Google's Gemini Can Now Book Rides and Order Food on Your Phone - No Tapping Required
Google is launching Gemini automation as a beta on Pixel 10 and Samsung Galaxy S26 - long-press the power button, describe a task, and Gemini navigates apps like Uber and DoorDash in the background to complete it for you.
Updated
5 min read
Google is turning Gemini into an autonomous phone agent. Starting in March, Pixel 10 and Samsung Galaxy S26 users can long-press the power button, describe a multi-step task - "book me a ride home" or "reorder my last DoorDash meal" - and Gemini will open the relevant app, navigate it, fill in the details, and prepare the order. All in the background while you keep using your phone.
The feature, which Google is calling "Gemini automation," launches as a beta in the US and South Korea. It is the first time a major phone platform has shipped an agent that controls third-party apps autonomously rather than just answering questions about them.
TL;DR
- What: Gemini can now autonomously navigate third-party apps on your phone to complete multi-step tasks (ordering food, booking rides, grocery shopping)
- How: Long-press power button, describe the task. Gemini opens apps in a sandboxed virtual window, scrolls, taps, and fills forms via screen automation
- Supported apps at launch: Uber, DoorDash, Grubhub, Kroger, Walmart
- Devices: Pixel 10 / 10 Pro / 10 Pro XL, Samsung Galaxy S26 / S26+ / S26 Ultra
- When: Beta rolling out March 2026, US and South Korea only
- Safety: Gemini never completes the final purchase - it prepares everything, then hands you the buy button
How It Works
When you issue a command, Gemini opens the target app in a secure, isolated virtual window running on the device. It reads the screen via screenshots, then scrolls, taps, and types to navigate the app - the same screen automation approach that powers Google's Project Mariner in Chrome, now applied to native Android apps. The processing happens in the cloud, powered by Gemini 3.
The task runs entirely in the background. You get progress notifications similar to Android's live activities. At any point you can open the virtual window to watch Gemini work, jump in to take manual control, or kill the task entirely.
The critical safety guardrail: Gemini will never complete a purchase on its own. It builds your cart, fills in your delivery address, selects your payment method - then stops and asks you to open the app and tap the final buy button yourself. No money leaves your account without your explicit confirmation.
Five Apps, Two Countries
The beta launches with a deliberately narrow scope:
| Category | Apps |
|---|---|
| Rideshare | Uber |
| Food delivery | DoorDash, Grubhub |
| Grocery | Kroger, Walmart |
That is it for now. Google says more apps will be added over time, but the limited rollout reflects both the beta status and the complexity of screen automation - every app has a different UI, and the agent needs to handle layout changes, error states, and edge cases reliably.
Geographic availability is restricted to the United States and South Korea at launch, which aligns with the Galaxy S26's launch markets and Google's existing Gemini feature rollout pattern.
Supported Devices
The feature requires specific hardware:
- Samsung Galaxy S26, S26+, S26 Ultra (shipping March 11, 2026)
- Google Pixel 10, Pixel 10 Pro, Pixel 10 Pro XL (rolling out March 2026)
APK teardowns suggest the feature requires Android 16 QPR3 or later. The internal codename is "Bonobo."
Samsung's "Agentic AI Phone" Pitch
The timing is not accidental. Google announced Gemini automation alongside Samsung's Galaxy Unpacked event, where Samsung branded the S26 series as the first "agentic AI phone." The S26 ships with a three-layer AI stack:
- Gemini as the primary agent (task automation plus an upgraded Circle to Search powered by Gemini 3)
- Perplexity as a secondary system-level search agent, accessible via wake phrase
- Bixby overhauled for contextual device controls
Samsung is also shipping Gemini-powered scam detection for calls and messages, a personalized AI daily briefing called "Now Brief," and a hardware-level privacy display on the S26 Ultra that blocks side-angle viewing.
The Bigger Picture
This is the most tangible step yet in Google's convergence of its agentic AI projects. Project Mariner handles browser automation. Project Astra handles real-world visual understanding through the camera. Gemini automation now handles native app control on phones. All three use the same underlying approach - screen reading, action planning, and sandboxed execution - just applied to different surfaces.
The gap between "AI assistant that answers questions" and "AI agent that does things for you" is closing fast. Google is not alone - Perplexity just launched Computer, a cloud-based agent that orchestrates 19 models to run projects for weeks. OpenAI has Operator for browser-based tasks. Anthropic has Claude computer use at the API level.
But Google has something the others do not: Android. With over 3 billion active devices, even a narrow beta on two phone lines puts agentic AI into more hands than any standalone product can reach. If the screen automation approach proves reliable, expanding from five apps to fifty is an engineering problem, not a research one.
The beta launches in March. Google I/O on May 19 will likely expand the scope significantly.
Sources:
- Android Multi-Step Tasks with Gemini - Google Blog
- Gemini Automation Coming to Pixel 10 and Galaxy S26 - 9to5Google
- Gemini Can Now Automate Some Multi-Step Tasks on Android - TechCrunch
- Gemini on Galaxy S26 and Pixel 10 Can Finally Control Other Apps - Android Authority
- Gemini Is About to Start Checking Off Your To-Do List by Actually Using Your Apps - Chrome Unboxed
Last updated
