In a clustering analysis for data use, the project team finds that the clusters are...

In the PMI approach to managing AI initiatives, clustering and other unsupervised techniques depend heavily ondata quality, completeness, and relevance. When clusters are not meaningful or actionable, the primary recommended action is toreassess and improve the underlying datarather than immediately changing algorithms. PMI guidance on AI data practices emphasizes that AI teams should “ensure that datasets are sufficiently complete, representative, and aligned with the business problem before drawing conclusions from models.” This includes identifyingdata gaps, missing attributes, bias, and noisy or inconsistent records, and then addressing these deficiencies through improved collection, integration, cleaning, and feature engineering.

The PMI-CPMAI content further stresses thatdata readiness assessmentsanditerative refinement of dataare critical tasks before and during model development. Poor or incomplete data typically leads to patterns that do not map to real-world segments or behaviors, which is exactly what happens when clusters lack business meaning. While algorithm selection and trade-off analysis are also important, PMI characterizes them as secondary to ensuring that data is “fit for purpose” for the targeted use case. Therefore, the project manager should lead the team toidentify data gaps and address deficiencies, which best aligns with PMI’s emphasis on data quality as the foundation of reliable AI outcomes.