Demystifying Agentic AI with a Single Toy App

Agentic AI is gaining attention, but what actually defines an agent?

In this talk, we break down the idea of agency by building the same toy app five different ways. Each version introduces a new level of autonomy, from basic LLM use to full multi-agent systems. Through a live demo, you will see how small design decisions, such as letting the model choose a function or control a workflow, change the behavior and complexity of the system.

We will explore tradeoffs in flexibility, control, and safety, helping you make better decisions about when and how to use agentic designs.

Vorkenntnisse

• Attendees should be familiar with generative AI tools, understand how LLMs are used via APIs, and be comfortable reading Python or pseudocode.
  
• It also helps to already know about function calling (aka tool use) and have some awareness of web automation or how modern websites are structured, but these are not strictly required.

Lernziele

Attendees

• will understand what makes a system agentic by showing concrete levels of agency in practice.
  
• will learn how LLMs can take on more control, what the tradeoffs are, and how to design safer, more effective systems by choosing the right level of autonomy for the task.