Validating Problem Statements with Subject Matter Experts

Overview
In the PMI Cognitive Project Management for AI (PMI-CPMAI) methodology, the first phase, Business Understanding (often described as Identify Business Needs and Solutions), focuses on making sure an AI project solves the right problem before money, data and model-building effort are committed. A key practice in this phase is validating the problem statement with Subject Matter Experts (SMEs). This guide explains what the practice is, why it matters, how it works, and how to handle exam questions about it.

1. What It Is
A problem statement is a concise description of the business pain point, opportunity or decision the AI solution is meant to address. It covers who is affected, what is happening, why it matters, and what success looks like.

Validation with SMEs means taking that draft statement to the people with deep knowledge of the domain, process, data or users, and confirming four things:
- The problem is real, correctly described and worth solving.
- The root cause, not just a symptom, has been identified.
- The scope, constraints, assumptions and success criteria are realistic.
- AI is actually an appropriate approach, rather than simple automation, process change or business rules.

SMEs can include:
- Business process owners and frontline operators.
- Domain specialists such as clinicians, underwriters or fraud analysts.
- Data owners and data stewards.
- Compliance, legal and risk experts.
- End users of the eventual AI output.

2. Why It Is Important
Prevents solving the wrong problem. Many AI projects fail because the team builds a technically good model for a poorly defined or irrelevant problem. SMEs anchor the statement in operational reality.

Distinguishes symptoms from root causes. For example, "customer churn is high" may really be caused by billing errors. SMEs reveal these underlying drivers.

Confirms AI fit. CPMAI stresses asking whether AI is needed at all. SMEs may know of rule-based or process fixes that are cheaper and less risky.

Identifies the right AI pattern. SMEs help clarify whether the need is recognition, prediction, anomaly detection, conversational interaction, personalization, goal-driven optimization, or autonomous systems.

Surfaces data realities early. SMEs know what data exists, its quality, how it is labeled and its known biases. This feeds the next phase, Data Understanding.

Defines measurable success. SMEs help set realistic KPIs and acceptable accuracy or error tolerances. For instance, a false negative in medical diagnosis costs far more than one in product recommendations.

Builds buy-in and adoption. When SMEs shape the problem, they are more likely to trust and use the solution. This reduces change resistance later.

Reduces risk and supports responsible AI. SMEs flag ethical, regulatory, privacy and fairness concerns tied to the problem domain.

Supports go/no-go decisions. Validation evidence gives sponsors a sound basis for continuing, reshaping or stopping the initiative.

3. How It Works (Typical Process)
Step 1: Draft the initial problem statement. Use stakeholder input and the business case. A common structure is: [Stakeholder] needs a way to [do X] because [pain or impact], measured by [metric], within [constraints].

Step 2: Identify the right SMEs. Map stakeholders and pick those with domain, process, data and regulatory expertise. Include the people who will actually use the output.

Step 3: Prepare structured validation sessions. Use interviews, workshops, process walkthroughs, observation (gemba) and focus groups. Bring evidence such as baseline metrics, examples and current process maps.

Step 4: Probe with targeted questions.
- Is this the real problem or a symptom? Use techniques like the 5 Whys or fishbone diagrams.
- How is it handled today, and what does it cost?
- What would a successful outcome look like, and how would we measure it?
- What data exists, who owns it, and is it trustworthy?
- What decisions will the AI output support, and who makes them?
- What are the constraints: regulatory, ethical, timing, budget, explainability?
- Could a non-AI solution work?

Step 5: Reconcile differing views. SMEs may disagree. Facilitate consensus, record assumptions, and escalate conflicts to the sponsor or product owner when needed.

Step 6: Refine and document. Update the problem statement, scope, assumptions, risks, success criteria and the chosen AI pattern(s). Keep traceability to SME input.

Step 7: Obtain confirmation or sign-off. Get explicit agreement from SMEs and the sponsor. This baseline guides the Data Understanding phase.

Step 8: Iterate. CPMAI is iterative and agile. If later phases (data, modeling, evaluation) reveal new insights, return to the SMEs and revalidate the problem statement.

4. Common Pitfalls
- Relying only on executives or the sponsor and skipping frontline experts.
- Treating validation as a one-time event instead of an ongoing loop.
- Leading SMEs toward a predetermined AI solution ("solution-first" thinking).
- Ignoring data SMEs until model building starts.
- Failing to define measurable success criteria.
- Confusing a technology wish ("we want to use generative AI") with a business problem.

5. Example
A bank wants "an AI to reduce loan defaults." The fraud and credit SMEs reveal two things: many defaults stem from income misreporting, and the underwriting team needs early-warning flags rather than automatic approvals.

The validated problem statement becomes: "Underwriters need a risk signal identifying applications with likely income misstatement at submission, reducing 12-month defaults by 10% while meeting fair-lending explainability requirements."

This statement is specific, measurable, constrained, and maps to a predictive or anomaly-detection pattern.

Exam Tips: Answering Questions on Validating Problem Statements with Subject Matter Experts
Tip 1: Business problem first, technology second. If an answer jumps straight to selecting algorithms, collecting data or building a prototype before the problem is validated, it is usually wrong.

Tip 2: Prefer engagement over assumption. When a scenario describes ambiguity, conflicting requirements or unclear value, the best answer typically involves consulting or workshopping with SMEs and stakeholders. The project manager should not decide alone.

Tip 3: Ask "Is AI needed?" CPMAI expects you to consider non-AI alternatives. An option that validates whether AI is appropriate is often correct.

Tip 4: Look for root cause language. Choose answers that separate symptoms from underlying causes using SME insight.

Tip 5: Measurable success criteria matter. Correct answers usually include defining KPIs, baselines or acceptable error thresholds with SMEs.

Tip 6: Remember iteration. If new data findings contradict the original problem, the right move is to go back to Business Understanding and revalidate with SMEs. Do not force the original scope.

Tip 7: Include the right SMEs. Watch for answers that involve end users, data owners and compliance experts, not just executives.

Tip 8: Resolve SME conflicts through facilitation and the sponsor. Do not ignore minority expert views, especially on risk, ethics or regulation.

Tip 9: Documentation and sign-off. When asked what happens after validation, think of an updated problem statement, documented assumptions and risks, and stakeholder agreement before moving to Data Understanding.

Tip 10: Spot distractors. Typical wrong answers include:
- Starting model development to "see what the data shows."
- Accepting the sponsor's solution request without question.
- Postponing SME input until testing.
- Choosing the most advanced AI technique by default.

Tip 11: Link to AI patterns. Expect questions where SME validation clarifies which of the seven AI patterns fits. Pick answers that tie the pattern to the validated business need.

Tip 12: Think risk reduction and ROI. The underlying reason for SME validation is avoiding wasted investment and project failure. Answers emphasizing early risk reduction and business value alignment are strong choices.

Summary
Validating problem statements with SMEs ensures an AI initiative targets a real, well-understood, measurable business need that AI can appropriately address. In the exam, favor answers that are:
- Collaborative.
- Problem-first rather than solution-first.
- Focused on root causes.
- Measurable.
- Iterative.
- Documented, with stakeholder agreement before moving forward.