Prioritizing Enterprise Generative AI Adoption Under Regulatory Constraint: A Fuzzy Analytic Hierarchy Process Study of a Financial Holding Company Grounded in the Technology–Organization–Environment Framework

Abstract

Enterprise generative AI adoption in financial holding companies is examined as an organizational governance problem rather than a technology deployment decision. The purpose of this study is to identify and prioritize the factors that should govern such adoption and to clarify the role of regulatory compliance within the resulting priority structure. In a single-case design, a fuzzy analytic hierarchy process framework grounded in the technology–organization–environment framework—four dimensions and twelve criteria—was applied to the judgments of nine cross-functional experts in a Taiwanese financial holding company after its broad deployment of Microsoft 365 Copilot. Consistency was verified at the aggregate and individual-expert level, and robustness was assessed through nine leave-one-out re-estimations, sixteen sensitivity scenarios, and alternative fuzzy specifications. Three findings survive every check: organizational management is the dominant dimension (0.4055), ahead of operational benefits (0.3365), technological risk (0.1403) and regulatory compliance (0.1177); departmental acceptance of AI adoption is the first-ranked criterion; and the criteria ranked second to fifth form a closely bunched tier of organizational and value-creation conditions whose internal order is not stable. The low weight of regulatory compliance is interpreted, on methodological grounds and on the evidence of the regulator’s own risk-based classification of internal-efficiency systems, as the signature of a constraint that precedes prioritization rather than a preference within it. Two refinements of the technology–organization–environment framework for generative technologies follow, together with a testable prediction and a phased, governance-oriented implementation strategy for regulated financial groups.

The authors' abstract, as published at the source. Systems, 2026 · DOI ↗