Leveraging Generative AI Through Vibe Coding: A Case of Simulation-Based Curriculum Redesign in Management Education

Abstract

Generative Artificial Intelligence (GenAI) tools such as Large Language Models (LLMs) have uncovered new possibilities for educators to develop interactive learning resources, yet practical guidance on harnessing these capabilities remains limited. This paper examines how GenAI and LLMs can support curriculum redesign through their capabilities in generating and testing code. We present a case study of a postgraduate operations management course redesigned to include simulation-based learning, to strengthen analytical and decision-making skills. The case demonstrates how a replicable prompt-driven workflow can enable educators to co-create HTML/JavaScript simulations with GenAI without programming expertise. Simulation-based learning has an established evidence base for improving student learning outcomes; the challenge has been accessibility. This paper’s contribution is not to re-validate that evidence, but to demonstrate that pedagogies once beyond the reach of resource-constrained educators are now meaningfully accessible through GenAI-enabled vibe coding. We document the design and implementation of this approach, and the opportunities and constraints encountered, to provide a practical roadmap for educators seeking to close the gap between what simulation-based education can offer and what has historically been possible to deliver.

1. Introduction

Since the early 21st century, the focus of tertiary education has gradually shifted from traditional transmissive pedagogy, where teaching involves a lecture system (Klein, 2008; Knight et al., 2023) to the development of transferable, work-relevant skills (Fulmore et al., 2023; Hilton & Pellegrino, 2012). In parallel, recent advances in GenAI offer educators new opportunities to overcome some of the constraints and technical barriers that have historically limited pedagogical innovation (Mollick & Mollick, 2024; Sharma, 2025). Our contribution illustrates how GenAI can operationalize simulation-based learning, a pedagogy long recognized for its value but often limited by cost and technical complexity (Hallinger & Wang, 2020; Pasin & Giroux, 2011). We further extend this work by providing a systematic, replicable workflow for educators underpinned by design thinking principles and operationalizing these principles into four prompting patterns, enabling educators with little programming experience to move from initial concept through to verified simulations that can be embedded into learning management systems for use in learning activities.

Despite this potential, GenAI adoption in education has been slow, constrained by limited policies, training, and practical guidelines (Kumar et al., 2024; Michel-Villarreal et al., 2023). Current discourse often emphasizes ethical concerns and academic integrity (Chan, 2023; Itani et al., 2025), sometimes at the expense of exploring GenAI’s role as an enabler of proven pedagogical innovations (Sharma, 2025). Efforts to integrate GenAI must therefore be context-sensitive, well supported, and aligned with human values to foster critical thinking and AI literacy (Verboom et al., 2025; Walter, 2024). In our context, simulation-based education has considerable alignment with operations management (Pasin & Giroux, 2011) and thus provided grounding for our work.

Simulations have a well-established record of facilitating learning and preparing students for professional practice (Jossberger et al., 2022; Kageyama et al., 2022; Keys & Wolfe, 1990). Historically, however, their adoption has been constrained by financial and technical barriers, limiting integration into mainstream curricula (Ionescu-Feleagă et al., 2025). Recent work in medical education demonstrates how LLMs can assist in developing digital simulations (e.g., Chow & Ng, 2025), making this approach more accessible. Building on this, our study leverages the emerging practice of vibe coding, a natural language, prompt-driven workflow that enables educators to generate executable code and artefacts without programming expertise (Barzanji & Loitsch, 2025; Karpathy, 2025). These simulations were implemented in a postgraduate operations management course at a regional Australian university across three non-sequential trimesters, each attracting a diverse cohort of primarily international students balancing study and work commitments.

