FAST-16 Lab Guide: Working Well with AI

Provider: 7formations | Course: FAST-16 | Instructor: Wadood Chaudhary | Three labs (about 45 minutes each), one per session Audience: trainees (steps, worksheets) and instructors (setup, answer notes, timing). Instructor-only notes are marked Instructor.

All companies, people, numbers, and documents in this guide are fictional, except the two public court and tribunal cases in Lab 3, which are cited. Labs run in the employer's approved assistant (Microsoft 365 Copilot Chat, ChatGPT Business or Enterprise, Gemini for Google Workspace, or Claude for Work) or the 7formations sandbox. Every lab has a paper version with printed AI outputs, because what is scored is your judgment, not the tool.

Suggested reading: Laurence Lars Svekis, AI Literacy for All Employees: Understanding the Role of Artificial Intelligence in the Modern Workplace (Kindle edition, 2026), chapters 6, 7, 13, and 15. The labs are original to 7formations.

0. Lab environment and safety rules (read before Lab 1)

Tools

• Sign in to the approved assistant with your work account. If none is approved, use the 7formations sandbox account (fictional data only, spending-capped, disabled at the end of the cohort).
• Outputs vary from run to run and between tools. That is expected and part of the lesson.

Data

• Fictional data only. Never paste a Massachusetts resident's name combined with a Social Security number, driver's license or state ID number, or financial account or card number (G.L. c. 93H, 201 CMR 17.00), health information, passwords, or confidential business data.
• When you use your own work in Lab 1 step 5 or Lab 3 step 5, describe the task in general terms and use placeholders such as [CUSTOMER].
• If you paste something you shouldn't have, stop, tell the instructor, and do not "ask the AI to forget." Report it, the same as at work.

Fictional organizations

• Cape Harbor Inn: a 40-room inn in Chatham, MA.
• Merrimack Freight & Logistics: a regional trucking and warehousing firm in Lowell, MA.
• Bayside Precision Parts: a CNC machine shop in Worcester County, MA.
• North Shore Builders (Beverly, MA) and Lowell BioSupply (lab supplies distributor), used in Lab 3 case files.

Completion. W1 to W3 are scored complete / incomplete against the criteria at the end of each lab. W4 is the final and is scored Not yet / Meets / Strong (Appendix B). One resubmission is allowed.

Lab 1 (Session 1): Prompt makeover

Scenario. Cape Harbor Inn's front-desk lead tried AI for three tasks and got generic, unusable answers. You will rebuild the prompts with the four elements and improve the answers with follow-ups.

Goal. Complete W1 Prompt makeover log: three weak prompts rewritten with context, instruction, structure, and audience, each improved with at least two follow-ups.

Data. Cape Harbor Inn fact sheet (fictional): 40 rooms; check-in 4 p.m., check-out 11 a.m.; shuttle to the Hyannis ferry at 8:15 a.m. and 1:30 p.m. (reserve at the desk by 9 p.m. the night before); dogs up to 40 lb allowed in garden rooms only, $35 per night; no cats; parking: 24 spaces on site; from June 1, overflow parking at the town lot on Main Street, a 4-minute walk, free with a pass from the desk.

The three weak prompts

1. "Write about parking."
2. "Make a check-in guide."
3. "Reply to a guest who wants to bring pets."

Steps

1. Diagnose (5 min). For each weak prompt, mark which of the four elements is missing: context, specific instruction, structure, audience.
2. Rewrite (15 min). Rewrite all three. Paste only the relevant lines of the fact sheet. Add a gap rule: "If the fact sheet does not cover something, write [CHECK]."
3. Run and compare (8 min). Run the weak and strong versions of prompt 3. Note the differences in accuracy, length, and tone.
4. Follow-up ladder (10 min). Improve one answer with at least two follow-ups, using any two of these techniques: step-by-step ("first list the rules, then write the reply"), iterative ("keep the facts, cut to three sentences"), role ("as a friendly front-desk lead"), constraint ("under 80 words, no promises beyond the fact sheet").
5. Your own task (7 min). Write a four-element prompt for one routine task from your own job, in general terms, with placeholders.

