Skip to content
Practice cases
Professional ServicesBeginner

Rank an AI Automation Backlog

Use one explicit criterion per comparison, preserve evidence and test the resulting ranking for cycles and sensitivity.

18 minutes

Scenario

A services firm can run one AI pilot next quarter. Four opportunities remain after compliance screening: proposal drafting, invoice exception triage, meeting-note extraction and knowledge-base maintenance. Stakeholders keep switching between value, ease and personal preference.

Your role
Transformation analyst facilitating prioritisation
Method
Pairwise Comparison

Evidence pack

e1

Proposal drafting

900 hours/year; moderate quality variance; requires client-data controls.

e2

Invoice triage

1,400 hours/year; labelled historical outcomes; errors delay cash collection.

e3

Meeting notes

1,100 hours/year; fast pilot; consent and retention rules vary by client.

e4

Knowledge maintenance

700 hours/year; stale content affects all three other use cases.

e5

Capacity

The pilot team has eight weeks and access to one process owner.

e6

Criterion

The sponsor defines first-priority value as verified reduction in manual exception-handling hours within 90 days.

e7

Uncertainty

Hours estimates have confidence bands of +/-15%, +/-10%, +/-25% and +/-30%.

e8

Dependency

Proposal and triage pilots can use current approved knowledge; meeting notes cannot enter production until retention guidance is resolved.

Constraints

  • Use the sponsor's single criterion for the main comparison round.
  • Every win must have an evidence note and confidence label.
  • A ranking is provisional if a material cycle or low-confidence pivotal comparison remains.

Case steps

Work through each prompt using the evidence pack. Answers and rubric weights stay protected in the interactive flow.

1
Open Response

Translate the sponsor's criterion into a comparison question and state the decision boundary.

2
Structured

Complete all six unique pairwise comparisons for the four opportunities, recording winner, evidence and confidence.

3
Structured

Count wins, produce a provisional ranking and identify any A > B > C > A cycle or tied result.

4
Open Response

Revisit the pivotal comparison under one alternative criterion and explain whether the pilot choice changes.