Teaching with AI: Practical Strategies for Case Classrooms

Artificial Intelligence is no longer an abstract idea, it’s part of daily classroom life. Students are using it to question, create, and explore. For faculty, it’s an opportunity to rethink how learning happens and to bring more creativity and experimentation into teaching.

To explore this topic, we’ve drawn insights from a recent webinar, as well as results from a Teaching with AI faculty survey (148 respondents across Qualtrics and LinkedIn), and discussions with three Ivey faculty members: Kyle Maclean, Assistant Professor of Management Science; Mazi Raz, Assistant Professor of Strategy; and Yasser Rahrovani, Associate Professor of Information Systems.

Together, these perspectives highlight a community eager to experiment and share what works. The following strategies show how educators can integrate AI into case classrooms in ways that enrich learning, encourage critical thinking, and keep human insight at the centre of teaching.

Supporting Case Preparation

Preparation is the foundation of every good case discussion. But these days, many faculty members worry that because of AI, students could be less prepared, not more. According to our recent survey, more than half of the respondents fear that AI will reduce the depth of classroom discussion.

Rahrovani understands these concerns. “The challenge with teaching AI today is that students can easily turn to a generative AI tool, get a decent answer, and move on,” he says. To flip this dynamic, he designs his classes like an experience, plotting a learning trajectory. “It's almost like directing a short film where students start on an easy, guided path with AI, only to face a surprising plot twist that forces them to rethink what they took for granted. That twist is what makes the learning memorable.”

Maclean offers his perspective. In his classroom, he frames AI as a sort of thinking partner that helps students identify what they might have missed while they go through a series of levels of case preparation. The process begins with individual analysis, followed by AI-assisted reflection, peer discussion, and finally, live classroom dialogue. Students first analyze the case independently, then use an AI tool to test their assumptions by asking questions like, “Is there anything I have missed here?” or “What assumptions am I making that I’m not aware of?”

This simple structure turns AI into a stress test for reasoning. By the time students reach the classroom, they are more aware of their biases and ready for richer debate. As Maclean has observed, when students use AI to surface blind spots before class, they arrive more confident in their thinking and more open to challenge.

Raz adds another dimension: AI can bring in perspectives that may not naturally surface in the classroom. He encourages faculty to use AI to introduce “missing voices” or an adversarial view that challenges assumptions. These additional angles help broaden the conversation and move case discussions beyond the obvious.

Reimagining Faculty Efficiency

For faculty, the challenge is often time. Between mentoring teams, grading, and updating course materials, even the most engaged instructors can struggle to balance quality with efficiency.

In Maclean’s Sports & Entertainment Analytics elective, for example, one of the most time-consuming tasks was conducting a series of research checkpoints where he met with every team to review their progress. To streamline the process, he built a custom GPT that students use instead of meeting with him at each checkpoint. They share their research question and the data they’ve found, chat with the GPT for five to eight minutes, and then submit a shareable link as proof of completion.

This small shift created accountability and gave Maclean “a quick window into who needed extra help,” allowing him to focus his time on teams dealing with more complex or nuanced challenges. The AI handled the recurring issues he saw every year, and, as he notes, some students even had several exchanges with the GPT because it pushed them to think more deeply about their approach. The setup is intentionally simple. “It is literally just setting up text prompt. There is no programming or integrations needed,” he explains, which makes it easy for instructors to replicate or adapt for their own courses.

Most of the work to create the custom GPT went into developing the prompt itself, but once created, it becomes fast to reuse: roughly 15 minutes to personalize and run one quick test. “This would work the same in Canvas, Brightspace, Blackboard, or Moodle. If custom GPTs aren’t available, instructors could just paste the prompt into any LLM and have students do the same,” he says. For those interested in trying it, you can find the full prompt used for his first checkpoint at the end of this article.

Rahrovani also uses an AI tool in his classes. He explains: “I’ve designed two cases that run back-to-back. In the first, students use a no-code/low-code AI tool (Gemini) to build a predictive model for recidivism for judges, predicting whether a defendant might reoffend. I give them a CSV file with 50,000 cases to train an AI and walk them through every step, from data setup to evaluating performance metrics. The results look impressive: 80% accuracy! A sense of achievement fills the room: ‘Yay! We built an AI model!’”

The next class brings a twist: students analyze a real-world example of algorithmic bias, revealing how that same “accuracy” can mask systemic inequities. The shift from excitement to reflection helps students see that understanding AI’s power also means recognizing its limits.

“In short, we as instructors must become directors of learning experiences: plotting our classes like films that move students from excitement to insight, from speed to deep thinking, and from concrete stories to conceptual understanding,” Rahrovani adds.

Survey data show that many instructors use AI to support grading and maintain feedback consistency. The time saved can be reinvested where it matters most: engaging with students, responding to live discussion, and adapting to new insights.

Enhancing Relevance and Real-World Connection

Things are moving fast in the business world, and these days, even the best-written case can age quickly. AI offers new ways to bridge that gap.

