What should change after the AI pilot?
A teammate reports frequent AI use. You have two dry runs over the same twelve fictional requests, with the same acceptance checklist.
The question
Northline Services and all people, records and outcomes in this case are fictional. You manage its internal-tools team. Use only the supplied work evidence.
| Requests | Drafting minutes | Review minutes | First-pass acceptances | |
|---|---|---|---|---|
| Manual | 12 | 120 | 30 | 11 |
| AI-assisted | 12 | 48 | 72 | 9 |
Later rework is not timed. These small dry runs do not predict production performance.
At least 11 of the 12 requests must pass the first review. The team may use approved AI with human review; usage counts are not a performance target.
“Choose your next management action and explain the evidence behind it.”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Count the whole measured workflow
What would you do here? Open the move you’d make.
Include drafting and review
The strong move. The measured time reduction is 20%, before later rework.
Report only the 60% drafting reduction
Tempting, but it misses something. That omits the extra review work.
See how a strong candidate says it
“Manual work takes 150 minutes; AI-assisted work takes 120. Drafting is quicker, while review takes longer.”
Why it works: The measured time reduction is 20%, before later rework.
Step 2: Check the quality floor
What would you do here? Open the move you’d make.
Keep the acceptance check
The strong move. Time saved cannot make a failed quality check pass.
Treat frequent AI use as success
Tempting, but it misses something. Using the tool is not the same as accepted work.
See how a strong candidate says it
“Manual work meets 11/12. AI-assisted work reaches 9/12, or 75%, so it misses the agreed floor.”
Why it works: Time saved cannot make a failed quality check pass.
Step 3: Coach the recurring error
What would you do here? Open the move you’d make.
Inspect errors and try a change
The strong move. The next action should build capability and test whether the error changes.
Ban AI after this sample
Tempting, but it misses something. Twelve cases do not establish that every AI use will fail.
See how a strong candidate says it
“Review the three failed outputs with the teammate, identify the repeated error and try a bounded change on another comparable packet.”
Why it works: The next action should build capability and test whether the error changes.
Step 4: Agree the next decision
What would you do here? Open the move you’d make.
Set an evidence-based checkpoint
The strong move. The manager owns the standard and the decision to expand.
Double the output target now
Tempting, but it misses something. The pilot has not shown that more volume will meet the standard.
See how a strong candidate says it
“Keep human review and the same floor. Compare total measured time and first-pass quality after the next dry run before expanding use.”
Why it works: The manager owns the standard and the decision to expand.
The answer, in short
- Decision
- Continue a bounded learning trial with review.
- Because
- The AI-assisted run saves measured time but fails the quality floor.
- What would change it
- A later comparable packet that meets quality and saves total time could support wider use.
Common traps
- Do not infer staffing savings from these twelve examples.
Check your answer
- I counted review time.
- I kept the quality floor.
- I chose specific coaching.
- I set a new evidence check.
Last reviewed Nov 10, 2026