Episode Details

Send us Fan Mail

AI is starting to feel less like a feature you bolt onto a product and more like a system you have to run. That shift is easy to miss until you try to build something real: a workflow that calls APIs, keeps context across sessions, coordinates tasks, pauses for human approval, and resumes later without breaking. Suddenly prompts are not the hard part. Architecture is.

I walk through what Google’s Agent Development Kit (ADK) reveals about the future of AI agents and agentic workflows. The core idea is event driven execution: a runner orchestrates the system while an agent emits events like “use this tool,” “update state,” “store an artifact,” or “request confirmation.” It’s a clean mental model for building an AI runtime with resumable execution, observable state, and tool integration that can actually survive production.

We also get practical about agent design. Not every agent should be an LLM free styling its way through a task. I break down LLM agents for reasoning, workflow agents for deterministic reliability, and custom agents for complex orchestration, then connect that to the deeper takeaway: the model is the decision engine, but tools are the capability. Rich tool ecosystems and clear interfaces will matter more than chasing ever larger parameter counts.

Finally, we talk governance and safety. Tool confirmation and human in the loop controls are not optional if agents can send emails, change data, or trigger real world actions. If you’re a leader, builder, or architect trying to scale enterprise AI responsibly, this is the mindset shift to make now. Subscribe, share this with a teammate, and leave a review with the guardrail you think every AI agent should have.

Want to join a community of AI learners and enthusiasts? AI Ready RVA is leading the conversation and is rapidly rising as a hub for AI in the Richmond Region. Become a member and support our AI literacy initiatives.