How Google wants Gemini to control Android Apps, taking AI assistance to the next level

Google is preparing to significantly expand the capabilities of its Gemini AI assistant on Android, potentially allowing it to directly control apps on users’ phones, according to reports based on an analysis of a recent Google app beta version.

The move would mark an important step towards so-called agentic AI — artificial intelligence systems that do not just answer questions, but can carry out actions on behalf of users.

From Assistant to Digital Agent

According to Android Police, Google has already been experimenting with AI-powered automation. It recently introduced an Auto Browse feature in Chrome for users on its paid AI tiers, allowing the browser to complete tasks autonomously.

Now, similar functionality appears to be coming to Android via Gemini. According to a teardown of Google app beta version 17.4 by 9to5Google, the feature — internally codenamed “bonobo” — would enable Gemini to interact with apps through screen automation.

In practical terms, this could allow users to ask Gemini to book a ride, order food, or complete other routine tasks without manually opening individual apps.

Limited Rollout and App Support

At least initially, Gemini’s app-controlling abilities are expected to work only with a limited number of apps. This is largely due to the fact that app interfaces change frequently, making automation more complex and error-prone.

Early support is likely to focus on first-party Google apps, as well as a small number of popular third-party services such as ride-hailing or food delivery platforms.

The feature is also expected to require Android 16 QPR3, as Google has already introduced foundational system-level support for screen automation in that release.

Users Remain in Control

Google stresses that users will remain fully in control while Gemini performs tasks on their behalf. Users will be able to stop or take over actions at any time.

However, the feature also raises privacy considerations. According to information found in the app’s code, screenshots taken while Gemini interacts with apps may be reviewed by trained human reviewers to improve Google’s services, provided the user has activity tracking enabled.

Google also plans to warn users not to enter sensitive payment information directly into Gemini chats and to avoid using automation features during emergencies.

Continue reading

Leave a Reply

Your email address will not be published. Required fields are marked *