AI-Powered Unlimited Practice

Theme: AI Integration

Summary of Initiative

The AI-Powered Unlimited Practice initiative moves beyond chatbots by embedding generative AI into our core educational infrastructure. The system uses AI to develop massive, varied question banks for both quantitative and case-based courses. This scalable system provides students with unlimited practice in new business skills, enabling deep engagement and improved learning outcomes.

Call to Action for Initiative

The initiative was born from a persistent pedagogical challenge within our school’s core Business Mathematics course. While mathematical foundations are a cornerstone of success in business education, this course has historically been a major hurdle for many students. We observed concerning patterns of course failures, program delays, and even dropouts stemming from a wide variance in students’ mathematical preparedness. Some students arrived with gaps in foundational algebra, while others came from different educational systems entirely, creating a classroom environment where it was difficult to meet everyone’s needs.

Our first step was to develop a digital practice tool, the Algebra Quiz, designed to provide students with a structured way to strengthen these foundational math skills. Powered by Python, this system offered unlimited practice opportunities, instant feedback, and gamification elements like weekly leaderboards and personalized progress emails. The impact on student engagement was immediate and profound. While a previous entry test had garnered only 355 attempts over a semester, the new gamified quiz system saw over 1,188 attempts in just the first four weeks, demonstrating that we had created a highly motivating and effective learning framework.

This success reinforced our core pedagogical belief: True understanding, especially in technical subjects, is not built through passive learning but through active, persistent practice. The process of struggling with a problem, failing, and trying again in a low-stakes environment is where deep learning occurs. Our system had created a safe space for this essential cycle of trial and error, but we were determined that this opportunity should not be confined to a single preparatory quiz.

Despite this success, the initial model faced a critical limitation in terms of scalability. The system worked exceptionally well for simple calculation-based mathematics problems, where questions could be generated algorithmically using simple randomization techniques. However, we recognized that the same need for structured, unlimited practice existed in other core business courses, such as Global Supply Chain Management. These subjects rely on qualitative reasoning and complex, case-based questions, whose generation cannot easily be outsourced to a routine based on classical procedural programming. Expanding our successful model to these areas would thus have required an unsustainable amount of manual effort from faculty, creating a significant bottleneck to innovation.

The true inspiration for the initiative’s current form came from the recent breakthroughs in generative AI. We realized that large language models (LLMs) offered the perfect solution to our scalability problem. Modern LLMs understand context and are able to generate nuanced and varied settings and produce the exact kind of text-rich, analytical questions that were previously impossible to create at scale.

The vision is no longer about a single course but transforming our proven pedagogical framework into a schoolwide ecosystem. We were inspired to move beyond using AI as a simple chatbot and instead embed it into our core educational infrastructure, automating the most resource-intensive part of personalized learning—content creation—for any subject.

Description of Initiative

The integration of artificial intelligence in our project represents a fundamental redesign of how personalized learning content is created and delivered. Critically, the AI is not a student-facing chatbot; it is a powerful content-generation engine embedded into the backend of our educational platform. This strategic implementation allows us to automate the most resource-intensive aspect of teaching at scale—the creation of diverse, high-quality practice problems.

Our system leverages Claude AI’s application programming interface through a process of carefully crafted prompt engineering. For each topic, we design detailed templates that define the learning objectives, desired question format, and key variables. These templates act as blueprints, guiding the LLM to generate questions that are not only unique but also pedagogically sound and perfectly aligned with the course curriculum. The AI then uses these blueprints to populate vast question banks that are imported into the Canvas learning management system, ensuring that when students practice, they are drawing from a pool so large they effectively never see the same problem twice.

A key technical innovation in our approach is the implementation of “tool use” to ensure accuracy, a significant challenge for many LLMs. For quantitative problems in business mathematics and supply chain management, we do not rely on the LLM’s internal calculation capabilities, which are rather limited for precise calculation involving either large numbers or complex operations such as exponentiation and square roots. Instead, the model is instructed to generate the textual scenario of a problem and then call on external, trusted routines for any complex operations. The model then intelligently integrates the precise result back into the question or solution. This sophisticated method separates the creative text-generation task from the logical calculation task, guaranteeing the mathematical integrity of every problem.

This AI-driven approach is versatile and has been deployed across different disciplines. In the Business Mathematics course, it generates complex word problems for topics like linear programming, requiring students to translate real-world scenarios into mathematical models. In our Global Supply Chain Management course, the same system creates nuanced mini-case studies on inventory control, logistics trade-offs, or procurement strategies. This allows students to practice applying theoretical concepts to varied, realistic business situations—a critical skill that is notoriously difficult to teach with static textbook examples.

Another important aspect of the system is a sophisticated gamification framework inspired by language-learning app Duolingo’s highly successful engagement model. Students compete in intimate leagues (bronze, silver, gold, and diamond) of 15 peers. This fosters healthy, fun, and localized competition where successful students are promoted to the next league weekly, while less engaged students are moved to lower leagues, creating a powerful incentive to practice. A real-time dashboard allows students to track their standing, and personalized weekly emails update them on their progress. This structure transforms solitary practice into a dynamic social experience that elevates learning from passive reception to active, continuous engagement.

Impact of Initiative

The initiative’s most immediate and measurable impact has been a dramatic transformation in student engagement and study habits. Previously, static quizzes saw minimal engagement. With the new AI-powered system, quiz usage tripled in just the first four weeks of the course. More importantly, we observed a fundamental shift in how students approached learning. Instead of single attempts, many students engaged in repeated practice to master the material and improve their leaderboard rank. Some dedicated students attempted the 15-minute algebra quiz up to 30 times—a massive investment of time that proves the system’s success in fostering the very cycle of persistent practice that builds true understanding.

Qualitative feedback confirms that a large proportion of students thoroughly enjoyed this competitive, gamified element. The heightened engagement has correlated with a remarkable improvement in academic performance. In the Business Mathematics course, the failure rate on the final theory exam fell from a historical high of 62 percent two years ago to just 27 percent in the most recent year.

While this significant improvement is the result of several concurrent course enhancements, we are confident that the AI-powered practice system was a major contributing factor. It provided a structured, motivating, and endlessly varied environment for students to build the mathematical fluency that they previously lacked, directly addressing the core challenge that led to high failure rates.

The impact of the initiative extends far beyond a single grade; it has a profound effect on the student lifecycle. More students are now able to pass a critical gateway course on their first attempt. This prevents program delays and reduces the significant stress associated with repeating courses, which often conflicts with heavy workloads in subsequent semesters or key opportunities like international exchange programs. By improving progression at this early stage, we are increasing the overall success rate of the BSc program and creating a more positive and sustainable academic path for our students.

Perhaps the most significant long-term impact is the creation of a scalable and sustainable model for pedagogical innovation. This transformative potential was formally recognized when the project won a competitive, schoolwide 10,000 EUR (nearly 11,000 USD) Vrije Universiteit Education Grant, awarded specifically for initiatives deemed “disruptive,” “scalable,” and “sustainable.”

Our initiative embodies these principles by demonstrating an effective way to embed AI into our core educational infrastructure, empowering faculty to overcome a universal challenge in higher education: providing personalized, engaging learning at scale without adding an unsustainable workload. The project has not just improved one course; it has provided a tangible and exciting path forward for creating a more adaptive and student-centered educational experience across the institution.