This paper addresses the following research question: How can GenAI be leveraged through a structured prompt-driven workflow to enable simulation-based curriculum redesign in a resource-constrained higher education context? To address this question, we employ a design-based research (DBR) approach (Barab & Squire, 2004; Hoadley & Campos, 2022; Sandoval & Bell, 2004), situating our study within a postgraduate operations management course at a regional Australian university across three non-sequential trimesters. Our DBR approach involved the creation of a theoretically grounded intervention in a naturalistic context and iterative refinement across multiple course iterations (Barab & Squire, 2004). We leveraged existing research in simulation-based learning as a valid means of developing analytical and decision-making skills through constructivist, active learning principles (Chernikova et al., 2020; Kageyama et al., 2022). Notably, the research establishing the value of simulation-based education is borrowed rather than tested here. This study contributes a structured GenAI-enabled workflow that can make proven but previously resource-intensive pedagogies (such as simulation-based education) accessible to educators without specialist technical skills or dedicated institutional funding. The primary contribution of this study is a design process output (Hoadley & Campos, 2022), referred to here as the SRVE framework, as a replicable methodology for GenAI-assisted artefact development. Reflections were collected from the educators throughout the three course iterations, and informal post-implementation conversations with the tutors who co-delivered the face-to-face sessions. These reflections were reviewed to identify recurring observations and patterns across the three course iterations, with the contribution directed at other educators seeking to replicate or adapt this approach. Together, these methodological choices define the nature of this paper’s contribution: a conceptual, practice-oriented design and implementation study. Its contribution is not to re-examine whether simulation-based learning improves student outcomes; that evidence is well established (Chernikova et al., 2020). Rather, we address the more fundamental constraint of educator accessibility. Simulation-based learning has long been recognized as pedagogically valuable yet practically out of reach for educators without specialist technical skills or dedicated funding (Pasin & Giroux, 2011). By demonstrating how GenAI-enabled vibe coding can close that gap, this paper offers a replicable roadmap for educators in resource-constrained contexts. Our approach prioritizes human oversight and contextual validation, aligning calls to frame AI as an enabler of educational transformation rather than a threat (Walter, 2024).

2. Background and Literature: GenAI as an Enabler of Educational Practice Transformation

Higher education has previously experienced how new technologies offer new possibilities for learning and teaching. Recently, GenAI has been touted as having great potential to enhance education practice (Zhu et al., 2026). Along with demanding new literacies from educators, GenAI opens up possibilities for the adoption of a wider range of pedagogical approaches, enabling adopters to rethink and reconfigure how higher education is delivered (Clegg & Sarkar, 2024; Lim et al., 2023; Walter, 2024). A critical gap persists in providing practical and scalable use cases that demonstrate how GenAI can translate pedagogical potential into measurable improvements in diverse higher education settings.

With the potential of GenAI comes the need for careful exploration and consideration of the challenges, both known and unknown, and how GenAI will impact education practice (Cotton et al., 2024; Mollick & Mollick, 2024). Throughout the history of educational technology adoption, there has often been insufficient consideration for how educators implemented such resources (Rudolph et al., 2023). Educators are often accused of being overly conservative, resistant to change, and quick to adopt the precautionary principle regarding new technologies (Marks & Al-Ali, 2022). Often the literature reports on the risks of GenAI, such as overreliance and dependency leading to misuse (Polizzi & Harrison, 2022; Zhang et al., 2024), ethical concerns (Itani et al., 2025; Ravi et al., 2025; Verboom et al., 2025), and academic integrity (Chan, 2023). However, most of these risks can be effectively mitigated if implementation is supported and informed by teaching and learning theory (Crawford et al., 2023) and educator reflective practice (Chan, 2023). Further, effective implementation of GenAI for education purposes requires human engagement at its core (i.e., a human in the loop), rather than considering it an ‘off the shelf’ tool to simply be implemented without oversight or scrutiny (Yang et al., 2024). Fortunately, the literature on GenAI best practice in education is gathering speed and momentum (Pang & Wei, 2025; Ravi et al., 2025), revealing some general use cases. For instance, educators delegate mundane, formulaic tasks to GenAI tools and free up time and energy to be invested in developing new educational designs and teaching practice enhancements (Kumar et al., 2024; Ritter et al., 2024).

From the institutional perspective, top-ranking universities invest in and actively encourage educators to use GenAI tools (Wang et al., 2024), while other educators face limited resources and guidance (Barrett & Pack, 2023), inhibiting adoption and integration into teaching practice (Ritter et al., 2024). For the latter cohort, there is a need for practical advice on the ways in which GenAI can be leveraged to enable curriculum redesign. Our case study is situated in the latter context: a regional Australian university where the initiative was entirely educator-led, without dedicated institutional support or funding. Fortunately, GenAI tools have begun to erode traditional resource constraints, offering educators practical pathways to implement innovative designs without extensive technical expertise or financial investment. However, guidance on how to leverage these affordances effectively remains limited. In our teaching context, GenAI was seen as a tool to help shift away from traditional lecture-based teaching and towards a more skills-focused form of education.

3. Curriculum Redesign Rationale

3.1. Emphasizing Skills in Operations Management Education

Operations management has traditionally been taught using verbal knowledge transfer and Socratic debates (Pasin & Giroux, 2011) coupled with the presentation and application of mathematical models to find solutions to problems. Such models of teaching practice often manifest as a ‘sage on a stage’ approach, where the educator is positioned as the central authority imparting knowledge to passive learners (King, 1993), an approach that is well understood to be inferior to more active forms of learning (Bonwell & Eison, 1991). This notion is amplified by the rise of GenAI tools, specifically their capability to provide instant, context-sensitive access to information, offering learners a more accessible model of information retrieval than the sage on a stage model (Clegg & Sarkar, 2024; Sharma, 2025). Research points to GenAI tools being best used as a teaching supplement, rather than a replacement (Kasneci et al., 2023). As such, the educator’s role extends beyond subject matter expertise, towards the leveraging of emergent tools to create conducive learning environments (Crawford et al., 2023) and enhance the student experience (Ritter et al., 2024). These pedagogical critiques and opportunities formed the basis of our curriculum redesign, prompting a shift toward active, skills-based learning supported by GenAI.

