AI as intern, not oracle: a design logic for human–AI partnership in experiential entrepreneurship education

Abstract Entrepreneurship education (EE) is increasingly extending across disciplines and institutional contexts, creating pressure to sustain the facilitation, reflection and contextual responsiveness required by experiential learning. At the same time, the rapid adoption of generative artificial intelligence (AI) in higher education raises questions about how AI can support such experiential learning without assuming authority over the judgement and decision-making central to entrepreneurial action. This study develops and empirically examines an AI-as-intern design logic for experiential EE. Rather than positioning AI as an authoritative “oracle”, the framework establishes an intentionally asymmetric human-AI partnership. Here AI provides prompts, alternatives, provisional representations and process support, while learners and educators retain responsibility for contextual interpretation, evaluative judgement and consequential decisions. The study examines two postgraduate entrepreneurship cohorts at a UK university ( n = 60) using a custom AI-enabled platform integrated into normal teaching. Evidence comprised session recordings and transcripts, observational field notes, a post-workshop survey, and 12 semi-structured interviews. Findings show that students used AI primarily as process scaffolding and iterative support. Learners accepted, rejected, modified and combined AI-generated material, while team interaction created additional opportunities for scrutiny and contextual evaluation. Our findings do not establish improvements in students’ judgement, creativity or cognitive load, but do show how these activities remained active within a bounded human-AI arrangement. The paper contributes a pedagogic design logic for AI-enabled experiential EE and provides exploratory propositions for the design of human-AI pedagogy concerning human authority, evaluative judgement, collaborative oversight, creative sequencing and bounded scaffolding under conditions of scaling pressure.