Instructor notes and reference answers are hidden in the learner view.
FAST-04 Lab Guide: AI Productivity with Microsoft Copilot and Google Gemini
Provider: 7formations | Course: FAST-04 | Instructor: Wadood Chaudhary | Four labs, one per session (Lab 4 has parts A, B, and C, covering the course design's Labs 4 and 5) Audience: trainees (steps, worksheets) and instructors (setup, answer notes, timing). Instructor-only notes are marked Instructor.
All companies, people, messages, files, and figures in this guide are fictional. Labs run in the 7formations sandbox tenant (Microsoft 365 or Google Workspace, matching the employer's suite) with a fictional mailbox, a recorded fictional meeting, and fictional files. Nothing connects to an employer's tenant or uses real customer, patient, or employee data.
Naming note (October 2026). Microsoft now brands its work assistant Microsoft Copilot (formerly Microsoft 365 Copilot); the in-app features are Copilot in Outlook, Teams, Word, PowerPoint, and Excel. Google's are Gemini in Gmail, Meet ("Take notes for me"), Docs, Slides, and Sheets. Menus and feature names change often. The instructor confirms current screens in the sandbox before each cohort.
0. Lab environment and safety rules (read before Lab 1)
Tools
- Microsoft track: sandbox tenant with Microsoft Copilot licenses, Outlook, Teams, Word, PowerPoint, Excel, SharePoint/OneDrive.
- Google track: sandbox Workspace with Gemini enabled in Gmail, Meet, Docs, Slides, Sheets, Drive.
- The instructor runs the employer's suite only and drops the other track's steps.
- Paper fallback: printed AI outputs for every lab. What is scored is the check, not the tool.
Data
- The assistants see what you can see: mail, chats, meetings, and files shared with you. That's why the sandbox exists.
- Never paste real customer personal information (G.L. c. 93H, 201 CMR 17.00), PHI, confidential pricing, bids, HR files, or passwords. Class uses fictional data only.
- Recording and transcription: follow employer policy and make sure everyone in the conversation knows and agrees. Massachusetts restricts secret recording (G.L. c. 272, §99). Training, not legal advice. Labs use a pre-recorded fictional meeting only.
- If something real appears in the sandbox, stop and tell the instructor.
Fictional organization
- Merrimack Freight & Logistics, a 70-person regional trucking and warehousing company in Lowell, MA. Routes: Merrimack Valley, North Shore, Metro West, Southern NH. Services: full truckload (FTL) and less-than-truckload (LTL). People: Dana (dock supervisor), Priya (dispatch lead), Luis (warehouse lead), Renee (ops manager).
Completion
- Submit W1 to W5. Completion requires Meets on every rubric criterion (Appendix B), including the live demo of two playbook tasks.
Lab 1 (Session 1): Inbox triage in the sandbox mailbox
Scenario. Monday morning at Merrimack Freight. The shared customer-service mailbox has five long shipper threads waiting. You'll triage them with the assistant without sending a single unchecked commitment.
Goal. Complete W1: summary, reply draft, commitments flagged, decision, and time for each of the five longest threads.
Data. Sandbox mailbox (fictional). The three threads used in the debrief:
T1 Granite Ridge Home Goods - load MFL-40817
"Can we move pickup from Monday to Tuesday? Tuesday 8-10 AM or Wednesday
afternoon works. Will the $150 Monday detention charge still apply?
FYI our new receiving contact starts next month."
T2 Lowell BioSupply - pallet count dispute, invoice INV-22931
"Invoice shows 14 pallets; our dock log shows 12. Please correct."
T3 Seacoast Outfitters - weekly LTL, request for rate review
"Volumes are up 20%. Can we discuss a lower LTL rate for Q1?"
Steps
- Summarize (8 min). For each thread: "Summarize: decisions, asks, deadlines, open questions." (Copilot in Outlook: Summarize at the top of the thread. Gemini in Gmail: Summarize this email in the side panel.)
- Check the summary (6 min). Compare with the thread. Does it narrow "Tuesday or Wednesday" to just Tuesday? Does it state the pallet count correctly?
- Draft (10 min). Draft each reply from the checked summary with a prompt that names what you can and can't promise (see activity F04-S1-A3).
- Commitment hunt (8 min). Highlight every price, date, refund, waiver, or guarantee. Remove it or mark it "needs approval: [role]."
- Follow-up (4 min). Ask for one follow-up reminder. Match every date to the thread. Don't auto-send.