Research points to GenAI tools as capable of automating routine tasks, allowing educators to redirect efforts towards assuming different roles (Ratten & Jones, 2023). For instance, the educator can contribute adaptive, ethical, and collaborative capacities and skills that traditional assessments are ill-equipped to measure (Mollick & Mollick, 2024). The work of Loaiza and Rigobon (2024) echoes such notions, pointing to human capabilities related to, for example, judgement, opinion, complex problem solving and critical thinking as not replaceable by GenAI. As such, we felt these human capabilities can serve as the target for our curriculum redesign. To develop these skills, we aimed to create learning activities that would emphasize these skills in an operations management context. Several fields of study have acknowledged the challenge of integrating both academic studies and the training of students to enter professional practice (Lehtinen, 2023). To address this challenge, we rebalanced the course to decrease (but not eliminate) the time spent delivering content about operations management and increased the time spent on the development of skills often required by operations managers in the workplace. Such a shift necessitated the underpinning of existing teaching and learning theories. In this space, we employed the work of Biggs (1996) on constructive alignment and the use of simulation-based education (Jossberger et al., 2022; Vermunt, 2023).

Our adoption of GenAI began with careful exploration and upskilling, and later as a tool to assist in curriculum redesign. Our cautious and purposeful use of GenAI tools was essential for inducing critical thinking, insightful reasoning, and other problem-solving higher-order skills in our students (in alignment with Kumar et al., 2024). Through our own upskilling in GenAI technologies and capabilities, we learned how it provides increased access to knowledge (as noted by Ratten & Jones, 2023), and reflected on our roles as educators. These reflections resonated with the notions of GenAI as a transformative opportunity in management education described by Dissanayake et al. (2024). Our experimentation with different teaching approaches eventually led us to settle on leveraging GenAI tools to code digital simulations as the focus of curriculum design.

3.2. Simulation-Based Learning

Simulation-based learning can serve to educate students and prepare them for work settings where the consequences of decisions are too high to allow for real-world experimentation (Jossberger et al., 2022; Vermunt, 2023). In its most basic form, a simulation is a model, a simplified representation of reality (object, system, or situation) with parameters that can be controlled or manipulated. When used in education, these models can provide a better understanding of the connections between variables in a system or to put different strategies to the test (Chernikova et al., 2020). Such simplifications are an important component to enabling learning, allowing students to interact with a simplified model of reality (Grossman et al., 2009; Vermunt, 2023). In operationalizing these simulations into our teaching and learning settings, the design element of the simulation itself was considered. For example, some simulated experiences can be concise, last a relatively short time and be highly structured, while others may be more involved, require users to make multiple decisions and be set in more complex and unstructured contexts (Bigelow, 2004; Heitzmann et al., 2023; Keys & Wolfe, 1990). However, simulations can also work against student learning, particularly if implemented in ways that are misaligned with student needs, either through poor user interface design and lack of guidance, overemphasis on high fidelity, or ignoring differences in student preferences (Davies, 2002).

To be effective as tools for learning, simulations should also be grounded in sound teaching and learning theory. For our curriculum redesign, we chose constructivism, a pedagogical theory that suggests humans actively construct knowledge and meaning through their lived experiences (Bada & Olusegun, 2015). Constructivism can be operationalized through learning activity designs that emphasize active learning and experiential learning (Beatty et al., 2021; Skritsovali, 2023). Active learning promotes teaching approaches that prioritize student participation over the passive transfer of an educator’s subject knowledge (Macvaugh & Norton, 2011). It emphasizes learner agency as a way to boost motivation, critical thinking, and the reflective skills needed for lifelong learning (Skritsovali, 2023). Similarly, experiential learning posits that learning is a process of constructing knowledge through experience (Kolb & Kolb, 2012). These pedagogical foundations are well supported in the management learning literature (e.g., Kluge, 2007) and informed our simulation design and learning activities. In operations management education, simulations and games have long been valued for their ability to support constructivist learning, yet their adoption has been limited by cost and development complexity (Pasin & Giroux, 2011).

