The increasing integration of artificial intelligence (AI)–based systems into professional environments has intensified the need for systematic pedagogical analysis of their functions in workplace learning and professional adaptation. Existing research often conceptualizes AI in education at an abstract level or evaluates its impact through human-centered studies, while the pedagogical functions embedded within AI systems themselves remain insufficiently examined in a systematic empirical manner. This article addresses this gap by treating AI-based systems as empirical pedagogical artifacts and examining their functional properties through comparative content analysis. The empirical basis of the study consists of a comparative analysis of five explicitly defined AI-based systems: ChatGPT, Gemini, Grok, Microsoft Copilot, and AI modules embedded in corporate Learning Management Systems (LMS). The study employs comparative content analysis and functional–pedagogical analysis, without involving human participants, surveys, or experimental interventions. A pedagogical function matrix was developed to analyze diagnostic support, feedback and formative assessment, support for reflection, individualization of learning and adaptation trajectories, pedagogical guidance and scaffolding, as well as pedagogical limitations and risks. The results reveal significant variation in how pedagogical functions are operationalized across AI systems. General-purpose generative AI systems demonstrate strong dialogic feedback and reflective support but lack persistent individualization mechanisms, while corporate LMS AI modules provide structured diagnostics and adaptation pathways with limited dialogic interaction. Microsoft Copilot exhibits embedded pedagogical functions within work processes, whereas Grok shows minimal pedagogical structuring across most functions. The findings underscore that AI-based systems are pedagogically heterogeneous and cannot be treated as functionally equivalent in professional adaptation contexts. The study contributes a replicable methodological framework for empirical analysis of AI systems as pedagogical agents and informs future research on AI-supported workplace learning.
Ovchinnikov et al. (Mon,) studied this question.