People, practice
& possibility.

Find what needs your attention. Turn employee feedback into better questions. Practice senior HR judgment through short conversations with Copilot—not one enormous prompt.

Before you begin

1. Open your work Copilot. Sign in to Microsoft 365 Copilot Chat with your work account. With the appropriate license and tenant settings, turn Work IQ on so Copilot can ground responses in accessible email, Teams discussions, meetings and files.
2. Choose safe work for weekly closeout. Pick a date range and an HR portfolio such as leadership development, workforce planning or learning delivery. Use approved project-level communications, not employee cases. Access alone is not permission: follow your organization's HR-data and AI policies.
3. Download the shared survey workbook. Use the FY27 Q2 QC Details demo for the second live lab. It contains 50 fictitious employee records and two managers. This is the lead dataset for the exercise, not a request to upload real reviews.
4. Point to sources. Use Add work content or type / to select relevant files or meetings where available. Give a project name and date range for email or Teams context. If retrieval fails, select a specific approved source or provide a permitted excerpt; never assume Copilot searched everything.
5. Run one prompt, then respond. Check a source or workbook row, correct a gap and make a choice before continuing. Replace bracketed text. Start a new chat between labs and reattach the workbook when changing surfaces; app history does not transfer automatically.

Choose the right surface.

Chat helps find the communication signal and explore written feedback. Copilot in Excel helps inspect rows, counts and rating comparisons. Analyst helps compare patterns and produce an evidence-backed summary. Open Analyst under Agents in Microsoft 365 Copilot and attach the demo workbook. Carry only reviewed notes between apps; chat history and access to original sources do not transfer automatically.

Availability gate: before delivery, confirm work-data grounding, Copilot in Excel and Analyst are generally available, appropriately licensed and enabled for these attendees. Basic chat alone does not guarantee mailbox-wide retrieval or agent access. If Analyst is unavailable, use the attached-file prompts in Chat and verify numbers in Excel; do not describe that as an Analyst run. If Excel Copilot is missing, use ordinary filters/PivotTables to check counts. No preview features are required. Later labs use Word and optionally Outlook where enabled.

Four ingredients. Across a conversation.

You do not need to repeat everything in every turn. Establish the context, ask a focused question, then steer the next result. Longer synthesis prompts are optional finishers—not substitutes for the interaction.

Find the signal.
Close the loop.

End the week with a clear view of unresolved asks, missed context and decisions that need your HR leadership—not just a tidier inbox.

1. Catch up on what matters Copilot Chat

Start a new work chat with Work IQ on where enabled. Add relevant work references with / or Add work content, then replace the dates, time zone and portfolio.

Help me close out [start date–end date, time zone] for [HR portfolio]. Review my accessible email, Teams discussions and referenced meetings. Surface five developments or requests I may need to address: what changed, the ask or decision, stated owner and due date, and source link. Distinguish explicit commitments from possible follow-ups. Exclude restricted personnel matters. Flag retrieval gaps; do not claim this is a complete inbox audit.

Your turn: open two cited threads and check their latest messages. Correct one finding or add a missing reference. “Not surfaced” does not prove that a message was read, ignored or never sent.

2. Find the loose ends Same chat

Which asks appear unanswered in these sources, which decisions remain open, and which items were resolved by a later message? Show the latest supporting date and source. Separate requests to me from dependencies on others. Mark absent owners or deadlines Not specified. Do not infer someone ignored a message. Ask me one question about the highest-impact uncertainty, then wait.

Your turn: answer the question and confirm whether the apparent gap still exists. Check Outlook or Teams yourself if the latest reply is missing. Remove resolved items from the proposed follow-up list.

3. Decide what deserves action Same chat

Using our verified findings, propose three priorities: address before closing the week, clarify with a partner, or carry into next week. Explain the impact on leader readiness, workforce plans or program delivery. Keep source links and distinguish stated deadlines from suggested checkpoints. Separate my actions from others' actions. Do not invent commitments, send messages or create tasks.

Your turn: change the priority of one item and explain why. Select one action you can responsibly take; confirm the recipient, audience and facts before drafting.

