From Pilot to Production: A Seven-Check Readiness Framework for Deploying Generative and Agentic AI in the Enterprise

Many enterprise generative AI initiatives stall between a successful pilot and dependable production use. Industry forecasts suggest that at least 30% of generative AI projects will be abandoned after proof of concept, and that more than 40% of agentic AI projects will be cancelled by the end of 2027. This paper argues that the gap is primarily a management and delivery problem rather than a model-capability problem, and proposes a practitioner framework for go-live decisions. The framework rests on a single organizing analogy, treating an AI system like a new hire, and translates it into seven readiness checks (value, ownership, data, evaluation, guardrails, operations and adoption), six hard stops that block launch regardless of overall score, an eight-part evidence pack, a 0–2 scoring rubric, and an additional control set for agentic systems that can take actions. The checks are mapped to established sources including the NIST AI Risk Management Framework, ISO/IEC 42001, the EU AI Act, and empirical work on machine learning engineering. Seven publicly documented cases are used to illustrate the failure modes each check is designed to catch. The paper closes with limitations and a proposed agenda for empirical validation. JEL classification: M15, O32, O33.