes from the author’s own use, and none is independently verified. Google gives no success rates, no failure data, and no list of which businesses or sites work well. It does not say how many calls Gemini can run at once outside the demo, what happens when a business hangs up on an AI, or how long task history is kept beyond saying users can delete chat history.
The post points to a technical paper on the defense strategy and a bug bounty through Google’s AI Vulnerability Reward Program. Those will be the first places to check whether the isolation and confirmation claims hold up. Google also says it is working with major websites and security providers on open, verifiable standards so that assistants are no longer blocked by default, an effort that will decide how often Gemini finishes a task without handing control back.
Sign Up for the Daily AI Pulse
One email a day. All the stories that matter.
By signing up, you agree to our Terms of Use and acknowledge the data practices in our Privacy Policy. You may unsubscribe at any time.
Alex CarterAI News & Big Tech CorrespondentAlex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.