4. Prepare one follow-up Copilot in Outlook

Open the selected source thread in Outlook and choose a reply draft. Open Copilot where enabled; the thread supplies the context for the prompt below. If unavailable, add the selected thread as a work reference or permitted excerpt in Chat, then manually transfer the reviewed reply. For synthetic practice, stay in Chat and select one fallback thread. Earlier chat context is not assumed to carry into Outlook.

Draft a concise reply to the latest unresolved request or decision in this email thread, using only the thread's context. Address the request and clarify the next step. If my intended action is unclear, ask me one question before drafting. Use a clear, collaborative senior HR tone. Distinguish proposed actions or dates from agreed commitments. Invent no facts or agreement; omit sensitive personnel details.

Your turn: compare the draft with the latest thread, edit its tone and check recipients. Leave it unsent during the exercise. Sending any real communication is a separate human decision.

Create a private weekly HR closeout brief using only the findings I verified and the priorities I selected in this chat. Keep it under 300 words.

Lead with the three issues that most need my attention. For each, state the business or people-program impact, the confirmed ask, the stated owner and due date, and the source. Separate resolved items, actions I accepted, dependencies on partners and proposed checkpoints for next week. Preserve Not specified and Needs validation labels wherever facts remain unclear.

End with two questions I should revisit next week and a short list of source gaps. Do not imply that you searched every message or know what I read. Do not send, schedule, create tasks or make commitments for anyone.

Reflect: Which follow-up changed what you would actually do before leaving for the week?

Listen closely.
Test the pattern.

What are employees asking for, and how are leaders responding? Use a shared fictitious quarterly-conversation dataset to move from comments to a coaching discussion—not a ranking of people.

1. Understand what is in the file Copilot Chat

Use only the attached fictitious FY27 Q2 workbook. Inspect Sheet1 / QC_Details. Confirm the employee count by Manager, review dates and available fields for comments, ratings, development items and tenure. Identify blanks and data-quality limitations. Use Worker aliases and worksheet rows for evidence. Do not treat missing responses as negative feedback or combine leadership and performance ratings. Do not analyze real work content.

Your turn: verify 50 data rows, excluding the header, and the 30/20 manager split. Find the support-response pair V/W and the Leader Insight column X. Ask Copilot to correct any mistaken field mapping before continuing.

2. Read the leader's feedback Same Chat

Summarize Manager 2's comments in Leader Insight : Manager, column X. What are the top three things this manager is asking the team to work on? Distinguish praise from requests. Support each theme with short excerpts, Worker aliases and worksheet rows. Count distinct employees per theme, allowing overlapping themes; show the denominator and blank comments. Do not infer manager effectiveness from this alone.

Your turn: read two cited cells in X. Challenge one theme: “Is this an actual request, or your interpretation?” Choose one follow-up from the question bank below if time permits; do not run every prompt in person.

3. Verify the operational signals Copilot in Excel

Open the original demo's working copy in Excel and select a cell inside QC_Details. Ask in the Copilot pane. If workbook editing is available, put formulas or PivotTables on a new analysis sheet; otherwise inspect the response and verify with ordinary filters.

Using QC_Details, count workers by Manager and exact Status. Separately, count Current Development Items - In-Progress = No by Manager, showing each team's denominator and percentage. Keep blanks separate from No. Show reproducible formulas or PivotTable steps and source ranges. Put any analysis on a new sheet; do not change the source table. Verify the totals reconcile to 50 workers.

Your turn: filter S to No: Manager 1 has 1 of 30 (3.3%); Manager 2 has 11 of 20 (55%). Clear the filter and check F separately: 28 and 16 Successfully Completed. A review status is not an employee performance rating.

4. Turn a pattern into a question Analyst

Start a new Analyst conversation and attach the same unedited demo workbook. Paste the verified counts if helpful, but have Analyst check them against the file. If this agent is unavailable, use Chat and label the route accurately; independently verify all numbers in Excel.

Analyze employee support requests in column V of the attached fictitious workbook by Manager. Define a small theme codebook, count distinct employees per theme and show team denominators, blanks and representative Worker aliases/rows. Allow multiple themes per response; explain why percentages may exceed 100%. Compare the teams without ranking managers. Separate observed requests from your interpretation and recommend one question HR should validate with each manager.

Your turn: inspect the theme definitions and two supporting rows. Ask Analyst to explain one coding choice or recalculate an unclear denominator. Workload/exposure and career/mentoring/feedback are hypotheses to check, not answers to force onto the data.

