JOURNAL: Post-Implementation Success of AI: A Configurational Analysis of Scaling and Evidenced Value

PURPOSE: This study investigates which configurations of post-implementation conditions are associated with the scaling of artificial intelligence in organisations, and with a more demanding outcome that combines scaling with evidenced value. Whereas prior research has focused predominantly on AI adoption and readiness, less is known about how organisations develop and benefit from AI after its initial deployment.

METHODOLOGY: The study employs a sequential qualitative-to-configurational design. Thirty structured interviews with professionals from organisations using AI were analysed through framework analysis, producing calibrated scores across six post-implementation dimensions: use breadth, use depth, scaling maturity, integration maturity, governance maturity, and value measurement. The calibrated sets then served as inputs for fuzzy-set Qualitative Comparative Analysis (fsQCA), applied through two models targeting different outcome definitions.

FINDINGS: The analysis indicates that AI scaling is configurationally explainable, with the combination of use breadth and use depth forming the core driver. Two complementary pathways to scaling were identified, including a learning-by-doing path that combines perceived value with low formal governance. By contrast, the more demanding outcome combining scaling with evidenced value did not yield a sufficient configurational solution: no case achieved full membership in the joint outcome, suggesting that the ability to formally demonstrate AI value lags behind its operational expansion. This finding implies that evidencing AI value requires organisational mechanisms distinct from those that drive scaling alone.

IMPLICATIONS FOR THEORY AND PRACTICE: For practice, the findings suggest that organisations should invest in parallel measurement infrastructure rather than assuming that scaling will automatically produce evidence of value. For theory, the results highlight the need to analytically distinguish AI scaling from value evidence, and indicate that existing maturity models should separate these dimensions rather than treat them as a single progression.

ORIGINALITY AND VALUE: The study contributes by shifting AI research from adoption to the post-implementation phase, by demonstrating that scaling and evidenced value represent qualitatively different organisational achievements, and by illustrating the potential of fsQCA methodology for studying configurational complexity in AI deployment.