Running example project: a week of work tooled by AI and a one-page usage charter

Running example project — Your augmented week + your team's charter

Goal: apply each level of the guide to YOUR real work week, bank your winning prompts, and deliver your team's AI usage charter.
Materials: the starter kit below (prompt template, use-case grid, charter frame).
Method: one step per level, on real tasks — measuring time saved and noting gray cases encountered.

Step 0 · Level 0 — Choosing your tools knowingly

Download the kit. Test the same prompt on 2 assistants, identify what your organization allows, set your account privacy, and deliberately provoke a hallucination (invented references) to calibrate your vigilance.

Step 1 · Level 1 — Forging your master prompts

Build 5 RCTF prompts for your 5 most frequent tasks (kit template). For each: lazy version vs RCTF version, then add a benchmark example (few-shot). Record the winners in your library.

Step 2 · Level 2 — A week of everyday uses

Produce with AI, on real material: 2 tricky emails (follow-up, refusal), 1 long-document summary as an actions table, 1 translation checked by back-translation, 1 complete brainstorm (divergence → matrix → devil's advocate), 1 reusable template. Note time saved each time.

Step 3 · Level 3 — Applying to YOUR job

Pick the 2 job cases that look like you (data/Excel, content, technical, HR/admin, monitoring) and document them with the kit grid: task, prompt used, result, verification done, measured gain, identified risks.

Step 4 · Level 4 — Scaling up with tools

If you have Copilot 365: redo your 3 best uses built-in (Word "/", Excel on table, Teams Recap). Otherwise: create a custom assistant with your instructions + 2 anonymized reference documents. Test it with 5 questions including 2 traps.

Step 5 · Level 5 — Delivering the charter

Compile: your tool decision tree, your data rules, your verification protocol and your week's gray cases. Have AI write the charter's first draft (kit frame), confront it with 3 real gray cases, get it validated at the right level, date the review in 6 months.

Success criteria

Your library holds 5 master RCTF prompts + 2 benchmark examples, proven on real material.
Time saved is measured on at least 4 tasks (before/after).
Each week's output was verified in proportion to its stakes.
The one-page charter is written, confronted with your gray cases and submitted for validation.