# FAST-15 Lab Guide: AI Foundations for Every Employee

Provider: 7formations  |  Course: FAST-15  |  Instructor: Wadood Chaudhary  |  Three labs (about 40 to 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, systems, and numbers in this guide are fictional. Every lab runs on paper first; where a step uses AI, it runs 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. No lab asks a trainee to enter work data.

Suggested reading: Laurence Lars Svekis, *AI Literacy for All Employees: Understanding the Role of Artificial Intelligence in the Modern Workplace* (Kindle edition, 2026), chapters 1 to 5. The labs are original to 7formations; they practice the ideas in those chapters with Massachusetts examples.

---

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

**Tools**

- This is a concepts course, so most lab work is sorting, explaining, and mapping on paper or a shared whiteboard. Two short steps (Lab 1 step 6 and Lab 2 step 5) use the approved assistant. If no tool is approved, the instructor shows the step on the projector from the sandbox.
- Use your work account in the approved tool. Do not use a personal account for any course step.

**Data**

- Fictional data only. Under Massachusetts G.L. c. 93H and 201 CMR 17.00, a resident's name combined with a Social Security number, driver's license or state ID number, or financial account or card number is personal information. Never paste that combination into any AI tool.
- When you describe your own job or tools, use general terms: "the scheduling feature in our time-clock app" is fine; a screenshot of a coworker's schedule is not.
- If you paste something you shouldn't have, stop and tell the instructor. Practice the same habit you would use at work: report it.

**Fictional organizations in this course**

- **Bayside Precision Parts**: a 60-person CNC machine shop in Worcester County, MA (advanced manufacturing).
- **Merrimack Freight & Logistics**: a regional trucking and warehousing firm in Lowell, MA.
- **Cape Harbor Inn**: a 40-room inn in Chatham, MA (hospitality), used in examples.
- **Pioneer Valley Mutual Insurance Agency**: an independent agency in Springfield, MA, used in examples.

**Completion**

- W1 to W3 are scored complete / incomplete against the criteria at the end of each lab. W4 is the final deliverable and is scored Not yet / Meets / Strong (Appendix B). A "Not yet" can be resubmitted once.

---

## Lab 1 (Session 1): Rules, prediction, or generation?

**Scenario.** Bayside Precision Parts' operations manager has asked for a one-page explainer that helps new hires understand which of the shop's systems "use AI" and which just follow rules. You will classify the shop's 12 systems, then write the explainer card.

**Goal.** Classify 12 systems into rule-based automation, predictive AI, or generative AI with a reason, and produce W1 How-AI-works explainer card.

**Data.** Bayside system list (fictional).

| # | System | What it does |
| --- | --- | --- |
| 1 | ERP reorder alert | Emails purchasing when stock of a part falls below a number set by the buyer |
| 2 | Spindle health monitor | Learned from three years of vibration data; flags spindles likely to fail within 30 days |
| 3 | Email junk filter | Learned from messages users marked as junk |
| 4 | Payroll overtime rule | Pays 1.5 times the regular rate after 40 hours |
| 5 | Copilot Chat | Drafts emails and summaries from a prompt |
| 6 | Quote estimator | Learned from 2,000 past quotes; suggests a price range for a new job |
| 7 | Badge reader | Opens the door for IDs on an approved list |
| 8 | Meeting recap | Writes a summary and action items from a call transcript |
| 9 | Vision inspection camera | Trained on labeled images of good and scratched parts; flags scratches |
| 10 | Shipping label printer | Prints the carrier label from the order's address fields |
| 11 | Customer grouping report | Groups customers by ordering patterns without preset labels |
| 12 | Image tool in the marketing app | Generates a background image for a trade-show flyer |

**Steps**

1. **Warm-up (5 min).** As a class, sort systems 1, 2, and 5 together. Ask of each: *Did a person write the rule, or did the system learn the pattern from data? Does it estimate an outcome, or create new content?*
2. **Classify (12 min).** In pairs, classify the other nine systems. Write a one-line reason for each.
3. **Learning type (6 min).** For each system you labeled predictive or generative, mark whether it learned from labeled examples (supervised), found groups without labels (unsupervised), or learned from rewards after actions (reinforcement). Use "not sure" when the description does not say.
4. **Why it can be wrong (6 min).** Pick two AI systems from your list. For each, name which of the four reasons for error is most likely: limited training data, outdated information, misread context, or bias in the data.
5. **Write the explainer card (10 min).** Complete W1 below for a new Bayside hire. Plain words, no jargon you have not explained.
6. **Explain it back (5 min).** Read your card to a partner. In the approved assistant, ask: "Explain the difference between rule-based software and machine learning to a new machine shop employee in four sentences." Compare its answer with your card. Note one thing it said better and one thing you would correct or add.

**Deliverable: W1 How-AI-works explainer card**

| Section | Your words (40 to 60 words each) |
| --- | --- |
| Rules versus patterns | |
| How AI learns (training) and how it is used (inference) | |
| The three families at Bayside, with one example each | |
| Why AI is probable, not certain (two reasons it can be wrong) | |
| What a person still does | |

**Completion criteria.** At least 10 of 12 systems classified correctly with reasons; every section of W1 filled in plain words; at least one correction or addition noted in step 6.

---

## Lab 2 (Session 2): AI Spotter inventory and myth-buster cards

**Scenario.** Merrimack Freight & Logistics is writing its first AI policy and needs to know where AI already shows up in everyday tools. Many staff say "we don't use AI" while their email, phones, and route planner use it every day. You will inventory Merrimack's tools, then your own, and replace six myths with accurate statements.

**Goal.** Complete W2 AI Spotter inventory (Merrimack plus at least five of your own tools) and W3 Myth-buster cards (six myths).