Maclean developed a “news widget” that automatically summarizes current business stories and links them to course objectives. Each day, it finds a recent headline, creates a short summary, and highlights how it connects to class topics.

This tool requires almost no maintenance but adds lots of value. It allows instructors to bring current events into the classroom, helping students apply frameworks to situations happening in real time. This helps with engagement and keeps discussions fresh.

Survey results reflect this growing interest in real-world applications. When asked, “What would you most like to learn or hear about AI?” the most common response was “practical examples.” It’s clear that faculty want practical examples that bridge theory and practice. For business schools, this means exploring how AI can help translate timeless learning objectives into timely conversations.

Designing for Dialogue, Not Dependence

While AI can enhance preparation and efficiency, it doesn’t replace the heart of case learning: dialogue. Live discussions depend on spontaneity, listening, and empathy; qualities no algorithm can replicate.

Maclean encourages students to use AI before class, but not during it. He explains that when students rely on AI in the middle of a case discussion, it can interrupt the rhythm of learning because they may end up reading out ChatGPT answers that are wordy or not quite accurate.

Raz builds on this idea and emphasizes that faculty should intentionally move discussions beyond predictable “AI answers.”

In his classes, the first 15–20 minutes often surface what everyone already knows, regardless of whether these are insights generated by generative AI or familiar first-pass analysis. Once those baseline ideas are on the table, he deliberately pivots the discussion to a perspective students haven’t prepared for: the customer’s view instead of the company’s, the competitor’s angle, or a stakeholder they overlooked. This shift forces students to reason in real time, engaging with the ambiguity and complexity that AI can’t resolve for them.

Getting Started: Small Steps, Big Impact

For faculty who are new to teaching with AI, the best way to begin is simply to experiment with manageable uses: brainstorming discussion questions, refining rubrics, or generating analogies for complex concepts.

Rahrovani recommends that instructors plan their teaching with AI with theory in mind. “Beneath the story lies a theory, a conceptual backbone that students should absorb. That’s the real substance. The story may eventually fade, but the timeless conceptual insights remain.”

Raz advises educators to see AI as a mirror for pedagogy and a tool that encourages ongoing reflection. “It’s constantly keeping me awake. I’m actually a lot more excited because every time I look at what I’ve been doing, I have to think very carefully how I’m going to do this differently this time,” he says.

Raz adds that transparency also matters. He encourages faculty to be open about experimenting with AI and to let students know when they are learning alongside them. This openness helps model curiosity and reinforces that responsible learning is rooted in exploration, not perfection.

As classrooms continue to evolve, the most powerful example educators can offer is the mindset they model: thoughtful, reflective, and ready to keep learning.

Try It in Your Own Classroom: Using a Custom GPT to Support Research Projects

Ivey Assistant Professor Kyle Maclean’s custom GPT offers a simple way to guide students through early research stages while saving faculty time. Teams spend a few minutes clarifying their research question and data sources, then submit a shareable link to their conversation, giving Maclean a quick sense of which groups may need additional support.

This workflow requires no coding, works in any LMS, and costs nothing beyond a regular ChatGPT Plus subscription. Instructors can paste the prompt into a custom GPT (or any LLM), share it with students, and have them upload the conversation link as an assignment. As mentioned earlier in this article, once the prompt is written, adapting it for another course takes about fifteen minutes.

See the full prompt below.

Context: You act as a professor for a course called "Sports and Entertainment Analytics". This course is an elective at the Ivey Business School. The students do not have formal math background. The course covers topics such as linear regression, logistic regression, game theory, linear programming. Their main project is in a group, and is to select a research question/topic, and write a research paper on it. You will be helping them at their first "checkpoint" meeting. You will have a conversation with the group about their research question.

Checkpoint Context: At this checkpoint, there are three common problems. The first is that the group will not have a clear research topic. I prefer "How does weather impact Broadway Gross?" to "What leads to success in a Broadway show". The second is that their research topic is too complex and open-ended. Students will ask a vague question but not realize that they don't know how to analyze it. The third is that they have don't know where they will get their data from.

Instructions: You first have to ask students what their research question is for the group project. If the question is not clear to you, ask follow-on questions until it is clear. Then probe the students via questions on whatever you feel is the weakest aspect of the question. For instance, if you think the question is vague, ask them what results they imagine, and help guide them towards a more specific question. You should also ask them where they imagine getting data from. Push them to be specific. What websites? Students will always sound confident, so you should feel free to push back on portions that may be more difficult than they think.

End Condition: Ask no more than 4 questions total. At that point, thank them and provide any weaknesses that you see. For most teams, they will need to be more specific about where they will get data from, and more specific about their research question. Be CLEAR that the conversation is over at this point.  

Format of Responses: Ask one question at a time. This should feel like a conversation. Only ask questions on their areas of weakness. Once they have answer in a satisfactory way, move on to the next question. Use minimal formatting! You should write like a human, occasionally making typos and speaking casually.

Discover more from Maclean, Raz and Rahrovani in our full catalogue. Log in or create an educator account to preview their case studies and other learning materials.