Compare employee goals in T with manager responses in U, by Manager. First define a transparent rubric for goal clarity: observable action, measure and timing. Show examples of vague goals and how each manager responds: adds specificity, acknowledges only, or leaves no recorded reply. Include Worker aliases and rows. Separate missing goals from vague goals. Treat your coding as a reviewable judgment, not a validated assessment.

Pair employee support requests in V with manager replies in W. By Manager, distinguish no employee request, request with blank reply, partial response and response addressing the request. Define your coding and cite Worker aliases/rows for ambiguous cases. Count blank replies separately from substantive gaps. A blank cell means no recorded reply here, not proof support was withheld. Ask me to review one ambiguous pair.

Across all 50 fictitious workers, summarize recurring accomplishments and learning themes in employee responses J. Use K only as separately attributed manager context. Count distinct employees per theme, show blanks and denominators, and give representative Worker aliases and rows. Distinguish reported accomplishments from independently verified outcomes. Do not add together unrelated project metrics or let repeated wording imply greater impact.

In the attached workbook, compare Length of Service AJ and Time in Position AK by Manager, keeping them separate. Show counts and denominators for each band. Explore descriptive patterns with leadership ratings, performance ratings and goal clarity in T; explain any clarity rubric. Do not infer causation, convert bands to exact months or average rating labels. Flag small groups and missing data before suggesting questions for HR.

From QC_Details, list fictitious workers whose Time in Position AK is 0-3 Months or 3-6 Months. Group by Manager and show Worker alias, exact band, employee support request V and worksheet row. Do not use Length of Service AJ as a substitute. Explain that these bands approximate the under-six-month question; exact dates and the six-month boundary cannot be resolved from this file.

In QC_Details, list reviews whose Status F is not Successfully Completed, grouped by Manager and exact status. Include Worker alias and worksheet row, then reconcile the list with status counts. Do not collapse Manually Advanced and Awaiting Manager Completion into the same explanation. Identify the recorded process step, not a judgment about employee performance; flag blanks separately. Preserve the source table.

Compare employee versus manager ratings in QC_Details separately for Leadership L/M and Performance P/Q. For this demo, order labels Leading, On-track, Off-track from highest to lowest. By Manager, count same, self higher, self lower and missing-pair cases. Show denominators and Worker aliases/rows for differences. Use formulas or reproducible steps; no numeric rating averages. Preserve QC_Details and put outputs on a new sheet.

Using only the fictitious workbook, list workers whose self-rating is higher than the manager rating, grouped by Manager and rating dimension. For this demo use Leading above On-track above Off-track; exclude missing pairs. Show both labels and summarize the corresponding manager comment: O for Leadership, R for Performance. Cite Worker alias and row. Mark blank comments No recorded explanation; do not invent a reason for disagreement.

Using Manager 2's employee goals T and manager replies U in this fictitious workbook, draft 1:1 talking points to help employees set measurable goals. Use two cited examples, an open question and a proposed rewrite for each. Label every new target or date as proposed for agreement. Keep the tone supportive. Do not diagnose employees, assume motives or turn examples into ratings or employment recommendations.

Create a one-page HR leadership summary from this fictitious workbook and the findings I verified. Compare the two teams without ranking managers or recommending employment actions.

Include team sizes, review-completion status, development-item gaps, employee support themes and three proposed coaching priorities for each manager next quarter. For every priority, show the supporting field/rows, a question to validate, a proposed next step and a proposed checkpoint—not an agreed commitment.

Include a labeled chart of manager-rated Performance categories from Q by Manager, showing counts and team percentages with missing ratings separate. Keep Leadership ratings out of this chart; do not average ordinal labels.

Retain denominators, uncertainty and source references. Note unequal team sizes, Manager 2's newer-in-role mix, missing comments and this single synthetic snapshot. Separate observations, interpretations and proposed actions. Do not claim causal effects or publish the output.

Reflect: Which finding survived a row-level check—and what would you ask before acting on it?

Better prepared.
Still accountable.

Close the week with clearer priorities. Bring better questions to the feedback discussion. The human who acts on the output still owns the decision.

The prompting
pattern.

Self-paced practice · not part of the two live exercises. Back to session close

Build an executive-ready ask across a conversation: goal, context, source and expectations. Help your CHRO or leadership team make a better decision—not simply read a shorter update.