4. Educator Blueprint for Simulation Design

4.1. Vibe Coding and Simulation Development

Recent developments in GenAI have challenged established approaches to teaching and learning, prompting educators to reconsider how curriculum and pedagogy can remain relevant (Walter, 2024). The emergence of GenAI has made capabilities such as coding more accessible to individuals with limited programming expertise. LLMs enable the creation, testing, and explanation of code in accessible ways, reducing barriers to implementing interactive learning activities (Moundridou et al., 2024). Although concerns about over-reliance on GenAI remain (Zhang et al., 2024), its capacity to replicate and scale code across contexts offers a practical capability for educators. On this basis, we integrated coding and simulation-based activities as part of our pedagogical adaptation.

In our work, ChatGPT 4, Gemini 3.1 Flash, and Claude Opus 4.5 LLMs were used for routine code generation, while the educator guided pedagogy, interpretation, iteration, and accuracy (Clegg & Sarkar, 2024). To generate the code for our digital simulations, we utilized a technique known as vibe coding (Adam et al., 2025; Ehsani et al., 2025; Karpathy, 2025), whereby an LLM was used to generate code prompted by natural language prompts and conversational code generation. Vibe coding represents a departure from writing code line by line, instead leveraging GenAI to produce the code based on natural language prompts and iterative refinement instructions provided by the developer. The informal nature of vibe-coding emphasizes spontaneous idea exchange, dialogic feedback and shared problem-solving between humans and the LLM (Barzanji & Loitsch, 2025). In the practice of applying vibe coding in education, Chow and Ng (2025, p. 1) noted “a shift towards a fluid, creative coding experience, where educational goals take precedence over technical barriers.” Vibe coding can foster motivation, critical thinking and engagement (Barzanji & Loitsch, 2025). Furthermore, by reducing technical barriers, vibe coding can allow educators to invest greater attention to the activity design process and activity alignment with constructivist principles.

Vibe coding, in its original conception, describes an informal mode of interacting with LLMs to produce code and one that tolerates imprecision in favor of creative momentum (Chow & Ng, 2025; Karpathy, 2025). However, we acknowledge that application in an educational design context demands greater structure and accountability. In this paper, we operationalize vibe coding through design thinking principles, applying a more structured approach. The result is the SRVE framework: a structured, replicable workflow that preserves the accessibility and iterative spirit of vibe coding while embedding the rigor, verification, and contextual alignment required of educational artefacts. To establish structure, we followed a design thinking approach (e.g., Bathla et al., 2025), attempting to match the needs with what is feasible (Brown, 2008). We adopted the Hasso Plattner Institute of Design at Stanford (2010) five-stage design thinking process (Empathize, Define, Ideate, Prototype, Test) as a framework for our curriculum redesign. In practice, this meant beginning with empathy for our learners’ challenges, defining the core instructional problem, ideating solutions with and through GenAI, and rapidly prototyping and testing digital simulations via iterative prompt refinement. The number of iterations varied widely, based on the complexity of the simulation and the ability of the LMS to render the simulations. Each cycle of prototyping and testing generated new insights, which were immediately fed back into the next round of prompt development, ensuring that both the simulations and the learning experience were continually improved in response to user feedback. Development times for simulations varied (between 5 and 50 h per simulation), largely based on LLM rate limits and educator prompting capability. We suspect that as LLMs become more capable, these times will be reduced dramatically. In early 2026, Gemini 2.5 Pro was able to create a working simulation in a single prompt in less than 10 min. We summarize the SRVE framework in Figure 1.

Table 1. Mapping of design thinking stages to actions and SRVE prompting patterns.

Design Thinking Stage	Actions in This Case	SRVE Stage
Empathize	Recognised that students’ time was the primary resource at stake, and that traditional knowledge-transmission teaching was an inefficient use of it. Reframed the educational problem as one of skills development: what students could do mattered more than what they knew.	Pre-SRVE
Define	Identified simulation-based learning as the target pedagogy—sufficiently close to existing pedagogical knowledge to be implementable, and well-supported in the operations management education literature. Third-party simulations were ruled out due to cost and licensing constraints; traditional coding ruled out due to technical barriers and absence of dedicated resources.	Pre-SRVE
Ideate	The emergence of vibe coding was recognised as an enabling opportunity—a newly accessible capability that made simulation development feasible without specialist technical knowledge or funding. This recognition, informed by pre-existing domain expertise, collapsed the ideation process into a single, well-grounded insight rather than a formal divergent options exercise.	Pre-SRVE
Prototype	LLM-assisted code generation using iterative natural language prompts, structured through the four SRVE prompting patterns.	Specify → Refine
Test	Simulations trialled with academic colleagues prior to implementation; refined iteratively across three non-sequential trimesters based on LMS rendering issues and tutor feedback.	Verify → Embed

maps each stage of the design thinking process to the specific actions taken in this case and to the corresponding SRVE prompting pattern. Notably, the first three stages—Empathize, Define, and Ideate—constitute a pre-SRVE framing phase, in which the pedagogical problem is understood and the design direction established before any prompting begins.

