Gemini’s new task automation is a clumsy glimpse into the future

Testing the feature on the Pixel 10 Pro and Galaxy S26 Ultra reveals a system that is currently in its infancy. Gemini acts as an agent, tapping and scrolling through human-centric menus to complete orders. While it successfully manages complex tasks—such as calculating that two half-portions equal a full chicken teriyaki order or cross-referencing calendar events to schedule airport rides—it is far from efficient. A simple dinner order can take nine minutes, with the AI occasionally struggling to locate clearly visible buttons on a screen.

This friction highlights a fundamental mismatch between current app design and AI capabilities. Applications are built with human-centric interfaces, full of ads, images, and clutter that are irrelevant to a machine. Google’s head of Android, Sameer Samat, notes that this reasoning-based approach is a stopgap. The industry is ultimately pushing toward more robust methods like the Model Context Protocol or native Android app functions, which would allow AI to interface with data directly rather than simulating human touch.

Despite the current clunkiness, the ability to use natural language to navigate these apps represents a shift from the limited, timer-setting assistants of the last decade. Gemini does not yet replace the speed of a human user, but it demonstrates a functional, albeit awkward, leap toward an era where AI manages the granular details of digital life.