In the banking sector, the adoption of Artificial Intelligence (AI) is often driven by a deterministic narrative that associates automation with direct increases in productivity and risk quality. However, empirical evidence signals a persistent "execution gap": value does not emerge automatically, remaining tied to fragmented initiatives and structural limitations of the operating model and governance. This paper addresses this gap by proposing a decision-making framework for the adoption of AI-driven Decision Intelligence, establishing theoretical continuity between integration technologies (SOA) and new paradigms of algorithmic decisionmaking. Through a qualitative design based on an archival single case study and the application of pattern matching logic, the research analyzes the case of Bank of America as an empirical contrast to demonstrate how the transformation of AI into a scalable decision-making capability requires systemic alignment between infrastructure and control frameworks. The analysis identifies five mechanisms of value generation: strategic anchoring (enterprise-by-design), the establishment of proprietary information assets (data moat), governance as an accelerator of institutional trust, socio-technical integration of personnel, and the decoupling of operating costs (operational leverage). The study offers a threefold contribution: theoretical, by evolving adoption models for integration technologies; empirical, by validating industrial scalability mechanisms; and managerial, by providing a strategic roadmap for the transition from experimentation to operational Decision Intelligence.