Google gives Gemini hands inside Android apps

For years, phone assistants have basically been glorified search engines. Ask them for the weather or to write a poem, and they're great. Ask them to reorder your usual takeout order, and they usually just punt you into the DoorDash app to finish the job yourself.

That limitation might finally be on its way out. Google rolled out the first Android 16 developer preview this week, and Android Authority's Mishaal Rahman spotted a new API dubbed "app functions." It essentially gives the Gemini assistant agent-like capabilities to take real actions inside third-party apps.

Turning Gemini into an actual agent

Google's official developer documentation describes an app function rather vaguely as "a specific piece of functionality that an app offers to the system." But Rahman dug up a concrete example buried in the code: food ordering.

By hooking into app functions, a developer could expose a specific transaction directly to the OS. That means you could tell Gemini to order from your local Thai joint, and the assistant handles the purchase without you ever seeing the delivery app's splash screen.

Right now, Gemini can tap into personal context if you grant it permission, but it's largely confined to searching your Gmail or checking Google Maps. It reads, but it doesn't really act.

Android vs. Apple: The battle of app frameworks

If this sounds familiar, it's because Google is following the exact playbook Apple laid out with its updated App Intents framework in iOS 18. Apple promised that a smarter Siri would eventually be able to tap inside apps on your behalf, though rumors suggest that functionality won't actually arrive until spring 2025.

Currently, iOS 18's biggest AI leap is just routing tough questions to ChatGPT. Neither tech giant has delivered on the original "do things for you" promise that voice assistants made over a decade ago. But both platforms are now building the actual plumbing required to make it happen.

Why it matters

For smartphone users, this represents the transition from AI chatbots to functional AI agents. An assistant that only chats is a novelty; an assistant that can handle the seven taps required to order dinner or call a ride becomes an indispensable interface layer.

However, the real bottleneck won't be Google's API design—it will be developer incentives. App businesses make their money off in-app screen time, banner ads, upsells, and impulse clicks. If an Android 16 user bypasses the app's interface entirely via Gemini, platforms lose their primary monetization funnel. Convincing companies to willingly give up their UI real estate is going to be an uphill battle.

We think this is the inevitable direction for mobile operating systems, but it sets up a messy ecosystem fight. Apple has the leverage of high-spending users and tighter developer control, yet even it is moving slowly. For Android, where adoption of new OS versions crawls and fragmentation is chronic, app functions will likely remain a neat concept long before it becomes an everyday habit.