What this course assumes
Who it is for
People who have used an AI assistant, found it impressive, and now want to give it real work — access to files, to email, to a system that matters. That step is where the interesting problems start, and it is where most of the advice runs out.
What I mean by "agent"
Throughout this course: an agent is a language model plus tools plus permission. A chatbot that only talks is not an agent; it can be wrong but it cannot act. Once the model can read your files, send a message, or change a record, its errors leave the conversation and enter the world. Everything in this course is about that transition.
No vendor, no hype
I do not name a preferred model or platform, because that changes every few months and because the three skills here transfer completely. I also do not tell you agents are about to replace anybody. What I can tell you honestly is what they are reliably good at, what they are reliably bad at, and how to arrange your work so that the bad parts are cheap.
The shape
- What an agent actually is — and the one property that explains most surprises.
- The brief — how to describe a task to something with no context and no memory.
- The bounds — deciding what it may touch, before it touches anything.
- The check — verifying work that always sounds right.
- Knowing when to stop — the tasks to take back.