Run the First Experiment

Choose row one unless feasibility is Partial and you honestly cannot unblock it this week. Then take the highest row you can run today.

The list is not the deliverable. The run is.

Run the first experiment in Copilot. Time-box it. Bring the output back to the same chat for a debrief: what worked, what was generic, what you would change.

If the experiment worked and will repeat, Practical AI Experimentation can turn it into a proper test. If it failed informatively, that is a successful Lesson 5.

Core principles

  1. One experiment. Not three in parallel.
  2. Use the starter prompt from the table, then adapt — do not start from a blank page.
  3. Debrief with evidence: the draft, the time, the edits needed.
  4. Optional: a second pass on personal life (home, learning) uses the same table — not required to finish.

If it will repeat: Practical AI Experimentation

Check yourself

When is this course actually finished?

Do this in Copilot

Run #1 (or the highest runnable row). Paste the result back and write one sentence on whether it earned a repeat.

Same Copilot chat as the table if you can. Fresh chat is fine if you paste the row.

Debrief the run

I ran the first experiment from my table. Ranked use case: [ROW]. What I did: [STEPS]. Output (or description if confidential): [PASTE OR DESCRIBE]. Time spent: [MINUTES]. Edits before it was usable: [NOTES]. Critique like a colleague: did this earn a repeat, a rewrite of the prompt, or a drop? Give me a two-sentence “tomorrow morning” card if it earned a repeat.

Techniques: Claim spot-check

  • If the output was generic, the next move is usually more context (identity or project pack), not a longer wishlist.
Open Copilot

Did you run this in Copilot? Mark complete when you have tried it.

AI

Part of the MillerKnoll collective

Studio · 2026