Deconstruct Your Workflows for Agentic AI

Hosted by James Gray

198 students

In this video

What you'll learn

Identify the workflows hiding inside your business processes

Name your process and the discrete workflows within it—the first step to making any of them automatable

Deconstruct one workflow into AI-executable steps

Use a meta-prompting technique to have AI write the detailed instructions for each workflow step

Generate your workflow SOP and process guide with AI

Capture your domain expertise as documentation you can delegate from, share, or automate

Why this topic matters

You describe workflows at the "what" level. AI needs the "how"—specific steps, decision points, and data flows. This gap is why agentic AI fails: you say "research prospects" and get generic results because AI doesn't know what qualifies a prospect for your business. Deconstruct one workflow into executable steps, then map them to the building blocks that operationalize it.

You'll learn from

James Gray

UC Berkeley AI Instructor | ex CIO & CPO

James Gray accelerates AI adoption globally, drawing on 30+ years leading data and AI teams as CIO, CPO, and engineering executive at enterprises and high-growth startups.

At Microsoft, he architected global data systems powering critical business decisions. He now delivers strategic consulting and hands-on learning experiences that drive results.

His Hands-on AI for Leaders course has trained hundreds of executives. As an instructor for UC Berkeley Haas Exec Ed, James helps leaders design actionable AI roadmaps for immediate implementation—bridging strategy and execution.

James holds a BS in Electrical Engineering from Union College, an MBA from Berkeley Haas, and an MS in Information and Data Science from UC Berkeley. He is a certified leadership coach through Brown University.

Worked at and Students at

Microsoft
UC Berkeley
Google
Meta
Amazon Web Services

Go deeper with a course

Hands-on Agentic AI for Leaders
James Gray
View syllabus