Agentic AI and Computer Use for Project Managers

Generative AI writes text and waits for a human to act on it; agentic AI plans, executes, observes, and adjusts toward a goal on its own. With capabilities like Anthropic's Computer Use, where Claude reads screenshots and drives the mouse and keyboard, automation reaches the project manager's desk. This course explains how agents work, where the security risks are, and how to use them responsibly.

Who this course is for

• Project managers who sat out earlier automation waves and want to understand what agents change
• PMs considering automating repetitive tasks that span several applications
• Anyone who will supervise or govern an AI agent deployment

What you'll be able to do

• Explain the difference between generative and agentic AI and the four-step action loop: Perceive, Plan, Act, Evaluate
• Identify three risks: prompt injection through visible content, scope creep in autonomous operations, and data exfiltration through legitimate channels
• Apply layered defenses: least privilege, human-in-the-loop checkpoints, detailed action logging, and sandboxed environments
• Use a clear decision rule to choose between Computer Use, APIs, and manual work
• Write agent goal specifications, assess failure modes, and design checkpoints for a small, supervised pilot

Complete the video and knowledge check to earn up to 2.0 PDUs, credited from your recorded learning time.