Abstract
Service organisations increasingly automate administrative processes to improve efficiency and overcome capacity constraints. Yet when automation proceeds in environments where infrastructure cannot be taken for granted, it may produce an unintended consequence: the gradual erosion of manual capabilities that organisations need most when systems fail. This paper examines the digital dependency paradox, defined here as the process by which automation can eliminate fallback capabilities in contexts where infrastructure failures make them indispensable. Through longitudinal analysis of administrative process automation at a Pakistani university over five years, drawing on 16 semi-structured interviews and archival records, we trace how deskilling under unreliable infrastructure appears to have created organisational brittleness rather than resilience. Infrastructure failures did not merely pause work; they eroded user trust, extending adoption timelines by approximately 18 months. Once adopted, the system substituted for cognitively demanding result compilation work but left staff with residual formatting tasks that required less domain expertise. Over time, manual compilation capabilities atrophied through disuse, and staff found themselves unable to revert to paper-based processes during system outages. The organisation thus became trapped in a cycle of dependency on a system it could not always rely upon, while lacking the capacity to function without it. We develop a framework linking infrastructure reliability, trust calibration, and skill transformation to explain how this dependency trap forms, and we identify conditions under which automation may enhance versus undermine service resilience. The findings extend service operations management theory by specifying mechanisms through which efficiency-oriented automation can compromise operational resilience, and they offer practical guidance for managing automation in resource-constrained service environments.