Question 12
Domain 2A team configures a 5-day Cowork task with one end-of-task review. What's the cadence critique?
Correct answer: C
Explanation
A 5-day Cowork task with only one end-of-task review creates too much unchecked drift; issues are found only after the work is mostly done, when fixes are more expensive. The cadence should include daily checkpoints with concrete artefacts so the human can "correct course while drift is cheap to fix."
Why each option is right or wrong
A. End-of-task review is sufficient for any task length.
Single end review leaves multi-day drift unchecked and increases expensive late-stage rework.
B. 5-day tasks should be split into 5 separate single-day tasks.
Daily checkpoints are needed, but not necessarily five fully separate one-day tasks.
C. 5-day task needs at least daily cadence — drift accumulates and is expensive to discover at end. Add daily checkpoints with concrete artefacts so the human can correct course while drift is cheap to fix.
For a 5-day Cowork task, a single end-of-task review leaves the entire work period unobserved, so any misunderstanding or divergence can compound across all 5 days before it is detected. The critique is that the review cadence is too sparse for a task of this length: the human should be checking at least once per day, with concrete intermediate artefacts, so course corrections happen before 24-hour blocks of drift accumulate.
D. The agent should self-monitor and self-correct.
Self-monitoring helps, but human oversight is still needed for goal alignment and quality control.