Build a reflective AI practice over time
In this unit you build a reflection routine you can use again and again. First, you study an annotated decision log that shows what good reflection looks like from the inside. Next, you practice by completing a partial log for a new scenario.
Role pathways
Review the pathway that reflects your professional role.
For teachers
Focus on a recurring classroom task where AI is involved, such as drafting feedback, generating practice items, or drafting communication. Choose an outcome tied to student learning, not just speed.
For coaches
Focus on helping a teacher or team document decisions and choose one manageable adjustment. Practice using reflection questions that keep responsibility visible and reduce blame.
For administrators
Focus on a recurring school workflow where AI support is used, such as drafting communication or summarizing notes. Choose an outcome tied to trust, privacy, and consistency.
Worked example
This sample completed decision log is from a real educator scenario. The annotations in the right column explain why each entry is written the way it is—not just what it says. Pay attention to the thinking behind the choices, not just the words.
Step 1 - Document the decision
| Task and purpose Used AI to draft a starting set of feedback comments for student persuasive essays. Goal was to free up time for in-person conferencing during the next class period. |
Document Names both the task and the reason. "Free up time for conferencing" is a specific value—not just "save time." This makes the decision easier to revisit honestly. |
| What the system did Generated one paragraph of feedback per student based on a short prompt I wrote about the rubric focus. Output included general praise and one suggested revision area. |
Document This entry keeps accountability visible. It names three specific actions—read, revised, protected. A log that just says "I checked it" is too vague to be useful later. |
| What I verified or protected Read each student's draft directly before reviewing the AI comment. Revised all comments to include one specific next step tied to the rubric. Didn't include student names or identifying details in the prompt. |
Document This entry keeps accountability visible. It names three specific actions—read, revised, protected. A log that just says "I checked it" is too vague to be useful later. |
| What I decided and why Used the draft comments as a starting point, then revised each one for accuracy and learning focus. Decided this was appropriate with oversight because students still needed targeted feedback based on their actual work—the AI draft alone wasn't specific enough. |
Why it works States the category (appropriate with oversight) and gives the reason in plain language. This makes it easy to explain to students or a colleague without sounding defensive. |
Step 2 - Notice outcomes over time
| Outcome I noticed Students revised more quickly than usual. But after one week, revisions became surface-level. Students were fixing the specific words I pointed to without explaining their reasoning or transferring the skill to new writing. |
Notice This is honest and specific. It names both a positive outcome (faster revision) and a concerning one (surface-level thinking). Vague outcomes like "it went okay" don't lead to useful adjustments. |
Step 3 - Adjust one condition
| One adjustment I tested Added a student reflection prompt after the revision step. |
Adjust One change only. Changing multiple things at once makes it impossible to know what worked. This is the whole point of adjusting one condition. |
| Prompt text "Name one change you made and explain why it improves your work. Name one place you still need help." |
Adjust The prompt is specific enough to generate real thinking. "Explain why" is the key demand — it moves students past surface editing into reasoning about their writing. |
| Why this adjustment fits It increases the learning demand without removing the support. Students still get AI-informed feedback, but they now have to do something cognitive with it — not just respond to it. |
Why it Works The explanation connects the adjustment back to the noticed outcome. This is what makes a reflection log useful — it shows the reasoning chain, not just the action. |
Three tools that support the document, notice, adjust loop
These Microsoft tools support you using the loop.
Microsoft Copilot - for the notice step
When to use: After you used an AI-supported task for one to two weeks and you want to make sense of what you observed.
Open Copilot and describe what you noticed in plain language—for example: "I used AI to draft student feedback. Students revised faster but stopped explaining their reasoning. What might be causing that and what are two adjustments I could test?" Use Copilot's response as a thinking partner, not a final answer. Evaluate whether its suggestions fit your specific students and context before choosing one.
Microsoft Forms - for the document step
When to use: Any time you make an AI-related decision and want to log it in under two minutes.
Build a five-question decision log form for yourself: Task and purpose, what the system did, what I verified or protected, what I decided and why, one outcome to watch. Submit a response each time you use AI for a significant task. Review the responses monthly to spot patterns in your own habits—what you tend to verify, what you tend to skip, and where outcomes are shifting.
Microsoft OneNote - for the adjust step
When to use: When you're ready to build a living record of what you tested and what happened.
Create a OneNote page called "Adjustments I tested." Each time you make one change to your AI routine, add a short entry: what you changed, why you chose it, and what you noticed after two weeks. Over a semester this becomes a genuine record of your professional learning with AI—specific, evidence-based, and yours. Share it with a coach or colleague to turn individual reflection into team learning.