Hands-on with Gemini task automation on mobile: it's super impressive despite being very slow and failing at some tasks; it can order food, book Ubers, and more
It took nine minutes to order my dinner, but it still feels like the future. … I've been testing out Gemini's new task automation …
Context & Ripple Effects
Gemini’s mobile automation moves its Android integration from answering and retrieving information toward completing transactions in third-party apps. The feature was introduced on select Pixel and Samsung devices in Google’s initial task-automation rollout, following earlier evidence that Gemini’s service integrations were useful but inconsistent with context in everyday Android use.
This hands-on provides the crucial operational test: the capability can complete consequential tasks such as food delivery and rides, but latency and failures still limit when delegation is preferable to doing the task directly. Later coverage groups automation within the broader Gemini Intelligence bundle, making reliability a product-level issue rather than a standalone demo feature.
First-order effects
- Users can delegate some food-ordering and ride-booking flows to Gemini, but a nine-minute dinner order and failed tasks make the feature impractical for time-sensitive or low-friction requests.
- Google gains a visible proof point for cross-app agent behavior, while also exposing execution speed and task-completion reliability as immediate constraints.
Second-order effects
- Uber, DoorDash, and other app partners face a shift in which the assistant, rather than their own interface, may become the user’s primary entry point for completing routine transactions.
- Assistant makers will be judged less on conversational quality than on whether they can reliably navigate app workflows; failures raise the value of confirmations, fallbacks, and clear handoffs to the user.
Third-order effects
- If cross-app automation becomes dependable, mobile assistants can evolve into an agentic transaction layer that reallocates interface control from individual apps to the operating-system-level assistant.
- The current friction suggests adoption will depend on trust mechanisms as much as model capability: users must be able to inspect, correct, and safely delegate actions that spend money or make bookings.
The trend: Mobile AI is shifting from integrated assistance to embedded agents that execute multi-app tasks, with reliability and control determining whether the shift becomes habitual behavior.