Our use of the GenAI tool employed four reusable patterns:

• Specify—prompts that declared the operations management model, inputs/outputs, constraints, and performance metrics. For example: create a turn-based simulation model for use in a university education setting that models inventory management, use the economic order quantity model as a basis, expect order quantity inputs from students and output financial performance metrics.
• Refine—prompts related to refining the user interface and usability, as well as input ranges and units. For example: A user interface should include a space to input order quantities (with specified maximum and minimum values), operational parameters such as holding and ordering costs (in dollars), performance graphs showing profit (in dollars) and service level (in percentage demand met) per turn, and a historical table of relevant inputs and outputs per turn.
• Verify—prompts containing boundary tests (including edge cases), seed randomization, and equation checks using manual calculation procedures to ensure rigor in calculations. If verification revealed errors, feedback was provided to the LLM and iteration continued until errors were resolved. For example: Test 10 different input values (including maximum and minimum values) and output the performance outcomes based on the stated equations and calculation algorithm, I will check these against manual calculations and tell you if any errors arise.
• Embed—producing the HTML and JavaScript code that is suitable for embedding into learning management system (LMS) content blocks. For example, if an element did not display properly, we provided a screenshot and used prompts such as the graphs are not displaying properly, see the following screenshot, diagnose the problem, and refactor the code accordingly.

Recent research found that engineering students struggled to effectively generate and debug LLM-produced code, highlighting the importance of domain expertise in the prompting process (Vieira et al., 2025). Our experience corroborates this finding: the SRVE framework’s effectiveness depended on the educator’s ability to specify disciplinary models accurately (Specify), evaluate outputs against known standards (Verify), and diagnose rendering or logic failures informed by pedagogical intent (Embed). This supports the broader argument that GenAI-assisted educational design is most effective when driven by educators who combine domain knowledge with pedagogical purpose.

The GenAI-produced HTML and JavaScript code was embedded in webpages hosted in the LMS (Moodle) and accessible to students and teaching staff via standard web browsers without additional software installations. In incorporating these resources into workshops and tutorials, face-to-face sessions shifted away from knowledge-centric problem-solving learning activities to simulation-based skills development and decision making. The simulations were implemented within a wider portfolio of learning methods (Beatty et al., 2021), after the presentation of pre-recorded lectures, practice quizzes, feedback on practice quizzes, and case studies. Each simulation was accompanied by a lesson plan, highlighting the inputs and decisions students should produce and learning objectives for each face-to-face session and discussion questions. A GitHub repository, containing the complete simulation code, lesson plans, and implementation guidance, is available as Supplementary Materials at https://github.com/amunozUOW/Operations_Management_Simulations (accessed on 17 March 2026).

Prior to implementation, the simulations were trialed with academic colleagues to seek feedback and adjust before embedding them in syllabi and presenting them to students and the tutors who were delivering the face-to-face sessions. The simulations were refined to reflect typical operations management problems, enabling students to make decisions with or without mathematical models in a controlled classroom environment. Educators facilitated these activities by prompting students to justify their choices and engaging them in structured discussions at key stages of the session. For example, when setting inventory levels, students were asked questions such as “Why did you choose that level?” or “Do you think this will be sufficient?” These prompts encouraged reflection on decision-making processes and supported iterative experimentation. After entering decisions into the models, students analyzed outcomes, compared alternatives, and repeated the process to deepen analytical skills often emphasized in management education (e.g., Kageyama et al., 2022) and practice (Sollosy & McInerney, 2022). Tutors reported that the simulations led to greater ease in creating a whole of class discourse, through an inquiry-based approach to the analytical skills targeted in each tutorial. Further, after the face-to-face session, students could still access the simulations via the LMS and continue their learning outside the classroom.

4.2. GenAI-Enabled Simulation-Based Learning for Critical and Analytical Thinking

In many operations management practical settings, decisions are underpinned by knowledge (underpinning the importance of knowledge retention and domain understanding) and complemented by whether a decision applies to the operation (underpinning the importance of theoretical application in each context). To ensure constructive alignment between subject-level and program-level objectives, the operations management subject was designed with explicit mapping of its learning outcomes to the broader course learning outcomes. The four subject learning outcomes—assessing and applying operations management theories (SLO 1), analyzing the role of inventory (SLO 2), communicating operations management principles (SLO 3), and applying quantitative techniques to assess performance (SLO 4)—each map to course-level outcomes spanning specialized business knowledge, integrated problem-solving, professional communication, and research capability. This alignment ensures that the simulation-based activities contribute to subject-specific competencies and to the broader graduate profile.

