Level 2 exam — Generative AI at Work

Exam — fresh questions

12 new questions: recall (25%) + application (50%) + arbitration (25%) on everyday uses.
Answer everything, then click Check: your score and explanations will appear.

2.1 · Recall

1. What is the rule for AI-assisted writing?

Never re-read a fluent text AI signs and sends in your place AI writes, you remain the author (proofreading, signature) Always send the raw text

The draft is a draft: tone adjusted, facts verified, sent under your signature.

2.1 · Application

2. Amine must send a firm follow-up to a contractor. What does he do?

He lets AI sign and send alone He sends the raw first AI draft He pastes the client email into a non-approved tool He requests a directed rephrasing then runs the recipient test

Same firmness on substance, impeccable form — then the recipient test before sending.

2.1 · Arbitration

3. Amine wants to prepare a reply to a client email on a non-approved public assistant. What should he choose?

Describe the situation in generic terms or use the company-approved tool Pour the whole mailbox in for context Publish the email to get more opinions Paste the client email as is to go fast

Client data = approved tool or systematic anonymization; generic often suffices.

2.2 · Recall

4. How to calibrate a summary with AI?

Randomly, according to mood To the use: decide, operate, or answer a question Always as short as possible Always the same generic format

The instruction specifies use and format: executive, actions record or targeted answer.

2.2 · Application

5. Amine receives a 40-page report before the committee. What does he request?

A single generic summary for the whole committee A full copy of the report Nothing, he will read the 40 pages in one go without help A 4-sentence executive summary + an actions table + an answer to his real question

Three readings of the same document: decide, operate, target — with verification against sources.

2.2 · Arbitration

6. Amine must decide on the basis of a summary mentioning amounts and deadlines. What does he do?

He has commitments, amounts and deadlines listed then re-reads these passages directly He signs on the faith of the summary He deletes the summary and decides without reading He requests an even shorter summary

The summary points to critical passages; it never replaces re-reading them.

2.3 · Recall

7. What is AI's great strength in translation versus conventional tools?

The subscription price Adapting register and tone to context The number of languages displayed Result display speed

Tone travels with meaning — provided register and audience are specified.

2.3 · Application

8. Amine translates an announcement for a key-account client. What does his instruction specify?

Only the text length Only the translation price Register, target audience and imposed glossary, with back-translation control Nothing, he just asks “translate”

2.4 · Application

9. What is the right use of AI in ideation?

Avoid it to stay creative Let it decide in your place Copy its first ideas without sorting Broaden YOUR thinking (diverge, converge, stress-test) — you decide

AI broadens, the devil's advocate challenges, humans decide.

2.4 · Application

10. Amine's first ideas for client welcome are banal. What does he do?

He imposes angles (zero budget, competitor) then has them sorted in an impact/effort matrix He stops there and picks the first one He asks AI to decide alone He abandons the topic

Good ideas hide in volume; imposed angles then sorting by criteria.

2.4 · Arbitration

11. Amine prepares a team brainstorming with AI. In what order?

AI instead of the team Human ideas first, then AI to broaden and challenge Neither, improvise AI first to frame, then the team

First, AI anchors the group and constrains creativity; second, it broadens it.

2.5 · Application

12. Amine wants to compare three contractors in a table. Who sets the columns?

Nobody, the table fills itself The chance of previous answers You (the analyst): cost, deadline, advantages, risks, recommendation — AI fills in AI alone, with no criteria

Specifying columns and criteria gives an exploitable, capitalizable table.

What's next?

Score ≥ 80%: move to Level 3 — business uses (data, content, code, HR-legal, monitoring) keeping your standard prompts.
Score 60-79%: review the missed sections (2.1 writing, 2.2 summaries, 2.3 translation, 2.4 ideation, 2.5 structures) and redo the 48h challenges.
Score < 60%: re-read the whole level applying each method to a real case, then retake the test.
In all cases: keep your personal library (templates + standard prompts) and human proofreading before any sending or decision.