Abstract
Generative artificial intelligence is increasingly embedded in higher education, yet limited research has examined how learner agency is enacted within structured human–AI interaction. Existing studies have focused primarily on adoption, perceptions, or learning outcomes, offering less insight into how students act within AI-mediated pedagogical settings. This study investigates how learner agency is accomplished in a voice-based generative AI sales simulation in management education. Drawing on an agency lens, we analyse over 160 pages of student–AI voice dialogue transcripts and follow-up interviews to examine how students respond to staged task progression, structured resistance, and embedded evaluative cues.
Adopting a qualitative, process-oriented design, we examine agency as observable interactional conduct rather than as self-reported intention. Findings indicate that agency is procedurally bounded yet variably enacted. Under identical algorithmic conditions, some students intensify initiative, adapt questioning strategies, and refine judgement across iterations, while others remain reactive and display limited progression. Repetition alone does not produce development. Instead, learning trajectories depend on how students interpret and mobilise the structuring logic of the system.
The study advances theoretical understanding by conceptualising learner agency in AI-mediated contexts as interactionally realised within digitally organised fields of action. It further provides design implications for management education by identifying how sequencing, calibrated resistance, and intelligible feedback signals shape opportunities for professional reasoning and skill rehearsal. These findings shift attention from questions of AI adoption towards the pedagogical configuration of structured dialogue in generative AI-supported learning environments.