1. State the goal Copilot Chat

Help me frame an executive discussion about [HR initiative]. The decision I need to support is [decision]. Restate the goal in one sentence and identify what a useful answer would help the leadership team decide. Do not research, infer results or draft the brief yet.

Your turn: correct the decision statement. If you cannot name the decision yet, say so and ask for one clarifying question before continuing.

2. Add leadership context Same chat

The audience is [CHRO, HR leadership team or cross-functional steering group]. We have [15 minutes] and need to balance [business outcome], [delivery capacity] and [employee experience]. The planning period is [dates]. What one ambiguity in this context would most change the decision brief? Ask me that question and wait.

Your turn: answer the question. Explain one constraint that a senior HR leader must own rather than ask AI to resolve.

3. Bound the sources Same chat · Work IQ

Add the approved initiative references using / or Add work content where available. Narrow the date range and read the source links; do not assume broad retrieval is complete.

Use only the initiative references I added and related program-level email from [start date–end date] where I am sender or recipient. Find three developments relevant to our decision, with dates and sources. Prioritize the latest supported information. Separate confirmed facts, interpretations and unresolved issues. Exclude restricted personnel matters. Flag missing access or conflicting evidence; do not fill gaps with assumptions.

Your turn: open two sources and correct an outdated or overstated finding. With the fallback, H02 supersedes H03's preliminary completion count; participation is not proof of performance improvement.

4. Set the output expectations Same chat

Using only our verified findings, write a discussion opener under 200 words: decision needed, three source-backed points, two risks or tradeoffs and one proposed next step. Use a direct, neutral senior HR tone. Preserve Needs validation labels. Separate recommendations from approved commitments. End with the most important question for the leadership team; do not decide, share or act for us.

Your turn: identify where goal, context, source and expectations entered the conversation. Revise one element and compare the result—not just the writing style.

Help me prepare a leadership discussion about [HR initiative] and the decision [decision needed]. The audience is [authorized leadership group], the discussion lasts [minutes], and the planning period is [dates]. Consider delivery capacity, business outcomes and employee experience without assuming a tradeoff has been resolved.

Use only [selected approved references] and the corrections I verified. Explain why this initiative merits leadership attention, then provide three evidence-backed points, two risks or alternative interpretations, and a proposed next action. Cite the original sources and retain all Needs validation labels.

Keep the discussion opener under 200 words, direct and neutral. Distinguish observations, recommendations and agreed commitments. Exclude restricted personnel information. End with one question that would improve the decision. Do not send, publish, assign work or make an employment decision.

Enter informed.
Ask better questions.

Self-paced practice · not part of the two live exercises. Back to session close

Prepare for a people-strategy or portfolio steering discussion with a clearer view of the decision, stakeholder input and assumptions—not a generic meeting agenda.

1. Establish the meeting context Copilot Chat

Using the upcoming meeting reference and approved papers I added, summarize its purpose, recent HR program developments and decisions that appear needed. Cite dates and sources. Distinguish the stated agenda from your interpretation of what leadership should resolve. Flag missing papers and outdated context. Use only this program-level scope, not individual personnel records.

Your turn: compare the result with the invitation and latest paper. Correct the decision question or supply one missing approved reference.

2. Check participant input Same chat

Which participants' input is important to this decision, and what relevant questions or positions are explicitly supported by our sources? Cite the message or paper behind each statement. Separate documented views from suggested questions for participants. Do not infer agreement, opposition, motives or decision authority from job title, silence or attendance. Mark anything unsupported Unknown.

Your turn: open one cited statement. A request for a cost check is not proof of opposition. Decide what you should confirm directly rather than assume.

3. Improve the leadership questions Same chat

Propose three questions I should ask as a senior HR leader: one about business value, one about delivery capacity and one about employee experience or equitable access to the program. Identify two assumptions the group should test and the evidence needed. Avoid causal claims and employee-level judgments. Ask which question would most improve the decision, then wait.

Your turn: choose and rewrite the most useful question. Share only the question's structure—not real meeting content—if comparing your approach with a colleague.

4. Build a private preparation note Same chat → your meeting notes

Create my private preparation note under 250 words using the context and questions I verified. Include the decision needed, three source-backed developments, documented participant input, my chosen questions and the two assumptions to test. Preserve Unknown and Needs validation labels. Separate my proposed opening from any agreed agenda. Do not invent a stakeholder position or distribute this note.

