Instructor notes and reference answers are hidden in the learner view.
FAST-09 Lab Guide: AI for Managers and Leaders
Three labs, one per session. Each lab produces a worksheet that feeds the final 90-day adoption plan. Labs run in the employer's approved tool (Microsoft 365 Copilot, ChatGPT Business/Enterprise, Gemini for Workspace, or Claude for Work). If there is no approved tool, use the provider sandbox with the fictional material below, or do the work on paper.
0. Rules for every lab
- Fictional or sanitized material only. No employee records, reviews, discipline notes, salaries, medical information, customer personal information (G.L. c. 93H, 201 CMR 17.00), bids, or unreleased financials.
- Sanitize before you paste. Replace names with roles ("a technician"), remove ID numbers, dates of birth, and anything that points to one person.
- Check everything you keep. Facts, numbers, dates, commitments, tone, fairness.
- You own the output. The tool drafts; you decide and sign.
Case companies used in the labs (all fictional): North Shore Builders (general contractor, Essex County, 38 employees), Pioneer Valley Mutual (insurance agency, Hampden County, 22 employees), Bayside Precision Parts (CNC machine shop, Worcester County, 64 employees), Cape Harbor Inn (seasonal inn, Barnstable County, 30 staff in season).
Lab 1 (Session 1): Own-use outputs with review notes
Objective: O1. Use AI for three management tasks and judge each output for accuracy, fairness, and tone. Time: 60 minutes (3 × 15 minutes of tasks, 15 minutes pair review). Deliverable: W1 Own-use review notes.
Task A: Team update from your notes (North Shore Builders)
You are the operations manager. Your notes (fictional):
- Starting Monday Nov 16: all site leads submit daily logs in the field app by 5 pm, not by text photo.
- Reason: lost logs on Hollis St job; owner asked for photos of every pour.
- Training: 30 min, Thursday Nov 12, 7 am, Danvers yard trailer.
- Questions: ops manager.
Prompt the tool to draft a short team update for 9 site leads. Then review it.
Task B: Vendor proposal summary (Pioneer Valley Mutual)
Paste the fictional vendor excerpt below and ask for a one-paragraph summary plus a list of price, term, renewal, and data-handling points.
ClaimDraft Pro (fictional) proposal excerpt. Pricing: $34 per user per month, billed annually, minimum 5 users.
Term: 24 months. Auto-renews for successive 12-month terms unless cancelled in writing 90 days before renewal.
Price may increase up to 9% at each renewal. Customer content may be used to improve vendor services unless
the customer opts out in the admin console. Data stored in U.S. data centers. Support: business hours, email only.
Task C: Talking points for a hard conversation (generic)
Without names or personal details, ask for talking points for this situation: "A long-time team member has missed three deadlines this month; the work is usually excellent; I want to understand what is going on and agree on next steps." Ask for an opening, two open questions, and a close.
W1 template
| Task | What I kept | What I fixed | What I discarded | Time with AI vs. without (estimate) | Fairness or tone issue? |
|---|---|---|---|---|---|
| A | |||||
| B | |||||
| C |
Lab 2 (Session 2): Use-case matrix and ROI worksheet
Objectives: O2, O3. Time: 75 minutes. Deliverables: W2 Use-case matrix, W3 ROI worksheet.
Part 1: Practice case (Pioneer Valley Mutual, 20 minutes)
Score each candidate 1 (low) to 5 (high). Priority = Value − Risk − Data sensitivity − (Effort ÷ 2); then sanity-check with judgment.
| # | Candidate use case | Value | Risk | Data | Effort |
|---|---|---|---|---|---|
| 1 | First-pass claim file summaries for adjusters (synthetic files in pilot; approved tool) | ||||
| 2 | Plain-language rewrite of standard policy-change letters (template text, no customer data) | ||||
| 3 | AI decides which claims to deny | ||||
| 4 | Summarizing public Division of Insurance bulletins for staff | ||||
| 5 | Chatbot on the website answering coverage questions directly to customers | ||||
| 6 | Drafting internal meeting agendas | ||||
| 7 | Uploading customer Social Security numbers to a free tool to "clean the list" | ||||
| 8 | Drafting renewal reminder email templates (merge fields filled in the agency system) |
Part 2: Your team (35 minutes)
List 8 to 10 candidate use cases for your own team (sanitized), score them on W2, place them on the value-versus-risk grid, and pick two to pilot. A peer challenges your top two: "What is the weakest score?"