Deliverable: W1 Prompt makeover log

Completion criteria. Three rewrites, each with all four elements and a gap rule; at least two follow-ups on one answer with a note on what improved; one prompt for your own task with no restricted data.

Lab 2 (Session 2): Error hunt and evaluation record

Scenario. Merrimack Freight & Logistics' operations manager asked an AI assistant to write a one-page on-time delivery report for March from a fact sheet. The report reads well. It also contains five seeded errors, one of each type, and one invented citation. You will find them, verify them, and record your evaluation.

Goal. Complete W2 Evaluation record: find at least four of the five seeded errors and the invented citation, name each error type, and record the verification technique you used.

Data. Merrimack March fact sheet (fictional): 1,240 deliveries; 1,091 on time (88%); Route 9 corridor 102 deliveries, 84 on time; the depot was closed March 3 and 4 for snow; two drivers were out on leave; February on-time rate 86%. The printed AI report (instructor copy has the answer key).

Steps

1. Read once for meaning (5 min). Read the report as its intended reader would. Note anything that "feels off."
2. Five-step evaluation (15 min). Work through the report in order: accuracy (check every number and name against the fact sheet), completeness (what is missing?), relevance (does it answer the request?), bias (any unfair or one-sided claims?), judgment (would you send it?).
3. Name the error types (5 min). Label each problem: factual error, incomplete information, context misread, logical inconsistency, or bias.
4. Verify the citation (8 min). The report cites "Northeast Freight Review, May 2025, p. 14" for an industry on-time average of 94%. Use SIFT: Stop; Investigate the source; Find better coverage; Trace the claim to the original. Search for the publication. Do not ask the AI whether it is real.
5. Fix and rerun (7 min). Rewrite the prompt to prevent the errors (paste the fact sheet; "use only these facts"; "if a figure is not in the facts, write NOT IN SOURCE"; "no outside statistics"). Rerun and check again.
6. Debrief (5 min). Which step caught the most? Which error would have reached the customer?

Deliverable: W2 Evaluation record

Plus: citation check result, revised prompt, and a keep / fix / discard decision with one line of reasoning.

Completion criteria. At least four of five seeded errors found and correctly typed; the invented citation identified with how you checked; a revised prompt; a decision with reasoning.

Lab 3 (Session 3): Case clinic and safe AI workflow card (final)

Scenario. Five Massachusetts employers (fictional, adapted from the kinds of cases in Svekis, ch. 13) and two public incidents show what goes right and wrong when AI meets real work. You will analyze one case in a group, then write your own safe workflow for a weekly task.

Goal. Complete W3 Case analysis for one case and W4 Safe AI workflow card (final) for one recurring task in your role.

Case files (one page each)

Steps

1. Case clinic (15 min). In groups of three, take one case (A to E) and one public incident. Answer: What was the benefit? What went wrong? Which success factor was missing (clear goals, trained employees, transparency, monitoring)? Which single safeguard would most likely have changed the outcome?
2. Share (8 min). Each group gives a 90-second readout.
3. Draft your workflow (12 min). Choose one task you do every week that AI could help with. Fill in W4 using the five steps: define (the task and whether AI fits), instruct (prompt and safe inputs), review (five-step evaluation), judge (would I put my name on it?), document (an AI-use note where required).
4. Peer test (7 min). A partner reads your card and tries to follow it for your task. They mark any step they could not follow. Fix it.
5. AI-use note (3 min). Write a one-line note for the work product, for example: "Drafted with [approved tool]; facts checked against the order system by [role]."

Deliverable: W3 Case analysis

Deliverable: W4 Safe AI workflow card (final)

Completion criteria. W3: all columns filled with a specific safeguard. W4: see Appendix B.

Appendix B: Rubric for W4 (Not yet / Meets / Strong)