How agentic systems really work — and how to decide when to use an agent, a workflow, or neither.
Everyone’s talking about AI agents. Almost no one can tell you what one actually is — or when you should use one.
What you’ll learn
- Create your mental model for AI agents.
- Classify any task as workflow, agent, or hybrid.
- Spot which augmentation (tools/retrieval/memory) a system needs.
- Right-size autonomy so you don’t over-build or under-build.
Course Content
- Introduction –> 3 lectures • 8min.
- The Augmented LLM –> 5 lectures • 13min.
- Workflows vs Agents –> 5 lectures • 12min.
- Explanation Through Example –> 4 lectures • 24min.
- Recap And Practice –> 1 lecture • 2min.
Requirements
Everyone’s talking about AI agents. Almost no one can tell you what one actually is — or when you should use one.
If you’ve ever nodded along to a conversation about “AI agents” while quietly wondering what really separates an agent from an ordinary chatbot, or whether your project even needs one — this course is for you.
It’s a concept-first, no-code course that gives you something most AI content skips entirely: a clear mental model of how agentic systems actually work, and the judgment to decide when to build an agent, when a simpler workflow wins, and when you need neither.
What you’ll walk away able to do
- Classify any task as a workflow, an agent, or a hybrid — and explain why.
- Spot which capabilities a system actually needs: tools (to act), retrieval (to know), memory (to persist).
- Right-size autonomy so you never over-build a simple problem into an expensive, unreliable one.
- See how it all fits through a single real example, built two different ways.
How the course works
In about 45 minutes of short, visual lessons, we build the mental model, one idea at a time. You’ll learn the one question that defines every system you’ll ever build — who decides the next step? — meet the “augmented LLM” and its three capabilities, and internalize the single most useful principle in the field: autonomy is a cost, not an upgrade. Then we take one real task — a refund assistant — and evaluate it as a workflow and as an agent, so you see every concept in action and understand exactly which to ship.
You’ll finish with downloadable practice exercises, a one-page cheat sheet, and a glossary so the model sticks long after the last video.
No code. No setup. No jargon for its own sake.
You don’t need to be a programmer, and you won’t write a single line of code. This course works at the level where projects are actually won or lost — the design decisions you make before anyone starts building.
A note on scope (so you know exactly what you’re getting)
This course makes you fluent in the design decision — what to build and whether to build it. The engineering that follows (cost, performance, integration, and actually shipping to production) is a deliberate next stage, not part of this course. Getting the design right first is the highest-leverage thing you can do — because the most expensive mistake in AI isn’t building slowly, it’s building the wrong thing well.