A Framework for Scalable Enterprise AI Adoption: The SDICG Model

Abstract

Despite near-universal deployment of artificial intelligence (AI) in enterprise settings, fewer than one-third of organisations successfully scale AI beyond the pilot stage. This paper does not claim to discover these adoption failure modes, which are independently documented elsewhere; instead, it presents the SDICG Framework as an integration and operationalisation framework, unifying them under one PM-actionable lifecycle-Strategic Alignment (S), Data Readiness (D), Cross-Functional Integration (I), AI Governance and Ethics (G), and Continuous Improvement (C)-and makes four original contributions: the Organisational Context Canvas (OCC) for pre-investment build-vs-adopt assessment; the Vendor-Domain Lock-in principle identifying a distinct AI build failure class; the Co-Creation Model requiring domain and technical experts to jointly own AI artefacts from inception; and the Dual Data Imperative formalising data quality and data security as inseparable, co-accountable prerequisites [25]. The framework is validated through a 13-respondent qualified practitioner survey (Cronbach's $\alpha$ = 0.941; 7-point scale), five successful and one discontinued first-hand use cases from a Global Regulated Medical Device Organisation (GRMD), and four published secondary cases. The framework aligns with UN SDGs 3, 4, 8, 9, 13, 16 and EU AI Act compliance.