AI-supported scenario-based learning in entrepreneurship education: A comparative classroom study of learning processes and outcomes

Entrepreneurship education is often limited by the difficulty of providing students with authentic, consequence-driven practice. This study examines how different forms of AI-supported scenario-based learning shape student learning and teaching in an entrepreneurship course. Using a quasi-experimental, three-condition classroom design, we compared regular instruction, static AI-assisted interaction, and a multi-agent dynamic simulation. The study involved 42 students, comprising 30 undergraduates and 12 graduate students, and adopted a mixed-methods approach that integrated pre- and post-tests, instructor-evaluated performance, reflection logs, and interviews. Quantitative analyses were used to compare learning outcomes and mechanism-related measures across groups, while qualitative data were coded and thematically analyzed. Results indicated that the static AI support condition was associated with only modest benefits, whereas the multi-agent dynamic simulation condition was associated with stronger gains in entrepreneurial self-efficacy, opportunity competence, and instructor-evaluated decision quality, as well as higher perceived scaffolding, authenticity, and engagement. Qualitative findings further indicated deeper reasoning and reflection in the multi-agent dynamic simulation condition, while also revealing reduced interpersonal communication and greater ambiguity in instructors’ roles. Because this condition combined several instructional features, the findings suggest that the value of AI in entrepreneurship education depends on the design of the broader instructional package rather than on technology alone.