Part 3: ROI worksheet (20 minutes)
Worked example (fictional, Pioneer Valley Mutual, use case 2):
| Line | Value |
|---|---|
| Letters per week (from 2 weeks of logs) | 40 |
| Minutes per letter now | 12 |
| Minutes per letter with AI draft, including checking | 5 |
| Hours saved per week | 40 × 7 ÷ 60 = 4.67 |
| Loaded hourly cost | $46.89 (BLS ECEC, private industry, June 2026; use payroll if known) |
| Working weeks | 48 |
| Gross annual saving | (280 ÷ 60) × $46.89 × 48 = $10,503 |
| Tool cost (already in suite) | $0 added |
| Training: 3 staff × 2 h × $46.89 | $281 |
| Maintenance: 1 h/month template review × $46.89 × 12 | $563 |
| Net first-year estimate | $9,659 |
| Weakest assumption | 5 minutes per letter including checking; test in the first 30 days |
Check your own math with a peer. Every assumption needs a source: log, pilot, payroll, or "estimate, to be tested."
Meets on W2/W3: at least 8 candidates scored; two picks with value and risk reasons; correct arithmetic; checking time included; all assumptions stated with sources; weakest assumption named.
Lab 3 (Session 3): Team guideline and 90-day plan
Objectives: O4, O5. Time: 75 minutes. Deliverables: W4 Team AI use guideline, Final 90-day adoption plan (draft, revised after peer review).
Part 1: Fix the weak guideline (15 minutes)
This draft (fictional, Bayside Precision Parts quality team) has at least six problems. Find them.
QUALITY TEAM AI RULES
1. Use whatever AI tools help you work faster.
2. Don't share anything secret.
3. AI can write inspection reports; just skim them.
4. Customer drawings are fine to upload if it helps with work instructions.
5. The manager will review everyone's AI chat history monthly.
6. If something goes wrong, the person responsible will be disciplined.
Part 2: Write your W4 guideline (25 minutes)
One page, these headings: Approved tools · Never allowed (data) · Encouraged / Allowed with a check / Off limits (tasks) · Who checks what · Reporting mistakes · Questions and review date. Check it against your employer's policy; where you don't know the policy, write "confirm with [role]."
Part 3: Draft the 90-day plan (25 minutes) and peer review (10 minutes)
| Phase | Milestones | Training | Measures (baseline → target) | Communication | Owner |
|---|---|---|---|---|---|
| Days 1-30 | Guideline published; pilot 1 starts | ||||
| Days 31-60 | Pilot 1 measured and adjusted; pilot 2 starts | ||||
| Days 61-90 | Results reviewed; keep, change, or stop |
Part 3 example (fictional, Cape Harbor Inn front desk)
- Days 1-30: Guideline posted Nov 2 (owner: front-desk manager). 45-minute training Nov 5. Pilot 1: AI drafts replies to common guest emails (directions, parking, pet policy) from an approved FAQ; staff edit and send. Baseline: 2 weeks of logs show 6.5 h/week on these replies.
- Days 31-60: Measure weekly team minutes and errors caught in review; target 3.5 h/week with zero wrong policy statements sent. Add pilot 2: shift-handover summaries from the desk log (no guest personal data).
- Days 61-90: Review with staff. Decide keep, change, or stop for each pilot. Update the guideline.
- Communication: kickoff in the Tuesday huddle (manager), weekly two-line update in the team chat, open Q&A at week 6, results shared at day 90. Staff ideas for pilot 3 collected all along.
- Measures stay team-level; no individual monitoring.
Meets on the final plan: milestones in all three phases; training before use; a baseline and target for each pilot; a communication plan with who, when, and how questions are handled; an owner for each step; consistent with the W4 guideline.