Your turn: move only the reviewed note into an approved private location if useful. Confirm every source and decide what, if anything, can be shared with the meeting audience.

Turn evidence
into a decision brief.

Self-paced practice · not part of the two live exercises. Back to session close

Turn a reviewed analysis into a one-page brief a CHRO or executive steering group can use: the decision, evidence, options, recommendation and risks.

1. Build the common handoff Copilot Chat

Start a new chat and supply your reviewed findings and permitted source references, or continue in the analysis conversation that produced them. Do not ask Copilot to fill evidence gaps.

From the findings I verified, create a Decision Brief Handoff under 250 words for [authorized leadership audience]. Include the HR decision question, three findings with source evidence, feasible options, a proposed recommendation, two risks or counterarguments and unresolved questions. Keep numbers, periods and denominators explicit. Mark unsupported items Needs validation. Proposed owners and checkpoints are not commitments. Use only my reviewed material.

Your turn: open the original sources and correct the handoff. Confirm the recommendation addresses the actual decision; remove unsupported impact or ROI claims.

2. Challenge the recommendation Same chat

Argue against the proposed recommendation using only our evidence and clearly labeled possibilities. What alternative explanation, delivery constraint or missing information could change the decision? Distinguish an observed risk from a hypothetical one. Identify the weakest claim and ask me one question before revising the handoff. Do not invent data, assume stakeholder agreement or recommend individual employment actions.

Your turn: answer or retain the gap as Needs validation. Approve a corrected handoff before changing apps; do not carry the entire private analysis by default.

3. Draft the executive brief Copilot in Word

Open a blank Word document in an approved location and paste the reviewed Decision Brief Handoff. Open Copilot in Word where enabled. If unavailable, draft in Chat and manually move the reviewed text into Word. No cross-app memory or original-source access is assumed.

Using only the reviewed Decision Brief Handoff in this document, draft a one-page executive HR decision brief. Lead with Decision Needed, followed by Situation, Evidence, Options, Recommendation and Risks. Use a direct, neutral tone for a CHRO or steering group. Preserve important numbers, source references and every Needs validation label. Distinguish proposed actions from agreed commitments. Add no facts, policy claims or personnel recommendations.

Your turn: compare the draft with the handoff and sources. Confirm the document's actual rendered length; a prompt cannot guarantee one printed page.

4. Refine for the leadership audience Same Word document

Tighten this reviewed brief to one page without changing the meaning. Retain evidence that materially affects the decision, the main alternative and all uncertainty labels. End with a proposed accountable role and next checkpoint, explicitly subject to agreement. Remove inflated claims and unnecessary jargon. List any material evidence you recommend omitting so I can decide; do not silently drop it or share the document.

Your turn: decide what to retain, check page layout and verify that the brief makes the decision clearer—not just shorter. Review audience and document classification before any later sharing.

Consolidate my approved handoff, brief and edits in this document into a one-page HR leadership decision brief. Use only the material I verified. Lead with the decision required, then include Situation, Evidence, Options, Recommendation and Risks.

Preserve the important numbers, time periods, denominators, source references and Needs validation labels. Retain the strongest counterargument and explain what additional evidence could change the recommendation. Separate observed outcomes from interpretations and proposed actions from agreed commitments. Keep any proposed owner and checkpoint explicitly subject to agreement.

After the brief, provide a separate private review checklist of unresolved evidence, audience restrictions and page-layout checks for me to complete. Do not strengthen claims, infer stakeholder agreement, make individual personnel recommendations, publish or distribute the document.

Make the decision
accountable.

Self-paced practice · not part of the two live exercises. Back to session close

Turn a specific past meeting into a reliable decision and owner record, then prepare a concise Outlook follow-up. Discussion is not approval; a proposed date is not a commitment.

1. Extract the decision record Copilot Chat

Using only the specific past-meeting record I provided, build a decision and owner table: decisions made, actions, explicitly named accountable roles, stated due dates, unresolved questions and items discussed but not decided. Cite the transcript passage or note reference for each item. Label the source type. Mark missing owners Unassigned and missing dates Not specified. Do not turn a proposal into an approval.

