A team is evaluating different AI models for their project.

Within CPMAI, model evaluation is never framed as a single-number decision. The methodology stresses that AI performance must be assessed using multiple technical and business metrics, not just error rate. In the Model Evaluation phase, guidance explains that model success “goes beyond raw accuracy” and must be aligned with ROI and cost–benefit criteria defined earlier in the project. This explicitly means that a team focusing only on error rate can easily miss critical aspects such as precision/recall trade-offs, class imbalance, latency, robustness, explainability, fairness, and business impact.

CPMAI materials also highlight that evaluation should answer whether the model is fit for purpose in the real context, which requires comparing different models across a balanced scorecard of metrics, including technical quality and business KPIs. Selecting a model based solely on error rate risks deploying a solution that looks good statistically but performs poorly in production, causes unintended bias, or fails to meet stakeholder expectations. Therefore, according to CPMAI-aligned evaluation practices, the outcome of using only error rate as the selection criterion is a potential to overlook other critical performance metrics, making option A the correct answer.