We designed many of our learning activities to elicit decisions from students, which required recall of operations management content coupled with an understanding of the simulated context. The resulting decisions were tested against the simulations to ascertain their effectiveness (via an inspection of performance outputs). By manipulating variables and observing performance outcomes, students received feedback on their decisions (Dittrich et al., 2022; Lehtinen, 2023). The intention of each exercise was to allow students to hone and develop their ability to leverage output to make informed and insightful decisions, sharpening analytical skills (Sollosy & McInerney, 2022). These analytical skills can be further broken down into four basic types of analysis: descriptive, diagnostic, predictive, and prescriptive (Rose, 2016). Using this typology of analytics to inform the discourse during face-to-face sessions allowed for a framework that could be consistently applied across all simulations. Each analytics type was linked to basic questions used to spur discourse during face-to-face sessions, as well as being linked to Bloom’s taxonomy verbs to ensure consistency with learning objectives (see

Table 2. Analytics categories of structured discussion questions used during operations management learning activities.

Category	Description	Example Question	Bloom’s Taxonomy Level
Content recall	Requires recalling and stating content previously covered.	What is the definition of concept X?	Remember
Calculation	Performing step-by-step mathematical procedures to arrive at a precise solution.	Calculate the value of X using the given formula.	Apply
Descriptive analytics	Involve identification of trends, describing characteristics, or articulating current state.	Describe any trends in the performance data.	Understand
Diagnostic analytics	Analyze and identify/diagnose root causes, behind observed outcomes or data patterns.	Why did product X sales spike during the holiday season?	Analyze
Predictive analytics	Making predictions, forecasting future trends, or evaluating outcomes based on evidence.	If the operation is managed in X manner, what is the most likely outcome?	Apply
Prescriptive analytics	Problem questions that elicit prescriptive courses of action, strategies, or solutions based on analysis of data and settings.	Given the data provided, what level of inventory would be appropriate to maximize profits?	Evaluate/Synthesize

).

The week-to-week structure of the course presented students with foundational content in the form of pre-recorded lectures, textbook chapter content, and practice quizzes. This content was assigned to students as ‘pre-work’ to be completed prior to the face-to-face sessions. Each face-to-face session featured a different simulation; each built from existing operations management theories and models featured in the foundational content and exemplified how operations following similar rules and behaviors performed. The simulations place students in risk-free scenarios where they repeatedly make operations decisions and evaluate the outcomes of individual decisions, receiving feedback that helps develop the causal interplay between decisions and performance outcomes. The iterative nature of these decisions is important as it aligns with stimulus-response theory (Knowles et al., 2005), where repetitive use of a skill (e.g., decision-making) leads to its development. To prevent over-repetition, multiple versions of each simulation were developed with different parameters (e.g., higher and lower demand scenarios) and scenarios.

Each simulation design followed guidance by Kageyama et al. (2022), including a case statement introducing the hypothetical problem being faced by an operations manager, followed by instructions and operational parameters (e.g., demand levels, starting cash-on-hand), performance metrics, and goals. Students input decisions as parameter selections via sliders and input boxes with time series charts and tables illustrating the relevant parameters. Furthermore, each simulation was interactive, requiring students to provide inputs to the simulation and then analyze the response provided after a ‘Next Turn’ or ‘Submit’ button was clicked. As our operations management course progressed, the simulations became more involved and represented more complex models. A summary of the portfolio of simulations, their corresponding operations management, theoretical content, and their objectives is provided in

Table 3. List of simulations developed, alignment with content, and learning objectives.

Week	Simulation	Content	Aligned SLO	Objective(s)
1	Number Guessing Game	None	4	Introduce students to the importance of decision-making in operations. Acquaint students with the analytical and critical thinking required to manage operations.
2	Productivity and Performance (Food truck simulation)	Introduction to operations management Operations performance	1, 3, 4	Evaluate performance (single and multi-factor productivity) and performance metrics. Understand the role of decisions in driving operations performance.
3	Little’s Law (Coffee shop)	Process Design	1, 3	Link capacity, productivity, and Little’s Law in operations.
4	Staffing (Supermarket check-out)	Capacity Management	1, 3	Utilize forecasting to make capacity decisions.
5	Inventory Management (Vending Machines)	Inventory Management	1, 2	Understand how multiple decisions affect inventory and service levels.
6	Inventory and Lean Operations	Lean Operations	1, 2	Demonstrate use of Economic Order Quantity and how it contrasts with lean principles.
7	Supply Chains (Centralized beer supply chain)	Supply Chain Management	1, 3, 4	Manage inventory across multiple locations in a supply chain to ensure synchronized movement of stock and information.
8	Quality Management (Staff and Stock management)	Quality Management	2, 3, 4	Understand and balance inventory and staffing decisions to ensure customers are served to expectations.
9	Project Management (Community Garden project)	Project Management	3, 4	Understand how dynamic allocation of resources affects budget, schedule, and work completion rates.

