Executive Summaries of Data Assessment Findings

Overview
In the PMI Cognitive Project Management for AI (PMI-CPMAI) methodology, the Identify Data Needs phase (aligned with Phase II, Data Understanding, of the CPMAI lifecycle) is where the team determines whether the data required for an AI solution actually exists. It also checks whether that data is accessible, sufficient in quality and quantity, and appropriate for the intended use. The team captures the results of this investigation in a data assessment. They then distill those results into an Executive Summary of Data Assessment Findings: a concise, decision-oriented communication for sponsors, executives, and key stakeholders.

Why It Is Important
Data drives AI projects. If the data is missing, poor, biased, inaccessible, or legally restricted, even the best model will fail. Most AI project failures trace back to data issues, not algorithm choice. The executive summary matters for several reasons:
- It enables go/no-go decisions. Leaders need a clear answer to one question: do we have the data to proceed? CPMAI is iterative, and the summary supports the decision to continue, adjust scope, acquire more data, or stop.
- It translates technical detail into business language. Executives rarely read data profiling reports. They need the implications: risk, cost, timeline, and feasibility.
- It manages expectations early. Data gaps surfaced early prevent expensive surprises during model development.
- It supports governance and accountability. It documents what was known, when, and what decisions followed. This matters for responsible and trustworthy AI, compliance, and audits.
- It aligns stakeholders. A shared understanding of data readiness keeps business owners, data owners, IT, legal, and the AI team aligned.
- It feeds back into the business understanding. Findings may require revisiting the problem definition or the AI pattern selected in Phase I.

What It Is
An Executive Summary of Data Assessment Findings is a brief, high-level document or presentation. It summarizes the results of evaluating data sources against the project's data requirements, and it highlights the key findings, risks, gaps, and recommendations. It is not a full data dictionary, profiling report, or technical appendix. Those may be referenced, but the summary focuses on what decision-makers need to know and do.

Typical components include:
1. Purpose and context: the business problem, the AI objective, and the AI pattern (e.g., recognition, predictive analytics, conversational, anomaly detection).
2. Data requirements recap: what data the solution needs (types, volume, granularity, labels, timeliness).
3. Sources assessed: internal and external sources, owners, and systems.
4. Key findings on:
  - Availability and accessibility: does the data exist, and can we get it?
  - Quantity: is there enough to train, validate, and test?
  - Quality: completeness, accuracy, consistency, timeliness, and validity.
  - Relevance and representativeness: does it reflect the real-world population and use case?
  - Labeling status: for supervised learning, is labeled data available, or what effort is needed?
  - Bias and fairness concerns.
  - Privacy, security, legal, and ethical constraints: PII, consent, regulations such as GDPR or HIPAA, licensing, and data sovereignty.
5. Gaps and risks: what is missing and the impact on feasibility, cost, schedule, and model performance.
6. Options and recommendations: e.g., acquire third-party data, generate synthetic data, launch a labeling effort, clean or transform data, narrow scope, change the AI pattern, or pause.
7. Overall data readiness rating: e.g., ready, ready with conditions, or not ready (often shown as Red/Amber/Green).
8. Decisions requested and next steps: the specific approvals, funding, or actions needed from executives.

How It Works (Process)
1. Start from business understanding. Confirm the problem statement, success criteria, and AI pattern. These define what data is needed.
2. Define data requirements. List the needed features, labels, volumes, frequency, and history.
3. Inventory and locate data sources. Identify owners, systems, formats, and access paths.
4. Assess the data. Perform profiling and exploratory analysis. Evaluate quality dimensions, volume, representativeness, and bias, and review legal, privacy, and ethical constraints with the relevant stakeholders (data stewards, legal, security, compliance).
5. Analyze gaps. Compare what is needed against what is available and quantify the shortfalls.
6. Evaluate options. Estimate the cost, time, and risk of closing each gap.
7. Synthesize. Draft the executive summary. Lead with the bottom line (readiness and recommendation), then supporting findings, using plain language and visuals where helpful.
8. Validate. Review the summary with data owners and the technical team for accuracy before presenting it.
9. Present and decide. Walk through the summary with sponsors, capture decisions, and update the project plan, risk register, and scope accordingly.
10. Iterate. CPMAI is iterative. If data is insufficient, loop back to business understanding or re-scope, then reassess.

