Mixing “fast Copilot Chat with no files” against “deeper Copilot in Word with the brief attached” teaches you nothing clean. Pick a design and write it down before you run anything.
If you change three things at once, you learn nothing.
Default design for this course: two contestants inside Copilot. Same prompt. Same attachments. Same web grounding on or off. Vary only the option — faster vs deeper, or Chat vs the app that holds the file.
Reasoning depth counts as a contestant. If Copilot lets you set it, freeze it to the same band for both — or deliberately compare low vs high as the experiment. Do not leave it on whatever each tab happened to default to.
Core principles
- Same prompt text. Same attachments. Same web grounding on or off.
- Vary one thing: option (fast vs deeper) or surface (Chat vs Word/PowerPoint/Excel).
- Do not paste one contestant’s answer into the next.
- If one run auto-searches and the other does not, that is a confound — note it, do not ignore it.
- Work files stay in Copilot. Do not export a customer brief to a consumer chat to “make the test fair.”
Check yourself
What must stay constant for a fair Copilot comparison?
If the prompt and the option change together, you cannot know what caused the difference. Hold the work constant. Vary one contestant.
Do this in Copilot
Write your comparison design in four lines: task type, contestant A, contestant B, what you will hold constant.
Keep the four lines. Lesson 3 is the battery.
Lock the design
Help me lock a fair Copilot taste-test design. Task type: [TYPE FROM LESSON 1]. Contestants I can actually open: [A] and [B]. Propose what I must hold constant (prompt text, attachments, web grounding, reasoning depth) and what I am allowed to vary. Flag any confound I already described. Output four lines I can paste into a note: task type, A, B, constants. Do not write the prompts yet.
Did you run this in Copilot? Mark complete when you have tried it.
RecordedNext lesson: Run a Small Battery →
Navigate: press j for next lesson,k for previous.