Gemini’s ‘Agentic’ Era is here, it can now automate multi-step tasks on Android apps

Google has launched a new update to Gemini that automates multi-step tasks on Android phones. The chatbot will be able to handle multi-step tasks like ordering from food delivery apps or booking a ride by navigating third-party apps autonomously.

Google is reportedly using a secure virtual window that transforms Android’s OS from a passive tool into a proactive AI agent. 

The tech giant is releasing the feature as an early preview, and it will soon be available as a beta feature in the Gemini app for specific Android phones such as Pixel 10, Pixel 10 Pro and Samsung Galaxy S26 series. Initially, Gemini will support selected apps in categories like food, grocery, and rideshare.

How will Gemini automations work?

Gemini’s new multi-step automation feature will act as a "personal agent" on Android devices that will navigate third-party apps to complete desired tasks. Firstly, users will have to trigger Gemini to give a command like "Order my usual pepperoni pizza from DoorDash.” Gemini will analyse the app, order history, and land on the checkout page for the final step.

Google assures that users “can monitor Gemini’s progress live via notifications, allowing you to view, jump in or stop the task.” The company claims that the automation runs in a Secure Virtual Window, so it can only see and interact with the specific app required for the task. Hence, it will save time and effort for users in making repetitive tasks.

The Gemini multi-step task feature is initially rolling out for users in the U.S. and Korea. With the feature, Google also joins the agentic AI race, where rivals like Anthropic and OpenAI are scaling their enterprise AI agent. However, Google’s move brings that same autonomous execution directly into the pocket.