The engine: a model that predicts the next move
You'll see why the same model that answers a chat prompt in seconds can take a full minute as an agent — pausing, opening a file, and working a goal like "pull the three numbers that changed most since last quarter, then draft an email to my team.".
The temptation to pick the tool first
You'll learn why searching "best AI agent" drowns you before lunch — every vendor claims to do everything, every list ranks differently, and half the names didn't exist a year ago — so you can choose without the noise.
A first run, start to finish
You'll hand a task over and get something usable back. By the end, an agent will have done one real job for you, and you'll know how to fix it when the first pass disappoints.
The three risks worth knowing by name
You'll learn to name the failure modes before they bite — including prompt injection and confabulation, the risks flagged in NIST's Generative AI Profile — so you know exactly what to watch for when you review an agent's work.