With the rise of Agentic AI, Uncle Ben's "with great power comes great responsibility" warning has never felt more relevant. AI agents can gather information, use tools, make decisions, and take action with limited human direction. Everybody's talking about them, but ask 10 people what an agent is and you'll get 10 different answers.

So, what makes a system agentic? What separates Agentic AI from Generative AI and ordinary automation? And where does human responsibility still lie?

Depending on the audience and format, participants either investigate AI agents through a game or build one themselves. Each game round reveals an agent's goal, prompt, and seemingly successful output, then challenges the room to uncover the failure, risk, or human effort behind it. General audiences get AI remixes of real business disasters, while technical audiences get cited engineering incidents. In the hands on version, participants interact with a working agent, write its instructions, choose its tools, and try to get it onto a shared live feed.

By the end, participants understand LLMs, RAG, MCP, RLHF, context, and context windows through firsthand experience. They see what Agentic AI can do, where it falls short, and why the growing field of AI Engineering still needs humans in the loop.

A 60 minute talk or hands on workshop for college students and working professionals, with versions for technical and nontechnical audiences.