Understanding AI Agents and Agentic Workflows

This course contains the use of artificial intelligence.

Someone in your organization has already said it: "We should get an AI agent to handle that." It sounds simple. But that one word covers very different systems, from a tool that drafts a reply to one that changes data, orders equipment and spends money on its own. Before anyone builds anything, someone needs to ask the right questions: Does this job need an agent at all? What should it be allowed to do without a person? What happens when something unexpected turns up?

This course teaches you to answer those questions. You'll follow a fictional field services company as its team moves from a chat assistant to a controlled request-handling workflow, one design decision at a time. You'll use diagrams, worked examples and pause-and-decide exercises, with no coding and no software to install.

You'll learn how to:

• Tell a chat assistant, a fixed automation, an agentic workflow and an AI agent apart by their behavior, not their label

• Name the seven building blocks of an agent and see why the model proposes while the system checks and acts

• Follow the action loop: choose a tool, check it's allowed, run it, look at the result, decide the next step, and know when it must stop

• Choose the right level of autonomy for each task, from look up only to draft for approval to act within limits

• Map a bounded workflow with decision points, human approvals, exception paths and clear completion criteria

• Decide whether one agent is enough, and write handoff contracts when work passes between steps

• Protect information from untrusted content and prompt injection, and keep sensitive data out of the model's reach

• Bound actions with least privilege, approval gates, evidence checks, retry limits and activity records

• Test with representative cases, set acceptance criteria, and plan a small pilot with a way back

The course ends with a final case where you write a one-page agent workflow brief for a new situation, then compare it with a worked answer and acceptable simpler alternatives. Risk language follows current security community guidance, including the OWASP Top 10 lists for LLM applications and for agentic applications. You'll get five downloadable resources in Word, Excel and PDF: the case pack, a workflow map worksheet, a test matrix, an agent workflow brief template and an agent design checklist.

This course is for business professionals, team leaders, IT staff and process owners who need to evaluate, scope or approve AI agents and AI-assisted workflows. It isn't a coding course, a guide to any AI product or platform, legal or compliance advice, or a certification. Your organization's own policies always come first.