Risk-Aware but Unprepared: Business Students, AI-Assisted Writing, and Lessons from a Professional Verification Failure

Abstract

The integration of artificial intelligence tools into document preparation has created unprecedented opportunities and risks for business professionals and students. This study examines the educational implications of AI-assisted writing through a mixed-methods design combining a quantitative survey of 50 international business students with semi-structured interviews of 10 students, situated alongside a critical examination of a high-profile professional AI failure. Quantitative results show that students hold strong, internally consistent risk-aware perceptions of AI-generated legal business text (Risk Awareness subscale M = 4.27, α = .73, t(49) = 6.11, p < .001) and qualified trust in its utility (Trust/Substitution subscale M = 3.84, α = .78, t(49) = 3.14, p = .003), while general awareness of AI's capabilities remains only moderate (General AI Awareness subscale M = 3.47, α = .87, t(49) = 0.57, p = .571) and substantive reliance on AI for actual legal drafting is significantly and strongly avoided (Substantive Reliance subscale M = 2.27, α = .83, t(49) = −8.59, p < .001, d = −1.21). Interview data corroborate and explain this pattern: students report deliberate avoidance of AI for high-stakes drafting due to accuracy concerns, alongside near-total absence of structured instructor feedback on AI use. A parallel analysis of the January 2026 West Midlands Police case, in which AI-generated fabricated evidence informed operational policing decisions and precipitated institutional crisis, illustrates the real-world consequences of the verification failures that students' own risk-aware attitudes already anticipate. Together, these findings indicate that the central educational gap is not attitudinal but structural: students possess the disposition toward caution but lack the formal training and institutional feedback loops needed to convert that disposition into reliable verification practice. The paper proposes a framework for AI literacy education integrating technical competency, critical evaluation skill, ethical reasoning, and legal understanding, and argues that curricula must close this gap before students enter professional contexts where verification failures carry institutional, not merely academic, consequences.