The potential of analytics to innovate work processes is often hindered by the socio-technical challenges of integrating models into complex, fragmented digital infrastructures. Drawing on a three-year Action Design Research (ADR) project with a large Danish manufacturer, this paper addresses the critical gap in existing methodologies that focus primarily on model building while neglecting the deployment and integration phases. We propose a development approach and six design principles grounded in a digital infrastructure perspective, which conceptualizes the analytics system as an integral part of the organization's "operational backbone". The research contribution is twofold: first, it provides prescriptive design knowledge that shifts the focus from model-centric to system-centric development; second, it advances design science by demonstrating how multiple demonstrators can lead to a nascent design theory that re-conceptualizes the "installed base" of legacy systems as a generative force for innovation. For practitioners, specifically managers and data scientists, the study provides a practical roadmap to mitigate "over-expectations" towards technologies like AI and ML. This is achieved by helping stakeholders understand and manage the critical trade-offs between analytical and infrastructural complexity. By recognizing that projects exhibiting high complexity in both areas are prone to failure, the suggested approach emphasizes an incremental strategy. This imply starting with simple visualizations to achieve "quick wins" before moving to complex predictive or prescriptive models. This ensures that analytics initiatives remain scalable, realistic, and effectively integrated into existing organizational workflows.
Bojer et al. (Sun,) studied this question.