Your turn: check two entries against the original record. With H05, the pilot stays at four sites pending checks; expansion and mentoring were not approved.

2. Resolve what the record cannot Same chat

Check the table for unsupported approval, ambiguous ownership, missing dates and contradictions within the selected record. Separate explicit commitments from items needing confirmation. Ask me one question about the most consequential ambiguity, then wait. If I suggest an owner or date, label it Proposed until the responsible people agree; do not rewrite it as a meeting commitment.

Your turn: correct the record or leave the gap explicit. Do not make up a due date to complete the table. Confirm the intended audience and remove private discussion that does not belong in a follow-up.

3. Prepare the follow-up draft Same chat

Using only the decision record I verified, draft a follow-up under 200 words for [authorized audience]. Put decisions first, actions with agreed owners and stated dates second, and open questions last. Keep Unassigned, Not specified and Proposed labels. Ask recipients to correct anything that does not match their understanding. Include a clear subject. Exclude restricted personnel details. Do not send, create tasks, schedule or add recipients.

Your turn: check every sentence against the verified record. Confirm that the message asks for correction without implying silence means agreement.

4. Review in Outlook Copilot in Outlook

Manually open an unsent message in approved work Outlook and paste the reviewed draft. Leave recipients blank during the exercise. If drafting Copilot is available, refine with the prompt below; otherwise edit manually. Do not assume Chat opens a draft or carries its source context into Outlook.

Refine this reviewed meeting follow-up for a senior HR leadership audience. Keep it under 200 words and preserve the decisions, source references, agreed owners, stated dates and uncertainty labels exactly. Make the tone concise and collaborative. Retain the request for corrections. Do not add commitments, infer recipients, change a proposal into an agreement, send or schedule anything.

Your turn: recheck the latest record, subject and audience. Leave the message unsent. Sending a real communication and assigning work require a separate human decision.

Accelerate the work.
Own the judgment.

Self-paced practice · not part of the two live exercises. Back to session close

Apply an executive review gate to a real draft. HR leadership accountability includes evidence, employee trust, audience boundaries and the consequences of acting—not just a confident answer.

1. Verify the claim Copilot Chat → original sources

Audit this HR program draft against only the sources I supplied. Separate facts, interpretations, assumptions and recommendations. For each material claim, show its source and whether it is supported, contradicted or not verifiable here. Check numbers, periods, denominators, attribution and implied causation. Preserve uncertainty; a citation is not proof the claim is correct. Do not add personnel information.

Your turn: open two original sources yourself and correct the claim table. If Copilot cannot access a source, treat that as a review gap, not verification.

2. Challenge the conclusion Same chat

What is the strongest counterargument to this draft's proposed action? Identify an alternative explanation, a missing perspective and a foreseeable employee-experience or implementation risk. Label hypothetical concerns as hypotheses, not evidence. Explain what information could change the recommendation without inferring protected traits, motives or individual performance. Ask which concern I want to test, then wait.

Your turn: choose a concern and either find permitted evidence or retain it as an open question. Do not use AI-generated possibilities as new facts.

3. Protect the audience boundary Same chat · human policy check

Review this draft for the intended audience [authorized audience]. Flag unnecessary identifiable details, small-group exposure, private source references and language that could imply a personnel judgment or unapproved policy. Suggest program-level wording that preserves material uncertainty. List permissions, classification and policy checks I must make myself. Do not claim that you verified access rights, achieved anonymity or certified compliance.

Your turn: apply your organization's HR, privacy and sharing rules. Check permissions in the original apps. Editing names or asking Copilot for a privacy check does not guarantee de-identification or authorization.

4. Own the next move Same chat → your decision

Using only the corrections and review decisions I explicitly made, create a private review note: claims I verified, claims I removed, unresolved evidence, audience restrictions and the decision still required from the accountable human. Keep proposed actions and checkpoints labeled Proposed. Ask me to choose Revise, Hold or Ready for human approval and explain why. Do not approve, distribute or act on the draft.

Your turn: make and record your choice. Copilot's note is not sign-off. Resolve any material gaps and obtain the required human approval before a real communication or decision.

Your next move: choose one recurring HR leadership workflow. Redesign it as catch up, prepare, analyze, decide, create and follow through. Where could Copilot improve the quality of a decision—not just its speed?