Practical guide
AI fluency for management consultants: build an issue tree and evidence plan
Turn a fictional performance question into a mutually distinct issue tree, evidence requests and hypothesis tests.
Management consultants can use AI to structure a problem while checking that branches do not double count and hypotheses are not presented as findings. This fictional exercise produces an evidence workplan, not client advice, a proven diagnosis or an assured outcome.
Translate the question into a decision
An issue tree should serve a choice, not merely organize interesting analysis.
A fictional service team asks why completion time increased and whether to change staffing, intake or workflow. Define the period, unit of work and decision owner. The packet does not establish that any of the three causes is responsible.
Ask AI for candidate interpretations, then select the one the supplied question supports. Record exclusions such as customer outcome and cost if they are outside scope.
Build distinct first-level branches
Branches should explain different mechanisms and cover the scoped question.
Use demand mix, available capacity and process time as first-level branches. A repeated handoff belongs under process time, while absence belongs under capacity. Define each branch so the same observation is not counted twice.
AI can critique overlap, but the worker should test whether branches remain useful with the available data. Do not call the tree exhaustive beyond the defined exercise scope.
Attach evidence to hypotheses
A hypothesis is useful only when evidence could change confidence in it.
For demand mix, request case type and arrival time. For capacity, request scheduled coverage and absence records in aggregate. For process, request dated stage transitions. Keep access and privacy constraints beside each request.
Do not invent a dataset or ask for personal details that the decision does not need. Mark which evidence exists, which requires owner approval and which cannot be obtained in the exercise.
Run a contradiction test
Evidence that challenges the favored branch should remain visible.
A fictional aggregate shows arrivals were stable while stage transitions lengthened. That weakens a pure demand-volume explanation but does not prove a workflow cause. Update the hypothesis table and identify remaining alternatives.
Ask AI to generate the strongest counterinterpretation, then verify it against the packet. Do not convert correlation or a single period into causation.
Deliver the workplan, not a verdict
The handoff should let the client owner choose the next analysis.
Include decision, issue tree, definitions, hypotheses, evidence requests, access owners, contradiction and proposed sequence. Clearly label all conclusions as pending until analysis occurs.
Review branch distinction and evidence sufficiency. The process-improvement guide diagnoses a supplied workflow trace; this page creates the prior consulting workplan and preserves client decision ownership.
Convert the tree into a bounded interview guide
Qualitative evidence should test branches without leading the source toward a favored diagnosis.
Draft fictional questions for the intake owner, coverage planner and process owner. Ask for dated examples, definitions and exceptions before asking for explanations. Avoid saying delays are caused by handoffs or staffing. Each question should map to one branch and note what answer would weaken as well as strengthen the hypothesis.
Use AI to critique leading language and duplicate questions, then review the revised guide manually. Record consent, access and attribution requirements as unresolved for real work. The exercise creates a research plan only; it does not authorize interviews or allow a model-generated response to stand in for stakeholder evidence.
Sources and scope
NIST: AI risk management. Skills England: workplace AI foundations.
These references provide background, not validation or endorsement of this exercise. The case details, calculations and suggested review questions are original instructional material. Use them to discuss observable work, not to infer customer outcomes, professional credentials or performance in every setting. Before adapting the exercise, confirm the relevant facts, approved tools, data permissions and decision owners. If you change the case, revisit the expected answers and checks as well. These examples describe practice tasks, not a promise that a particular product includes the fictional features.
Sources: [1] [2]
Sources
- 1.AI RMF Core · NIST
- 2.AI foundation skills for work benchmark · Skills England