Meridian's team built a small offline demo of the fraud-detection model against a static sample dataset purely to test technical feasibility, with no plan yet to touch production data or live systems. What is this best classified as?

Differentiating between pilots and proofs of concept is an explicit CPMAI Business Understanding enabler. A proof of concept is characterized by testing technical feasibility in isolation — using a static sample dataset, with no integration into live production systems or operational processes — which matches this scenario exactly. A pilot, by contrast, involves a smaller-scale but real integration into live operations, typically with real-time or near-real-time data and actual users, to validate not just technical feasibility but operational fit. The pilot option claiming it qualifies because it involves data drawn from real fraud cases is the closest distractor and the one candidates most often get wrong: using real historical fraud data does not make something a pilot if it isn't integrated into live systems or workflows — the defining line is production integration, not data realism. The full production deployment option is clearly premature; nothing here has touched production. The Go/No-Go assessment option confuses a project-management checkpoint with a type of validation activity; these are different concepts entirely. Getting this distinction right matters because proof of concept success alone should never justify skipping straight to full production, as covered in the earlier POC-pitfalls enabler.