In this session, we’ll build an 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝗳𝗼𝗿 𝗔𝗻𝗱𝗿𝗼𝗶𝗱 𝘂𝘀𝗶𝗻𝗴 𝗚𝗼𝗼𝗴𝗹𝗲 𝗔𝗗𝗞 and Kotlin, and see how an agent can move from understanding user intent to selecting tools and executing a multi-step task. The session will include a 𝗹𝗶𝘃𝗲 𝗲𝗻𝗱-𝘁𝗼-𝗲𝗻𝗱 𝗱𝗲𝗺𝗼, showing the a𝗴𝗲𝗻𝘁 𝘄𝗼𝗿𝗸𝗶𝗻𝗴 𝗶𝗻𝘀𝗶𝗱𝗲 𝗮 𝗿𝗲𝗮𝗹 𝗔𝗻𝗱𝗿𝗼𝗶𝗱 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻 rather than just discussing agent architecture. 𝗪𝗲’𝗹𝗹 𝗲𝘅𝗽𝗹𝗼𝗿𝗲: • How to build an 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝘄𝗶𝘁𝗵 𝗚𝗼𝗼𝗴𝗹𝗲 𝗔𝗗𝗞 𝗮𝗻𝗱 𝗞𝗼𝘁𝗹𝗶𝗻 • How to expose Android capabilities as tools an agent can use • How agents use context, reasoning, tool calling, and execution loops • How an agent can complete a multi-step task autonomously • How to design guardrails so the agent can act without giving up application control By the end, developers will understand not only how to build an Android AI agent, but also how to think about the architecture required when Android applications become action oriented instead of purely UI-driven.

Google’s Android CLI and AI agents promise three times faster workflows, but moving from the ideal to the practical is full of friction. Shifting away from heavy reliance on the traditional IDE frequently introduces major headaches with dependency management, CI/CD pipeline complexity, and over-engineered automation scripts that break under real-world conditions. Instead of watching polished workflow demos, this circle discussion is a reality check on modern automation. You should definitely come to discuss concrete experiences on which terminal-driven and agentic workflows actually save time, which ones create more maintenance overhead than they are worth, and where developers draw the line between command-line speed and IDE power.