Impact of Data Governance on Artificial Intelligence Project Success

This thesis examines how data governance practices influence the success of artificial intelligence (AI) and machine learning (ML) projects. Despite growing organizational investment in AI, project failure rates remain high, with industry reports and practitioner experience consistently pointing to data-related challenges rather than algorithmic limitations as a primary cause. Yet, the relationship between governance maturity and AI/ML project outcomes remains insufficiently understood, and existing maturity models have largely been developed for traditional data management contexts with limited consideration of AI/ML-specific governance requirements. This study identifies the governance dimensions most critical to AI/ML success, analyses how governance maturity influences project outcomes, explores the organizational mechanisms through which governance translates into practice, and synthesizes these findings into a structured assessment framework. A convergent mixed-methods design is adopted, combining a systematic literature review, a practitioner survey, and semi-structured interviews with AI and data professionals across a range of industries. Data quality management, metadata, and lineage emerge as the most critical dimensions, with documentation a notable gap; governance shortcomings frequently manifest as apparent algorithmic failures; and decision rights and cross-functional collaboration act as mediating mechanisms. The findings also reveal a maturity perception gap. These findings inform a maturity model spanning five governance dimensions, five maturity levels, and two mediating mechanisms, mapped to AI/ML outcomes. An independent panel of eleven professionals validated the model's relevance and applicability, while identifying runtime governance as the primary direction for future work.