Abstract
Abstract
This article examines the complexity of university dropout in the context of digital transformation and growing labor market uncertainty associated with the rapid development of artificial intelligence. Designed as a conceptual and methodological contribution rather than an empirical study, it evaluates the diagnostic potential of institutional information systems, using the CloudA platform as an illustrative case, and proposes methodological improvements intended to support student retention. The first part reviews established theoretical models of student dropout and considers their relevance to the contemporary use of learning analytics. The central contribution of the article is an original classification of direct indicators available within institutional information systems, organized into academic, behavioral, and institutional dimensions. The paper also proposes composite indicators, including the Moment of Trouble and Financial Risk scales, which are intended to transform transactional data into measurable early warning signals. These indicators remain conceptual and require future empirical validation. The discussion identifies important limitations in the capacity of institutional information systems to capture psychosocial factors. It therefore emphasizes the need to combine data-based analysis with human-centered institutional support. The effectiveness of dropout prevention is likely to depend not only on the accuracy of analytical indicators, but also on the speed, appropriateness, and quality of the institutional response to the risks identified.