Characteristics of a Good Executive Summary
- Concise (often one to two pages or a few slides).
- Bottom line up front (BLUF).
- Business-oriented, connecting data issues to business impact.
- Honest and transparent about limitations; it does not hide bad news.
- Actionable, with clear recommendations and decisions required.
- Evidence-based, referencing detailed technical reports in appendices.
- Tailored to the audience: executives, not data scientists.

Common Pitfalls
- Filling the summary with technical jargon and statistics.
- Downplaying data gaps to keep the project moving.
- Omitting privacy, legal, or bias considerations.
- Giving no clear recommendation or decision request.
- Skipping validation with data owners.
- Treating the assessment as one-time rather than iterative.

Exam Tips: Answering Questions on Executive Summaries of Data Assessment Findings
1. Think audience first. If a question asks what to include or how to communicate, choose the answer that focuses on business impact, risk, and decisions, not technical detail. Executives need conclusions and implications.
2. Favor transparency. When data issues are found, the correct action is almost always to communicate them clearly and early, with options. Avoid answers that hide, delay, or minimize problems, or that proceed to modeling anyway.
3. Remember the go/no-go purpose. The summary supports a decision about whether and how to continue. Answers mentioning feasibility, readiness, and recommendations are usually strong.
4. Data before model. CPMAI stresses that data understanding precedes data preparation and modeling. Be wary of answers that jump to algorithm selection or model building when data readiness is unconfirmed.
5. Iteration is acceptable. If the data is insufficient, the best answers often involve revisiting business understanding, re-scoping, acquiring or augmenting data, or pausing. Do not force the project forward.
6. Include responsible AI considerations. Privacy, consent, regulatory compliance, bias, and representativeness belong in data assessment findings. Answers that ignore them are likely wrong.
7. Distinguish summary from detail. A data dictionary, profiling statistics, or schema documentation is supporting material, not the executive summary itself.
8. Look for actionable recommendations. A good summary ends with clear next steps and decisions requested. Choose answers that include options with their trade-offs (cost, time, risk).
9. Engage the right stakeholders. Validating findings with data owners and stewards, and involving legal and compliance where needed, is best practice.
10. Watch for scenario keywords. Phrases like sponsor asks whether to proceed, data quality issues discovered, insufficient labeled data, or stakeholders confused by the technical report point to the executive summary as the right communication tool.
11. Eliminate extremes. Discard options that cancel the project immediately without analysis, or that ignore the issues entirely. The best answer usually assesses, communicates, recommends, and lets the stakeholders decide.
12. Link to the risk register and plan updates. Findings should update the project risk log, scope, schedule, and budget. Answers reflecting integration with project management artifacts are often correct.

Sample Question Approach
Scenario: The data assessment reveals that only 40% of the required historical records contain the target label, and some records include customer PII without documented consent. What should the project manager do next?
Best answer pattern: Prepare an executive summary for the sponsor. It should state the labeling gap and the privacy compliance risk, assess their impact on feasibility, and present options such as a labeling initiative, consent remediation or anonymization, or re-scoping. It should then request a decision before proceeding to data preparation.
Wrong patterns: Start training on the labeled 40% anyway, remove the PII silently and continue, or send the raw profiling report to executives.

Key Takeaway
The Executive Summary of Data Assessment Findings bridges technical data evaluation and business decision-making. It concisely and honestly communicates data readiness, gaps, risks, and recommendations, so leaders can make informed, responsible go/no-go and scoping decisions. This is central to the data-first, iterative philosophy of PMI-CPMAI.