- Log (4 min). Complete W1, including time.
Deliverable: W1 Inbox triage log
| Thread | Summary errors found | Commitments flagged | Decision (send / hold / route) | Route to | Time (min) |
|---|---|---|---|---|---|
| T1 | |||||
| T2 | |||||
| T3 | |||||
| T4 | |||||
| T5 |
Completion criteria. Every summary checked against its thread; every commitment flagged; no draft promises fees, discounts, credits, or delivery dates without a named approver; T2 routed to billing (count dispute) and T3 to the account manager (rate review); time logged.
Lab 2 (Session 2): Meeting recap with marked errors
Scenario. Renee recorded the Monday Oct 5 ops meeting (fictional, all participants agreed, per company policy). The assistant produced a recap. Before it goes to the team, every line must be checked.
Goal. Complete W2: every AI error marked and corrected; a corrected recap and task list.
Data. Transcript excerpt (the full recording is in the sandbox):
RENEE: Nashua's been late three weeks running. Do we add a second truck?
PRIYA: Maybe, but let's price it first. I'll get quotes from two carriers
by Wednesday the 7th.
RENEE: OK, no decision until we see quotes.
LUIS: Dock schedule needs updating for the holiday volume.
DANA: I'll take the dock schedule. Thursday.
RENEE: Safety refresher: sign-ups due the 16th. Dana, can you send the link?
DANA: Yes.
RENEE: Last thing: Billerica yard lease is up in January. Let's think about it.
AI recap (as generated, contains errors):
Decisions: Add a second truck to the Nashua route. Freeze overtime for Q4.
Actions: Luis to update the dock schedule by Thursday.
Priya to request carrier quotes by Wed Oct 7.
Dana to send safety refresher sign-ups by Oct 30.
Open questions: Whether to renew the Billerica yard lease.
Steps
- Consent check (3 min). Confirm from the meeting notice that everyone was told and policy allowed it. Note which meetings at your workplace would not be transcribed.
- Generate your own recap (5 min). Use the improved recap prompt (activity F04-S2-A3). (Teams: Copilot in the meeting recap. Meet: Gemini notes doc.)
- Line-by-line check (15 min). Mark each line of the provided AI recap and of your own: correct / wrong owner / wrong date / discussed-not-decided / not in meeting.
- Correct (8 min). Write the corrected recap.
- Tasks (6 min). Convert actions to tasks (Planner/To Do or Google Tasks) with owner, date, and a link to the recap. Note that owners agreed in the meeting.
- Share the most common error (3 min).
Deliverable: W2 Meeting recap error marks
| Recap line | Mark | Evidence (transcript line) | Correction |
|---|---|---|---|
Completion criteria. All five planted errors marked (Nashua decided→discussed; overtime freeze invented; dock schedule owner Luis→Dana; safety date Oct 30→Oct 16; and the Thursday date lacks a calendar date: accept "Thu Oct 8" as the correction); corrected recap contains only verified lines; each task has owner, date, link.
Lab 3 (Session 3): Document-to-deck briefing
Scenario. Renee needs a six-slide Q3 service briefing for the owner's monthly review, plus a version of one section for drivers.
Goal. Complete W3: a grounded draft, one audience rewrite, a six-slide deck, and a check sheet tracing every number, date, and name.
Data. Outline and two sandbox files:
OUTLINE: 1 Headline results 2 Where we were late 3 What customers said
4 What we're changing 5 Asks of drivers 6 Decisions needed
Q3_OnTime_Report (fictional): Q3 loads 450; late 27; on-time 94.0%.
Q2 loads 412; late 37; on-time 91.0%. Late by route (Q3): Nashua 11,
Metro West 7, North Shore 5, Merrimack Valley 4. Main causes: dock
congestion at receiver (12), traffic (9), equipment (6).
Customer_Survey_Summary (fictional, 62 responses): communication 4.1/5;
tracking updates 3.6/5; top request: proactive delay alerts (29 mentions).
Steps
- Draft (10 min). Word/Docs: "Draft sections 1-6 of my outline using only Q3_OnTime_Report and Customer_Survey_Summary. Mark anything not found in them [NOT IN SOURCE]." Section 4 and 6 will mostly be [NOT IN SOURCE]: fill with Renee's notes (provided on paper), not the AI.
- Trace (8 min). Highlight every claim not in the files (expect an industry average or a "best in New England" line). Remove it.
