AI-driven analytics products at scale: A managerial and technical perspective

Organizations continue to struggle with scaling AI-powered analytics products from successful research prototypes into enterprise-grade, production systems. Prior research typically addresses this challenge through two partially disconnected streams: a technical stream focused on MLOps, data infrastructure, and model lifecycle automation; and a managerial stream focused on governance, organizational capability, and data-driven decision-making. This separation limits both theory and practice by under-explaining how technical and managerial systems interact during scaling. 
This study proposes an integrated socio-technical framework—the Integrated Scaling Architecture for AI Analytics (ISA-4)—that conceptualizes scaling as a co-evolutionary process across four interlinked layers: Data Infrastructure, Model Operations, Governance, and Organizational Alignment. Building on the Resource-Based View, Dynamic Capabilities, and Socio-Technical Systems theory, the framework defines scaling performance as a multiplicative function of resource endowments, reconfiguration capability, and the quality of socio-technical coupling across layers. 
The study advances eight propositions and operationalizes coupling mechanisms as a measurable construct across three dimensions: artifact, process, and normative coupling. The framework is applied to a structured multi-case investigation of pioneering AI-driven firms, illustrating how different coupling configurations explain variation in scaling speed, reliability, and organizational coordination.
This work contributes (1) a socio-technical co-evolution theory of AI analytics scaling, (2) a structured framework linking technical and managerial subsystems, and (3) a practitioner-oriented diagnostic lens to identify bottlenecks across the AI scaling lifecycle and inform platform strategy.