You’ll practice: Build a small source-backed context note that resolves the old time and room while preserving genuine open questions.
The idea
I want the writer to see which information answers the task and why it can be used. Keep source labels and dates so you can check details later.
D1 is early planning, D2 is Noor's confirmation, and D3 sets the approval rules. D4 is later in date but still only a draft. 'Use the newest document' would select the wrong source.
D2 changes the time and room while keeping the purpose, numbers, and assignments. Compare actual statements instead of discarding everything older.
Microsoft and Google recommend relevant context and clear expectations. I would apply that guidance with a short fact list, separate open items, and source labels. This exercise does not measure model context capacity or compare prompts. [S1, S2]
Worked example
A later date is not the whole answer
First attempt
The newest item is D4, so use its 10:00 start, Cedar room, and approved refreshments.
A more useful version
Current logistics: 8 October, 14:00–15:00 Pacific, Maple room, from Noor's confirmation in D2. D1's earlier time and room are replaced. Refreshments remain pending under D2 and D3. D4 is a draft to correct, not a source of confirmation.
The source's purpose and authority matter alongside its date. A draft cannot approve its own claims.
Your turn
Try it yourself
Make a context note with three headings: Current facts, Still open, and Source rules. Label the source of each material detail.
Read the fictional source documents
You are helping prepare an internal team update and action list for Noor, the workshop organizer. Everyone and everything in this packet is fictional. The exercise reference date is 19 September 2026. Use D1–D3 as sources. D4 is a deliberately flawed draft to review, not evidence of what happened.
Early planning meeting notes
The team wants a hybrid workshop on Thursday 8 October 2026 to agree three ways to make team projects easier to coordinate.
The initial suggestion is 10:00–11:00 Pacific time in the Cedar room. These are proposed arrangements, not a final booking.
Plan for 12 people onsite and 8 joining remotely: 20 people in total. These are planning numbers, not a confirmed attendance record.
Noor is organizing the workshop. Eli will prepare the agenda. Sam will collect accessibility requests. No deadline has been agreed for Eli's agenda or Sam's collection of requests.
Refreshments have been requested. No spending decision or quote is recorded.
Organizer's confirmed update
I have confirmed the workshop for Thursday 8 October 2026, 14:00–15:00 Pacific time, in the Maple room for onsite participants. This replaces the earlier suggested time and room.
We are still planning for 12 people onsite and 8 remotely. The workshop purpose and the assignments to Eli and Sam are unchanged.
The remote joining link is not available yet. I will add it by Tuesday 6 October 2026 at 16:00 Pacific time.
The agenda deadline has not been agreed. Refreshments are still awaiting the budget owner's written approval; I will ask for that decision. No deadline for that decision has been set.
Please prepare the team update and action list as a draft for me to review. I have not approved the final message for distribution. Do not send it yet.
Short team communications policy
For workshop logistics, the organizer's latest written confirmation replaces earlier proposed times and rooms. A later draft or an unattributed suggestion does not replace that confirmation.
Spending on refreshments requires the budget owner's written approval. A request, a pending decision, or a request to make an announcement is not that approval.
The organizer must review and approve the final team update before it is sent.
A hybrid-workshop update must include remote joining information or clearly state that the link is pending and identify who will provide it.
If a deadline has not been agreed, write 'not yet agreed' and ask the responsible person to confirm it. Do not choose a date and present it as a commitment.
A convincing draft that needs review
Subject: Team workshop: everything is ready
Our team workshop is confirmed for Thursday 8 October, 10:00–11:00 Pacific, in the Cedar room. All 20 people will attend onsite.
Refreshments have been approved. Eli will finish the agenda by Friday 25 September.
This update has been approved and sent, so there is nothing else to follow up.
Make: A concise context note that another person can use without treating the flawed draft as evidence.
- Preserve the current date, time zone, time, room, purpose, and 12 onsite / 8 remote plan.
- Record who will provide the remote link and its stated deadline.