- Rewrite (7 min). Rewrite section 5 for drivers (activity F04-S3-A2). Compare side by side: numbers identical?
- Deck (10 min). PowerPoint: Create a presentation from file. Slides: Create with Gemini or Help me create. Cut to six slides; key points only; replace generic images with the route table or nothing.
- Fact and figure check (10 min). Every number, date, and name on W3.
Deliverable: W3 Document and deck check sheet
| Slide | Item (number, date, name) | Source file and location | Match? | Fix |
|---|---|---|---|---|
Completion criteria. Draft has no untraced claims; rewrite keeps every number; deck is six slides; every figure on the deck appears on W3 with a source and a match.
Lab 4 (Session 4): Spreadsheet Q&A, permissions fix, and playbook demo
Scenario. Renee asks which region dropped the most last quarter. Meanwhile, sandbox search surfaced a file that shouldn't be visible. And it's playbook day.
Goal. Part A: W4 (three questions, one manually verified). Part B: one over-shared file found and corrected. Part C: W5 playbook, two tasks demonstrated live.
Data. Sandbox workbook Revenue_by_Region_Q2Q3.xlsx (Google: Sheet). A filter on Service = LTL is deliberately left on.
| Region | Service | Q2 revenue ($) | Q3 revenue ($) |
|---|---|---|---|
| Merrimack Valley | FTL | 48,200 | 45,900 |
| Merrimack Valley | LTL | 21,400 | 22,100 |
| North Shore | FTL | 36,500 | 37,800 |
| North Shore | LTL | 18,900 | 12,600 |
| Metro West | FTL | 41,000 | 33,700 |
| Metro West | LTL | 15,200 | 16,000 |
| Southern NH | FTL | 22,300 | 23,100 |
| Southern NH | LTL | 9,800 | 9,400 |
Part A: Spreadsheet Q&A (25 min)
- Ask three questions: "Which region's total revenue dropped the most from Q2 to Q3, in dollars?" "Show that as a percentage." "Which service line drove the drop?" Ask it to show the formula or pivot each time.
- Verify the first answer yourself: clear filters, then build a pivot (Region × sum of Q2 and Q3) or SUMIFS, and compute Q3 − Q2.
- If answers differ, find out why (filter? method?) and log it.
Part B: Permissions fix (15 min)
- Search the sandbox for "driver rates."
Driver_Pay_Rates_2026.xlsx(owner: HR) appears from the Ops All-Staff site (Microsoft: accessible to "Everyone except external users"; Google: general access "Anyone in Merrimack Freight"). - Don't download or screenshot contents. Check its sharing.
- Fix: Microsoft: Manage access → remove org-wide access, restrict to the HR group (or move to the HR site). Google: Share → General access: Restricted; add the HR group.
- Tell the owner (sandbox HR mailbox) and log a report per the sandbox policy.
- Search again as the test user to confirm it no longer appears.
Part C: Playbook demo (20 min plus demo slots)
- Finish W5: five recurring tasks from your job (described generally) with steps, prompt, check, and honest time.
- Demonstrate two tasks live to the instructor in the sandbox, including the check.
Deliverable: W4 Spreadsheet verification
| Question | Assistant answer | Formula/pivot shown? | My manual result | Match? | If not, why |
|---|---|---|---|---|---|
Deliverable: W5 Productivity playbook (one row per task)
| Task | Steps | Prompt | Check | Time by hand | Time with AI incl. checking | Keep / adjust / stop |
|---|---|---|---|---|---|---|
Completion criteria. Part A: one assistant answer manually verified, differences explained. Part B: the file's access restricted, owner told, report logged, re-search confirms. Part C: five tasks complete with checking time; two demonstrated correctly.
Appendix B: Rubric (Not yet / Meets / Strong)
| Criterion | Not yet | Meets | Strong |
|---|---|---|---|
| Accuracy checking | Outputs used unchecked or errors missed | Every AI output checked; errors marked and fixed | Errors traced to a prompt change that prevents them |
| Permissions awareness | Over-shared file not found or not fixed | Over-shared file found, restricted, owner told | Re-search verified; trainee lists files to check at work |
| Applied playbook | Fewer than five tasks or no checking time | Five tasks with steps, prompt, check, time incl. checking | Includes a "stop" decision or a measured improvement |
| Live demo | Demo fails or skips the check | Two tasks performed correctly, check shown | Demo includes catching and fixing a live error |