No time for a textbook? This crash course gives working adults the practical core of agentic AI in one fast pass: how agent workflows are built, how tools get used, how outputs get evaluated, and how automation stays safe when a human is not watching every step.
Douglas Barrett moves quickly and concretely. Agents are defined by what they do, plan, call tools, observe results, and repeat, and each part of that loop is taught with the failure it causes when skipped. Tool use is covered as an engineering habit: clear inputs, checked outputs, logged actions.
Inside, you will find:
• Agents in one working definition
• Workflows that plan, act, and check
• Tool use without the chaos
• Evaluation: testing what the agent produced
• Safe automation and sensible guardrails
• Human oversight where it actually counts
• A first build, sized to finish this week
Evaluation gets unusual weight for a short book, because it is the skill that separates a demo from a system: test cases, spot checks, and the metrics that tell you whether the agent is helping or quietly drifting.
The course assumes general computer confidence but no specialist background, and it is honest about speed: you will understand the field's working core, and you will also know exactly which parts need slower study before real deployments. Team leads who need the whole group literate before a pilot will find the course teachable: one sitting for the concepts, one build for the hands, and a shared vocabulary, workflow, tool, guardrail, evaluation, that survives the meeting where the real decisions get made. Readers who finish wanting the slower version will find the path marked: the course names the topics that deserve a full book of their own, security, evaluation at scale, and multi-step system design, and says plainly why each one matters.
One focused pass, and agent headlines turn into engineering you can judge. Start tonight; finish with a build.