Google launched AppFunctions, a Jetpack API that transforms Android apps into functional building blocks for AI agents. Available in early beta on Galaxy S26 devices, the system lets developers expose app capabilities that AI assistants can call directly—like asking Gemini to "Show me pictures of my cat from Samsung Gallery" and having it retrieve, display, and keep those images in context for follow-up actions. For apps that don't integrate AppFunctions, Google built a UI automation fallback that can handle complex multi-step tasks like placing pizza orders or coordinating rideshares through the assistant interface.
This represents Google's most aggressive push toward an "agent-first" Android experience, directly challenging Apple's upcoming iOS agent capabilities and positioning Android as the platform where AI assistants become primary interfaces. The on-device execution model addresses privacy concerns while reducing latency, but the real test will be developer adoption. Google's dual approach—voluntary API integration plus automated UI manipulation—shows they understand the chicken-and-egg problem of agent ecosystems.
What's notable is the limited initial rollout and Google's emphasis on user control mechanisms like manual overrides and purchase confirmations. The UI automation platform doing "zero code" heavy lifting for developers suggests Google recognizes that forcing API adoption would slow ecosystem growth. However, the sources provide minimal technical details about AppFunctions' actual capabilities, security model, or how it handles complex app states and error conditions.
For developers, this creates an immediate decision point: integrate AppFunctions early for better user experience, or rely on potentially clunky UI automation. The Android 17 wider rollout timeline means most users won't see these features for months, giving developers time to evaluate whether agent-driven interfaces will actually change user behavior or remain a novelty feature.