Abstract:

This exploratory study examines how engineering students develop the ability to work critically with generative AI when solving strategic management cases. We analyse 42 reflections from students of the Industrial Organization Engineering degree and the Master’s in Industrial Engineering at the Universitat Politècnica de València. Using Rolfe’s What–So What-Now What model, students documented (What) their prompts, iterations, and outputs from several large language models (GPT, Gemini, Copilot, Claude, DeepSeek, among others) while applying frameworks such as PESTEL, Porter, Ansoff, and Abell to real firms. In the So what stage, they evaluated the theoretical validity of AI-generated analyses and identified errors. Students generally perceived the tools as capable of reproducing core strategic concepts and correctly naming models, primarily when they provided lecture slides or highly structured prompts. However, they recurrently reported generic, overly verbose answers, superficial reasoning, and difficulties adapting the theory to less-known companies. Hallucinations appeared mainly in firm-specific data, competitors’ details, and in some cases, in the mixing of categories within analytical frameworks, which students considered critical for decision-making. In the Now what phase, they rated the initial strategic usefulness of AI outputs on a 1–5 scale (with most assessments clustering around 3–4) and articulated action plans. Across programmes, students converged on viewing AI as a starting point and “analytical copilot”, but never as a final authority. Proposed practices include systematic verification of facts using company reports and specialised databases, sourcing information from the LLM, triangulating answers across tools, and refining prompts by specifying models, format, length, firm context, and acceptable uncertainty. Several reflections highlight ethical and confidentiality concerns, arguing that teachers should explicitly train students in “critical AI literacy” rather than banning these tools. Overall, the findings suggest that guided use of Rolfe’s reflective model helps future engineers transition from an uncritical reliance on AI to a more nuanced stance that combines efficiency gains with human judgment, accountability, and deeper conceptual learning in strategic management.

Keywords:

Generative artificial intelligence, large language models, strategic management education, engineering students, critical reflection, Rolfe model, AI literacy, human–AI collaboration.