Course catalog
You are the context layer. Everything else follows from that.
Find a lesson
Before you start: how lessons work
How it works
Every lesson ends with a prompt you actually run.
Each lesson covers an idea, then hands you a prompt to run in Copilot, not to read about, to run. A mediocre output isn't a dead end; it's the next lesson. You figure out what the prompt was missing, adjust, and try again. That loop builds the skill more than any concept here.
-
Works in any model
What you practice here works in Copilot, ChatGPT, Claude, or whatever comes next. Prompting is prompting. The model changes. The thinking doesn't.
-
Your data, not a demo
These prompts run in Copilot, so responses pull from your actual O365 environment. Real emails, real meetings, real documents, not a generic example about a fictional company.
-
Thinking, not just clicking
The goal isn't feature fluency. It's judgment, knowing when AI helps, when it doesn't, and what to do when the output is almost right but not quite. That's the part that takes practice.
-
Safe to try, safe to fail
Copilot is a low-stakes canvas, nobody sees your prompts. Run it, see what happens. That's the whole method.
The longer game
Each course ends with something that isn't a lesson.
At the bottom of every final lesson, there's a short piece of writing that isn't a task, isn't an assignment, and isn't more curriculum. It's a provocation, a peek at what the skill you just built is actually the beginning of.
Some of them are about your organization: what the research says separates the 10% of companies genuinely getting value from AI from the 90% that have the tool and plateau. Some of them are more personal: what it means to keep bringing your own judgment to the work, what happens when people stop, what kind of manager your team already knows you are even if your dashboard doesn't. You can skip them entirely and the course is complete. Or you can read one and find it lands somewhere it wasn't supposed to.
Start with Foundations
The on-ramp: three courses, in order. Everyone starts here.
Foundations
- 01 Getting Started
Six short lessons on what AI is at MillerKnoll, where you fit, tools, safety, footprint, and next steps. Mostly reading.
- 02 Working with AI
First skills path: Step 0 plus five lessons on prompting, context, iteration, verification, and judgment in Copilot.
- 03 Prompting
Draft, review, ship · five lessons in the Prompting edition.
More paths (11 courses)
Specializations to pick up once Foundations is behind you.
All courses
- 04 Copilot Basics
Five lessons on mental model, GCSE prompting, daily apps, verification, and building a practice (20–30 min each, in order).
- 05 Prompt Engineering
Five lessons on why prompts fail, frameworks beyond GCSE, advanced techniques, a prompt library, and prompting as a skill.
- 06 AI Judgment
Five lessons on how AI gets wrong, verification tiers, data privacy, high-stakes contexts, and calibrated trust.
- 07 Agent Building
Five lessons on what agents are, three Builder paths, five fields, testing, and sharing (20–30 min each, in order).
- 08 AI for Managers
Five lessons on model use, psychological safety, team experiments, champions, and measuring adoption without compliance theater.
- 09 Practical AI Experimentation
Five lessons on framing a question, running an honest test, comparing outputs, debugging bad runs, and reporting what you learned.
- 10 Multimodal AI at Work
Five lessons on choosing text vs slides vs meetings vs images, alignment across formats, and building one multimodal workflow.
- 11 Copilot in PowerPoint — Client Decks
Five lessons from mental model through template setup, section-by-section build, and verify before share.
- 12 Copilot in Excel — Data Analysis
Five lessons on what Copilot sees, clean on a copy, brief sheet, session habit, and insights with verification.
- 13 Weekly Update Agent
Five lessons on chat draft, read as manager, minimal agent, personalize for your boss, and test with messy notes.
- 14 Client Research & Meeting Prep
Six lessons on chat prep doc, cite-or-omit rules, grounding docs, configure agent, and run and verify per meeting.