FAST-01 Projects: Applied AI Literacy at Work
Provider: 7formations | Course: FAST-01 | Instructor: Wadood Chaudhary | One capstone and two mini projects
Projects extend the labs into a work product trainees use after the course. They are completed inside the course window (mini projects between sessions, the capstone during and after session 3) and use fictional or sanitized material only. Scoring uses three levels: Not yet, Meets, and Strong. Completion requires Meets on every capstone criterion; the mini projects are formative unless the employer asks for them to be scored. One resubmission is allowed.
All companies and data are fictional. Never include real customer, patient, or employee information (G.L. c. 93H, 201 CMR 17.00; HIPAA for health information).
Capstone: My AI-at-Work Starter Kit
Brief. Build a starter kit that lets you (and a new coworker in the same role) use AI safely on day one. It pulls together your finished AI Use Map, three tested prompts, your four-step check log, and your workstation data-safety checklist, and adds a one-paragraph "how I decide" statement. The kit must be specific to your role and usable without you in the room.
Deliverables
- Finished W1 Personal AI Use Map: 8 to 10 tasks, ratings, reasons, and checks.
- W4 Three tested prompts, each with its practice source (sanitized or fictional), one checked output, and the prompt change made after review.
- A four-step check log for the final applied task with a keep / fix / discard decision.
- W5 Workstation data-safety checklist (one page).
- "How I decide" statement (120 to 180 words): when you use AI, when you don't, and who you ask.
Step plan
| Step | When | What you do |
|---|---|---|
| 1 | After session 1 | Finish W1 draft; add hours per week and a check for every "care" task |
| 2 | Session 2 | Use the data sort and drafts to write your role's "never" and "remove first" lists |
| 3 | Between sessions 2 and 3 | Draft three prompts on paper with sources picked out |
| 4 | Session 3 | Run, peer review, fix, and rerun the prompts; complete the final applied task |
| 5 | Within 7 days of session 3 | Write the "how I decide" statement, assemble the kit, submit |
Rubric
| Criterion | Not yet | Meets | Strong |
|---|---|---|---|
| Task fit judgment | Fewer than 5 tasks, or reasons are benefits only ("faster") | 8+ tasks with correct ratings and reasons tied to data, checkability, or harm | Every "care" task names the source checked against and the reviewer |
| Prompt quality | Prompts lack a source, reader, or format | All three prompts name role, reader, task, safe source, format, and gap rule | A partner ran at least one prompt unchanged and got a usable result |
| Output checking | Steps missing or decision not recorded | Four steps recorded; decision matches findings | Errors traced back to a prompt fix and rechecked |
| Data protection | Any restricted data present, or lists generic | Role-specific "never" and "remove first" lists; placeholders used | Includes a slip-reporting path with names or roles |
| Communication | Statement unclear or missing | Statement is plain, accurate, and role-specific | A new hire could follow it without questions |
Example of strong vs weak work (fictional front-desk associate, Cape Harbor Inn)
| Element | Weak | Strong |
|---|---|---|
| Use Map row | "Guest emails: good fit, saves time" | "Guest emails about check-in, pets, shuttle: use with care. Data: no card numbers needed. Check: every fact against the FAQ sheet. Harm if wrong: missed ferry, pet refused at desk." |
| Prompt | "Answer this guest about pets." | "I'm a front-desk associate at Cape Harbor Inn. Using only the FAQ text between the lines, write a 3-sentence friendly reply to a guest asking whether they can bring two dogs and a cat. If the FAQ doesn't answer something, say so and offer to check with the manager. --- [FAQ text] ---" |
| Check log | "Looked fine. Keep." | "Source: FAQ pasted. Facts: dogs OK, garden rooms only, no cats (match). Numbers: 40 lb, $35/night (match). Tone: friendly, no promise beyond FAQ. Keep." |
| Checklist | "Be careful with data." | "Never: guest name + card number, passport or license numbers, health details. Remove first: guest names, room numbers in complaints. Slip: tell the manager on duty and IT the same shift." |
Mini project 1: The Spot-the-Error Dossier
Brief. Collect and explain five AI errors on fictional or public material relevant to your job. Show how each was caught and which prompt change prevents it. The dossier becomes a teaching aid for your team.
Deliverables
- Five error cards (one per error): the prompt, the source, the AI output with the error highlighted, which of the four steps caught it, and the prompt fix.
- A one-paragraph pattern note: what kinds of errors repeat?
Step plan
- Pick a safe source (a public document, the course's fictional FAQ, or a fictional fact sheet).
- Ask 10 to 15 questions, including some the source does not answer.
- Check each answer with the four steps; save the five most instructive errors.
- For each, write the prompt fix and rerun to confirm it helps.
- Write the pattern note.
Rubric
| Criterion | Not yet | Meets | Strong |
|---|---|---|---|
| Error identification | Fewer than five real errors, or "errors" are style only | Five factual or numeric errors, each traced to the source | Includes at least one subtle error (a transposed number, a reversed rule) |
| Diagnosis | Step not named | Correct step named for each | Explains why the error was easy to miss |
| Prevention | No fix | A prompt fix for each, rerun once | Fix tested and the before/after compared |
Strong vs weak. Weak: "AI said the wrong thing about parking." Strong: "Asked about oversized vehicles. AI said 'street parking on Main St.' Source says 'lower lot on Harbor Lane.' Caught at Facts. Fix: 'Use only the FAQ; if not there, say not in source.' Rerun gave the correct lot."
Mini project 2: Teach It in Five Minutes
Brief. Prepare and deliver a five-minute explanation for a new hire of one AI limit and one data rule, using an example from class. Record it (audio or a written script) or deliver it live in session 3.
Deliverables
- A script or outline (250 to 400 words) with one limit, one data rule, one example, and one "what to do instead."
- Two questions you expect a new hire to ask, with answers.
Step plan
- Choose the limit (for example, fluent but false) and the data rule (for example, name plus account number never goes in).
- Pick a class example that shows both.
- Write in plain words; test it on a partner who is not in your role.
- Revise based on the partner's questions.
Rubric
| Criterion | Not yet | Meets | Strong |
|---|---|---|---|
| Accuracy | Describes AI incorrectly | Accurate, plain explanation | Accurate and anticipates a common misconception |
| Example | None or generic | Concrete class example | Example tied to the listener's job |
| Action | No "what to do instead" | Clear action | Action plus who to ask |
Strong vs weak. Weak: "AI can be wrong sometimes, so be careful." Strong: "AI writes what sounds likely, not what it checked. In class it told us the shuttle leaves at 9:15; the FAQ says 8:15. So before you tell a guest anything, find it in the FAQ. And never paste a guest's card number: that's protected under Massachusetts law. If you're unsure, ask the shift manager."