# FAST-16 Projects: Working Well with AI

Provider: 7formations  |  Course: FAST-16  |  Instructor: Wadood Chaudhary  |  One capstone and two mini projects

Projects extend the labs into habits trainees keep. Mini projects run between sessions; the capstone is completed during and within seven days after session 3. All material is fictional or sanitized. Scoring uses **Not yet**, **Meets**, and **Strong**; completion needs Meets on every capstone criterion. One resubmission is allowed.

Suggested reading for the projects: Laurence Lars Svekis, *AI Literacy for All Employees* (2026), chapters 6, 7, 13, and 15.

Never include real customer, patient, or employee information (G.L. c. 93H, 201 CMR 17.00).

---

## Capstone: My Ask-Check-Own Playbook

**Brief.** Build a playbook for one recurring task in your role that a coworker could follow on their first day: how to ask the AI well, how to check what comes back, and how to own and document the result.

**Deliverables**

1. Two tested prompts for the task (four elements and a gap rule), each with a sample output and the follow-ups that improved it.
2. A W2-style evaluation record for one real-type output (fictional or sanitized), with at least one error found and typed, and one verification technique used.
3. W4 Safe AI workflow card (final) for the task.
4. A one-page lesson from a case: pick one of the seven cases from Lab 3 and explain how your workflow would prevent the same failure.
5. W5 Ask-check-own pocket card.

**Step plan**

| Step | When | What you do |
| --- | --- | --- |
| 1 | After session 1 | Pick the task; draft and test two prompts |
| 2 | After session 2 | Run a five-step evaluation on one output; record it |
| 3 | Session 3 | Build and peer-test W4 |
| 4 | Within 7 days of session 3 | Write the case lesson; assemble and submit |

**Rubric**

| Criterion | Not yet | Meets | Strong |
| --- | --- | --- | --- |
| Prompting | Elements missing; no gap rule | Four elements, gap rule, two useful follow-ups | A partner reused the prompt unchanged |
| Evaluation | Steps skipped | All five steps recorded; error typed correctly | Traces the error to a prompt fix and rechecks |
| Verification | None | At least one technique (source trace, recalculation, lateral reading) | Uses two techniques, including checking a citation exists |
| Workflow | Steps vague | All five steps specific to the task | Escalation path and discard trigger stated |
| Case lesson | Summary only | Lesson linked to a specific safeguard | Links to a policy or law where relevant |

**Example of strong versus weak work (fictional inside-sales rep, Bayside Precision Parts)**

| Element | Weak | Strong |
| --- | --- | --- |
| Prompt | "Write a lead-time email." | "I'm an inside-sales rep at a CNC shop. Using only the order status below, write a three-sentence email to the customer's buyer giving the new ship date and the reason. If something isn't in the status, write [CHECK]. --- [status, no account numbers] ---" |
| Evaluation | "Looks good." | "Accuracy: ship date matches ERP. Completeness: missing the partial-shipment option, added. Relevance: answers the buyer's question. Bias: none. Judgment: send." |
| Document | (none) | "Drafted with Copilot Chat; dates checked against ERP by me." |

---

## Mini project 1: Prompt Before-and-After Gallery

**Brief.** Collect five weak prompts (your own or invented, no private data) and rewrite them. For each, show both outputs and say what changed.

**Rubric**

| Criterion | Not yet | Meets | Strong |
| --- | --- | --- | --- |
| Rewrites | Fewer than four elements in most | All five rewrites use the four elements | Each uses a different technique |
| Comparison | Outputs missing | Both outputs shown with one change noted | Notes an error the weak prompt caused |

---

## Mini project 2: Verify Three Claims

**Brief.** Ask an approved assistant three factual questions about your industry in Massachusetts (for example, a licensing rule or a safety standard). Verify each answer with SIFT and record what you found.

**Rubric**

| Criterion | Not yet | Meets | Strong |
| --- | --- | --- | --- |
| Verification | Asked the AI to confirm itself | Each claim traced to an official or independent source | Notes one claim that was partly right and how |
| Record | Missing sources | Source, date, and verdict for each | Explains which SIFT move did the work |