Agentic AI
The Challenge: We're moving past passive, chat-based search. Modern businesses need digital teammates that don't just answer questions, but proactively plan, coordinate, and take action to automate complex workflows.
A voice-activated drone delivery system that allows users to interact directly with an autonomous flight agent. Built using the Gemini Live API, the Agent Development Kit (ADK) and a custom created MCP server, the application allows users to provide natural language instructions—such as "Pick up the blue block and move left"—and translates these vocal inputs into precise flight coordinates. By integrating ADK with a custom MCP toolset, the system manages the drone’s flight path and operates the onboard electromagnetic cargo magnets. This setup illustrates the practical application of agentic workflows by bridging the gap between verbal commands and physical hardware execution in real time.

Empowering Autonomous Execution and Real-World Action
Natural language interfaces: Replace complex, legacy software dashboards with simple vocal interactions that non-technical staff can use immediately.
Secure process automation: Enable secure, governed AI agents to interact safely with physical hardware, APIs, and databases, automating multi-step tasks with strict guardrails.
Who is building with Agentic AI on Google Cloud today?
Starling Bank: Built the Starling Assistant — one of the UK’s first agentic financial assistants. This in-app companion allows customers to manage day-to-day budgets, set savings goals, and organize bill payments via natural speech and text.
Vodafone: Built the Vodafone Business AI Concierge. Using a multi-agent architecture within ADK, the concierge uses voice and data to interact with customers, resolve complex business inquiries, and execute bookings autonomously while localizing data handling for European sovereignty.
Interested to find out more?
