12 questions to validate the integrations section: Copilot in Office, Outlook & Teams, Work search, custom assistants and agents, and reasoned tool choice.
Answer everything, then click Check: your score and explanations will appear.
4.1 · Recall
1. What does built-in Copilot in Word, Excel and PowerPoint bring versus a general assistant?
Same prompt methods, but your files serve as direct sources.
4.1 · Application
2. Amine wants to analyse his tracking table in Excel with Copilot. What must he check first?
Without a structured table, Copilot's findings are fragile.
4.2 · Application
3. Amine wants the Copilot recap of a Teams meeting. What must he do?
No announced transcription = no recap; transcription must be announced.
4.2 · Application
4. What should be done with the generated meeting recap before releasing it?
Action assignments are sometimes wrong: check before releasing.
4.3 · Application
5. Back from a trip, Amine wants the status of lot B. Which question is most effective?
The transversal question crosses emails, files and meetings in one sourced answer.
4.3 · Arbitration
6. In Work mode, Amine gets a budget document he should not have seen. What must he do?
Revealed over-sharing is flagged, not exploited.
4.3 · Recall
7. What is the difference between the Copilot chat's Work and Web modes?
You must always know which mode you question before concluding.
4.4 · Application
8. Amine creates an assistant for the site's recurring questions. What must he engrave in it?
Instructions + knowledge base + “if you don't know, refer to a human”.
4.4 · Arbitration
9. Should Amine's assistant alone send an acceptance approval to the subcontractor?
Reasonable boundary: automate preparation, keep the human on decisions and sending.
4.4 · Application
10. Before sharing his assistant with the team, which test must Amine run?
An assistant's quality is judged by its out-of-scope refusals as much as its answers.
4.5 · Recall
11. What is the no. 1 tool-choice question?
Data sensitivity beats tool performance.
4.5 · Arbitration
12. Amine wants to reformulate an “anonymised” internal procedure in a public assistant. What should he decide?
Pseudonymisation is not anonymisation: if in doubt, approved tool.
What's next?
Score ≥ 80%: move to Level 5 — confidentiality, checking, bias and the AI usage charter.
Score 50-79%: revisit the failed sections (Office 4.1, Outlook & Teams 4.2, Work search 4.3, assistants 4.4, tool choice 4.5), re-read the explanations above, then retake the test.
Score < 50%: restart the level from 4.1 with the Word → Excel → PowerPoint circuit and the decision tree, then retry the exam with a fresh mind.