North Quay suspends operations.
A morning bulletin reports a closure. It does not name the supplier’s shipment or establish a delay.
WORKED EXAMPLE / 001
When the evidence changes,
the assessment should too.
A fictional supplier disruption, from assignment to decision brief.
Illustrative walkthrough · synthetic information
The scenario, documents, and responses below are authored examples. This page does not run agents or report measured performance.
01 / TEAM DESIGN
Give each perspective a job.
Research
Extract claims and dates from the supplied documents. Preserve their identifiers.
Analysis
Compare exposure, inventory, and options. State what remains unknown.
Independent review
Challenge the recommendation and test whether the sources support it.
02 / EVIDENCE PACK
Separate a report from its implications.
Five days of stock remain.
An internal planning note estimates five days of coverage. Actual demand and arrival dates need confirmation.
Backup capacity carries a premium.
An alternative supplier quotes a 20% premium. Its delivery window has not been confirmed.
03 / CHALLENGE THE ASSESSMENT
Introduce a conflicting update.
The controls switch between two authored examples of a brief. They do not generate an AI response.
INITIAL ASSESSMENT
Prepare the backup option; verify exposure before committing.
S1 establishes a reported closure, but not a shipment delay. S2 suggests a limited buffer. S3 makes backup capacity an option with unresolved cost and timing tradeoffs.
Confidence: Low. The shipment’s route and arrival time are unverified.
Human action: Request route and arrival confirmation; approve no purchase on this evidence alone.
S4 / SYNTHETIC CARRIER UPDATE
The carrier states that the shipment uses South Terminal, which remains open. This challenges the assumed exposure to North Quay, but does not guarantee arrival.
04 / MY RESPONSIBILITY
Own the decision boundary.
I would check the source dates, question the assumed route, and require confirmation before spending money. The team’s brief supports that judgment; it does not authorize the action.
05 / WHAT A REAL CASE STUDY MUST PROVE
Measure the work, not the animation.
A recorded run should retain its inputs, actual outputs, tool traces, reviewer corrections, elapsed time, and cost. It should test unfamiliar cases against a fixed baseline and record failures as well as successes.
Those measurements are not available for this illustration. Dojo provides the evaluation and promotion foundation for collecting stronger evidence.
Explore the evaluation workbench ↗