**Data.** Merrimack tool list (fictional): Outlook with suggested replies; Teams with meeting recap; a route planner with live traffic rerouting; a fuel card portal with fraud alerts; a warehouse scanner app with "suggested pick path"; a dispatch phone system with voicemail transcription; the HR portal's résumé keyword filter; a spreadsheet with an "analyze data" button.

**Steps**

1. **Spot the AI (10 min).** For each Merrimack tool, record what the AI does, what data it sees, and what a person still decides. Use the five workplace categories: content and communication, information processing, decision support, workflow automation, and personal productivity.
2. **Automation or augmentation (5 min).** Mark each as automation (AI does the step) or augmentation (AI helps a person do it).
3. **Your own tools (10 min).** List at least five tools you use at work, in general terms, and fill in the same columns. If you are not sure whether a feature uses AI, write what makes you think it might (suggests, predicts, summarizes, ranks, generates).
4. **Myth-buster cards (12 min).** In teams of three, write the six cards below. Each card has the myth, the reality in plain words, and one workplace example.
5. **Check one claim live (5 min).** Take myth 6. In the approved assistant, open the settings or help page for memory and data controls (the instructor shows where). Record what your tool actually does on your plan.

**The six myths**

1. AI thinks and understands like a person.
2. AI is always accurate, especially when it sounds confident.
3. AI can replace human judgment.
4. AI understands context perfectly.
5. AI is objective because it is math.
6. AI learns automatically from every conversation, or its flip side: "nothing I type is kept."

**Deliverable: W2 AI Spotter inventory**

| Tool (general terms) | What the AI does | Data it sees | Category | Automation or augmentation | What a person still decides |
| --- | --- | --- | --- | --- | --- |
| e.g., Email suggested replies | Proposes short replies | The message thread | Content and communication | Augmentation | Whether to send, and what to change |

**Deliverable: W3 Myth-buster card (one per myth)**

| Myth | Reality (plain words) | Workplace example | What I will do differently |
| --- | --- | --- | --- |

**Completion criteria.** All eight Merrimack tools and at least five own tools recorded with what a person decides; six myth cards with an accurate reality statement; the memory and data-control check recorded for the approved tool.

---

## Lab 3 (Session 3): AI role and responsibility chart (final)

**Scenario.** Bayside Precision Parts uses or is piloting ten AI uses. Before it adds more, the owner wants a chart that shows, for each use, how far the AI is allowed to go and who is responsible. You will build that chart for Bayside, then for your own role.

**Goal.** Complete W4 AI role and responsibility chart (final): place 10 Bayside uses and at least three uses from your own role on the five-rung assistance ladder, and name who decides and what they check.

**The assistance ladder**

| Rung | The AI... | Example |
| --- | --- | --- |
| 1. Inform | shows information; it does not suggest a decision | Scrap-rate dashboard |
| 2. Suggest | recommends; a person chooses | Quote estimator's price range |
| 3. Draft | creates a first version a person edits | Assistant drafts a customer email |
| 4. Act with approval | prepares an action that runs only after a person approves | Agent fills a purchase order; buyer clicks approve |
| 5. Act alone | acts within set limits; people review afterwards | Auto-reorder of gloves under $250 |

**Data.** Ten Bayside AI uses (fictional): spindle failure alerts; quote price range; vision camera that rejects scratched parts automatically; meeting recaps; drafting customer emails; résumé ranking for machinist openings (pilot); auto-reorder of shop supplies under $250; AI agent that books freight pickups after the shipping lead approves; chatbot answering "where is my order?" from the ERP; safety-incident report summaries for the monthly meeting.

**Steps**

1. **Place the ten (12 min).** In pairs, put each Bayside use on a rung. For each, name who decides (a role, not a person's name) and what that person checks.
2. **Red flags (8 min).** Circle any use that is higher on the ladder than its stakes allow. Ask: *What happens if it is wrong? Who is affected? Can the person check it?* Propose a lower rung or an added check.
3. **Where AI helps and where it hurts (5 min).** The instructor shares the jagged-frontier result: in the Harvard Business School study, consultants using AI did better on tasks inside the AI's frontier and were about 19 percentage points less likely to be correct on a task outside it. Mark any Bayside use where AI might look capable but be outside its frontier.
4. **Your own role (12 min).** Add at least three AI uses from your own job (current or likely), each on a rung, with who decides and the check.
5. **Peer check (5 min).** Swap charts. Your partner questions one rung or one check. Change it or defend it in one sentence.

**Deliverable: W4 AI role and responsibility chart (final)**

| AI use | Rung (1 to 5) | What happens if it is wrong | Who decides (role) | What that person checks | Limit or safeguard |
| --- | --- | --- | --- | --- | --- |
| e.g., Quote price range | 2 Suggest | Lost job or a loss on the job | Estimator | Material cost, cycle time, customer history | Estimator signs every quote |

**Completion criteria.** See Appendix B. Meets requires all ten Bayside uses placed with a decider and a check, at least three own uses, and at least one red flag with a fix.

---

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

| Criterion | Not yet | Meets | Strong |
| --- | --- | --- | --- |
| AI concepts | Families or learning types confused | Every use correctly described as rule, predictive, or generative where relevant | Explains why one borderline use is hard to classify |
| Ladder placement | Fewer than 10 Bayside uses placed or rungs unjustified | All 10 placed with reasons; 3+ own uses | Rungs tied to stakes and checkability for every use |
| Decision authority | Decider missing or "the AI" | A role named for every use | Names what happens when the decider is absent |
| Checks and safeguards | Checks vague ("review it") | A specific check for every use | Adds limits (spend, scope, sample review) that match the rung |
| Red flags | None found | At least one red flag with a fix | Links a red flag to a law or policy (for example c. 151B for hiring) |