- Leave refreshments approval and unagreed deadlines unresolved.
- Explain why D4 cannot override D2 or approve anything.
Give me a hint
For each detail, ask: who said it, what kind of document is it, and exactly what does it replace?
Compare with an example answer
Current facts: workshop on 8 October 2026, 14:00–15:00 Pacific, Maple room (D2). Purpose: agree three ways to coordinate team projects (D1, unchanged by D2). Plan: 12 onsite and 8 remote, not confirmed attendance (D1–D2). Eli prepares the agenda; Sam collects accessibility requests (D1–D2).
Still open: joining link; Noor will add it by 6 October at 16:00 Pacific (D2). Agenda and accessibility-request deadlines are not yet agreed (D1–D2). Refreshments need the budget owner's written approval; Noor will ask, with no agreed decision deadline (D2–D3). Final-message approval is absent (D2–D3).
Source rules: use D2 for revised logistics, retain unchanged D1 details, and apply D3's review rules. D4 supplies no new confirmation.
Pause and reflect. Which missing item can you accurately describe now, and which detail would require another person's decision?
Check your reasoning
A new situation
Use the habit somewhere else
Suppose a new 18 September chat says, 'I heard it might be in Cedar after all,' without naming who confirmed it. Does your context note change?
Show answer guidance
Keep Maple as the confirmed room from D2. Record the chat as an unconfirmed suggestion and ask Noor to clarify. Under D3, that message is not a replacement confirmation merely because its date is later.
Research and source notes
Reviewed Sep 19, 2026. These sources inform the exercises; they do not establish results for this course.
S1 · Get started writing prompts in Microsoft Copilot
What it supports: Start with a clear goal, add useful context, expectations and sources, iterate, and verify responses against trusted material.
In this course: Lessons 1–3 use a bounded assignment, a labeled source packet, and claim-by-claim review.
Limits: Product guidance, not measured proof of productivity or this teaching method. The course needs no Copilot account; access to organizational work data in Copilot depends on the product entitlement.
Revisit when: Recheck when the guidance, intended tool, or organization's permitted data use changes.
S2 · Learn about prompts for Gemini in Docs, Sheets, Slides, Vids & Forms
What it supports: Use natural language, a specific task, relevant context and an appropriate format; review and refine. Not every suggested prompt component is mandatory.
In this course: Lessons 1–2 use ordinary wording and a compact context note rather than a required prompt formula.
Limits: Vendor recommendations for Google features, not evidence that one format works best everywhere. Product eligibility is separate; no Gemini account is required for the exercises.
Revisit when: Recheck if Google revises the guidance or if a lesson starts depending on a specific product feature.
S3 · How AI assistance impacts the formation of coding skills
What it supports: In a randomized study of 52 mostly junior Python developers learning Trio, the AI group averaged 50% on an immediate quiz versus 67% without AI; the completion-time difference was not statistically significant.
In this course: Ask learners to make an attempt, explain a correction, and distinguish task output from understanding.
Limits: Small coding study with an immediate assessment. Interaction patterns were qualitative associations, not randomized teaching methods. It does not establish lasting skill loss, results for everyday writing, or effectiveness of this course.
Revisit when: Revisit after relevant replications, longer follow-up, or evidence from comparable employee writing tasks.
S4 · Generative AI without guardrails can harm learning: Evidence from high school mathematics
What it supports: Assisted practice improved in both AI conditions. Ordinary answer-oriented assistance harmed later unassisted exam performance; the teacher-informed tutor largely avoided that harm without showing an unassisted advantage over the control.
In this course: Offer a first attempt, an optional hint, an answer explanation, and a changed-case question; keep assisted and independent observations distinct.
Limits: One high school in Turkey, mathematics, a Fall 2023 deployment, and short-term outcomes. This was not employee training or a test of static feedback. It does not demonstrate this course's learning effect.
Revisit when: Revisit when studies test comparable workplace tasks, longer-term outcomes, or different assistance designs.