In CPMAI’s Executing the Business Understanding Phase, teams first “formulate AI-specific business questions” and “estimate time-to-ROI for various AI project types” before matching business needs to cognitive patterns . This ensures ROI-driven objectives guide the selection of one or more of the Seven Patterns of AI, rather than the reverse.

=========

In Phase II: Data Understanding, CPMAI urges teams to rigorously assess data feasibility—asking whether the data is available, sufficient in quality, and properly aligned with business goals—and to perform a Go/No-Go decision before proceeding . Expanding project scope in the face of data issues violates the methodology’s iterative, scope-controlled approach. Instead, CPMAI recommends either down-scoping (Option C), verifying existing data sufficiency (Option B), or identifying necessary data sources (Option D) to resolve issues without amplifying risk.

In CPMAI™ v7’s Phase III: Data Preparation, teams are instructed to construct the final modeling dataset through a variety of enhancement activities—including data augmentation, which specifically covers transforming existing records or generating entirely new records to increase volume and variety. This “augmentation” is described as “constructive data preparation operations such as the production of derived attributes or entire new records, or transformed values for existing attributes” .

Moreover, under the Training & Test Data Requirements task, the Workbook explicitly asks project teams to determine “What transformation or multiplication activities can be done to increase training data volume while maintaining quality” . Performing data transformation (e.g., image rotations, color jitter, cropping) and multiplication (synthetic record generation) directly addresses the lack of labeled samples without incurring the cost or delay of third-party purchases, making option B the correct approach.

=========

Phase III: Data Preparation is dedicated to all data-centric tasks, including cleaning (“Clean data”) and enhancement (“Enhance & Augment data”) of raw inputs so they’re ready for model consumption.

=========

CPMAI’s Domain VI: Trustworthy AI is designed to span the entire project lifecycle, embedding ethics, transparency, fairness, privacy, and security checks into every phase—from Business Understanding through Operationalization—so that trustworthy practices are not an afterthought but a continuous, integrated activity .

=========

In Phase III: Data Preparation, the Select Data task instructs teams to choose only the records and attributes needed for modeling—documenting inclusions and exclusions to reduce volume and complexity. This selective pruning of columns and rows is the primary mechanism for trimming excessive data before modeling.

=========