. Each simulation was accompanied by a lesson plan, provided only to the tutors delivering the face-to-face session. Each lesson plan included background, learning objectives, step-by-step instructions for using the simulation, and potential discussion questions.

5. Reflections on Implementation

5.1. Educator Insights

We anticipated that introducing simulations would require careful scaffolding, as many students initially expected a traditional “sage on the stage” approach (as described in Fischer & Dobbins, 2024). During the early parts of the course, students were somewhat hesitant to interact with the models and were expecting an educational experience with more knowledge transmission. To gradually introduce students to interacting with the simulation, we offered tutors the option to facilitate interaction by providing live demonstrations at the early stages of each tutorial. Over time, however, these expectations shifted as students began to articulate strategic reasoning during simulations. Positive informal feedback highlighted perceived relevance, with students noting the different approach being a more attractive form of learning and better aligned with their future workplace expectations. Overall, student engagement varied considerably, revealing diverse learning preferences. Some students embraced the simulations individually, exploring decision-making and performance outcomes autonomously or with little need for facilitation. Others preferred collaborative engagement, forming small groups where one student acted as the “driver” of the simulation while peers contributed inputs, interpreted outputs, and discussed strategies. A third group adopted an observational stance, paying attention to the tutor’s facilitation and monitoring interactions without active interaction with the simulations. While these patterns are anecdotal, they reiterate that simulation-based learning accommodates multiple engagement modes (Dittrich et al., 2022; Hallinger & Wang, 2020), aligning with constructivist principles of learner agency (Kageyama et al., 2022).

Tutors reflected that simulations transformed their teaching experience. Rather than delivering content, they adopted coaching roles. Post-implementation debriefs with tutors were important as insights into how tutorials became easier to run as students became comfortable throughout the trimester. Specifically, tutors reported observing students increasingly interacting with the simulations as a safe space for experimentation. This shift supported richer dialogue and critical thinking, consistent with literature advocating for active learning strategies (Bonwell & Eison, 1991; Kageyama et al., 2022). Tutors particularly valued the clarity of lesson plans accompanying each simulation, noting that these resources increased confidence in guiding discussions. However, they also acknowledged that preparation time was significantly greater, as moving away from established models required more comprehensive planning. This insight highlighted a vital trade-off: while the redesigned curriculum did not alter subject or course learning outcomes, it introduced new opportunities for students to practice analytical and decision-making skills in risk-free environments—benefits that tutors felt justified the additional effort.

5.2. Technical and Pedagogical Challenges

Despite these successes, technical robustness emerged as a significant challenge. LMS software updates disrupted simulation rendering on three separate occasions, necessitating rapid troubleshooting in either the HTML code or the JavaScript. Development and debugging were also more time-intensive than anticipated, particularly when addressing edge cases in calculation logic. These experiences revealed that scaling simulations beyond our own operations management context still require significant resource investment, not only for development but for ongoing maintenance. At the same time, once simulations were thoroughly tested and stabilized, their scalability became a clear advantage. Reusability across multiple course iterations reduced preparation workload and allowed tutors to focus on facilitation rather than technical setup, confirming arguments in the literature about the long-term efficiency of technology-enabled learning (Hallinger & Wang, 2020). This trade-off between upfront complexity and downstream efficiency was one of the most salient lessons from implementation.

This educator-led approach demonstrates that GenAI can enable greater access to simulation development, enabling innovation without dedicated institutional resources. However, it also highlights sustainability challenges and the need for institutional support structures to scale such initiatives. While autonomy fosters creativity, it exposes educators to risks that could be mitigated through institutional investment in training and infrastructure (Ravi et al., 2025). Upon reflection, we recognize that future initiatives would benefit from clearer governance structures and resource allocation to support scaling. GenAI enhanced confidence in designing innovative learning experiences, yet being the sole adopter of this application of GenAI in enabling simulation-based education within the institution created a sense of professional isolation. There were moments when we questioned whether the effort was sustainable without broader peer engagement or institutional support. This reflection underscores that innovation in this space can be both technical and cultural, requiring communities of practice to share insights, experiences, and normalize experimentation while avoiding the sense of isolation often felt when being one of the few who experiment with new technologies.

