Smarter Business ESP with Generative AI: From Generic Role‑Plays to Case‑Based, Domain‑Authentic Lesson Design

Abstract:

This paper examines how generative artificial intelligence (GenAI), particularly large language models (LLMs) such as ChatGPT can support lesson planning and activity design in Business English for Specific Purposes (Business ESP). Based on classroom implementations with cycles of 100 undergraduate business learners, I report mixed outcomes: AI generated role plays accelerated preparation and sometimes boosted engagement, yet frequently lacked domain authentic constraints and situational detail (e.g., stakeholders, metrics, risk, and policy). Synthesizing recent research (2023–2025) on GenAI in language education and Business ESP, I identify three recurrent failure modes—genericity, instructional opacity, and domain slips and propose a three layer design framework that combines (1) structured prompt engineering for specificity, (2) teacher authored case scaffolds with verifiable data, and (3) targeted use of complementary AI tools (adaptive feedback, corpus assisted term mining, multimodal simulation). Business focused lesson vignettes illustrate how the framework transforms “thin” role plays into realistic scenarios (procurement negotiation with conflicting KPIs; stakeholder email chains; crisis briefings with media Q&A). An evaluation protocol and rubrics (authenticity, clarity, complexity, ethics, evidence backing) are provided, with responsible use guidance aligned with UNESCO’s 2023 recommendations and the EU AI Act’s AI literacy obligations. Findings support the view that GenAI augments but does not replace pedagogical expertise; its value depends on careful integration and ongoing validity checks for accuracy, fairness, and academic integrity.