Abstract
Artificial Intelligence (AI) systems embedded for efficiency also reshape organizational capacities like perception, judgment, learning, and authority, which AI evaluation frameworks often treat as stable. Drawing on a case study of a digital sleeper monitoring system in a Nordic rail organization, this paper examines how predictive AI shapes how organizations notice and respond to their operational environment. I identify four dynamics that narrow routines, conceptualized as algorithmic narrowing: (i) algorithmic filtering limits what the organization notices, (ii) cognitive recalibration limits how practitioners interpret, (iii) institutional lock-in limits what the organization learns, and (iv) accountability displacement limits what the organization can do with what it knows. The paper makes three contributions. It shifts the view of algorithmic limits from a question of fidelity to a question about process, maps this process across four core routines, and shows that procurement, incentive, and verification retirement decisions structurally connect the dynamics, placing them within AI governance.
Recommended Citation
Washik, Maryam, "Algorithmic Narrowing of Organizational Routines" (2026). 17th Scandinavian Conference on Information Systems. 5.
https://aisel.aisnet.org/scis2026/5