6. Discussion, Limitations, and Theoretical Contribution

Management education scholarship has called for critical discourse that can adequately manage the excitement and concern for GenAI and its impact on educational practice (Clegg & Sarkar, 2024; Lim et al., 2023; Ratten & Jones, 2023; Sharma, 2025). Such discourse is likely to be enriched by efforts to enhance GenAI proficiency among educators (Walter, 2024) to help realize the gains reported to date (See recent systematic reviews by Belkina et al., 2025; Zhu et al., 2026). However, it is important that such efforts are deliberate, supported by strong teaching and learning theory, and aligned with the educational context (Verboom et al., 2025). In our case, employing GenAI tools to generate HTML and JavaScript code to construct digital simulation models for operations management skills development was one of many ways to conduct a curriculum redesign towards simulation-based education. In doing so, we contribute to education practice by demonstrating how GenAI, operationalized through vibe coding and design thinking principles, can enable educators to develop learning experiences that bridge theoretical knowledge and professional practice. This approach also facilitated a shift from knowledge transmission to skill development, aligning with broader calls for future-ready education. Challenges remain, particularly in managing student expectations against institutional constraints while supporting tutor engagement and ensuring technical reliability.

Our implementation experience underscores that successful GenAI-enabled curriculum redesign requires educator agency in driving innovation, ongoing upskilling in AI capabilities, and a strong grounding in teaching and learning theory to ensure pedagogical rather than merely technical outcomes. Much of this thinking falls in line with the observations of STEM educators: the use of LLMs as coding assistants cannot be effective without substantial expertise (Vieira et al., 2025). As such, to successfully navigate a future where the use of GenAI tools is increasingly embedded in educational practice, educators require safe and supportive environments in which to experiment and leverage existing expertise to uncover the pedagogical potential of GenAI (Dissanayake et al., 2024). Such experimentation enables the testing and integration of new ideas and experiences, empowering educators to make informed decisions aligned with desired educational goals (Carvalho et al., 2022; Sharma, 2025). Furthermore, the collective contributions of use cases, when properly grounded in teaching and learning theory, hold significant promise for ensuring that higher education institutions continue to cultivate the skills students need to create value in their future workplaces.

This study has several limitations that bound the generalizability of the findings. First, as a single-institution case study focused on postgraduate operations management education, the applicability to other disciplines and institutional contexts requires further investigation. Specifically, the use of other forms of interaction with LLMs may yield easier access to implementing other pedagogies, a notion that was not explored in this paper. Second, the study does not measure student learning outcomes, and this is a deliberate rather than incidental feature of the contribution. The paper’s purpose is to document a design and implementation process and to demonstrate that simulation-based pedagogies, long supported by an established outcomes evidence base, are now accessible to educators who previously lacked the technical and financial resources to adopt them. Whether SRVE-structured vibe coding produces simulation artefacts of sufficient quality to realize those documented learning benefits, and whether the accessibility gains are sufficient to meaningfully expand the population of educators who can implement simulation-based learning, are important directions for future empirical research. The present study provides the naturalistic implementation evidence that grounds those future investigations. Third, the simulations were developed by educators with pre-existing interest in GenAI, and replication by educators less familiar with these tools may present additional challenges. Finally, the rapid evolution of GenAI capabilities means that specific prompting strategies and tool affordances may shift over time, requiring ongoing adaptation of the framework. These limitations point to the need for ongoing professional development and institutional support structures that empower educators to experiment safely and reflectively. Future research could explore how these simulations can be adapted to serve as assessment instruments, whereby the simulations themselves provide context from which analytical questions can be asked of students, further embodying the constructive alignment principles espoused by Biggs (1996) and applied in Pereira et al. (2024).

This case contributes to emerging literature on GenAI in education in three ways. First, it extends the application of vibe coding (Barzanji & Loitsch, 2025; Chow & Ng, 2025; Karpathy, 2025) from medical education to operations management education, demonstrating cross-disciplinary transferability. Second, it offers a replicable framework (Specify-Refine-Verify-Embed) that operationalizes design thinking principles for prompt-based simulation development. The framework’s contribution extends beyond disciplinary transfer: by integrating design thinking’s iterative prototyping with structured prompt patterns, it provides structure to a GenAI-enabled education design process. The iterative nature of the framework includes a verification phase that mitigates many of the well-documented concerns about GenAI reliability in educational contexts, while the embedding phase bridges the gap between prototype development and sustainable classroom implementation. Third, it responds to calls for educator-centered perspectives on AI integration (Zawacki-Richter et al., 2019), providing a practical roadmap grounded in tried and tested pedagogies.