Experiments in Teaching with GenAI: Spring 2025
Detailed examples and reflections on activities this term and year
I have been experimenting in my undergraduate and graduate marketing classes with ungraded generative AI (GenAI) exercises in class since Spring of 2023. Beginning in Summer 2024, I incorporated formal graded exercises and training about GenAI in all my classes. And I have been “traveling” with GenAI partners professionally and personally for over a year. Following is a review of my most recent class activities and their outcomes with the hope that it helps other faculty with their ideas about when and how to use GenAI in classes.
I taught two 14-week classes in Spring 2025. One was an undergraduate elective in marketing management with 37 students ranging from sophomores to seniors. Most were marketing concentrators, often with a dual degree from other disciplines such as design or psychology.
The other class was a required advanced course for marketing concentrators in our full-time MBA program. This class consisted of 20 students, all of whom had previously had me as an instructor in the general core MBA marketing class for all students.
Both courses are case and project-based. Most class sessions consist of discussions of 10–30 page business school cases that describe a business situation that the students and I then discuss how to analyze and make decisions about. Group projects include formal and informal presentation of cases along with a capstone project where students produce a full marketing plan for a new product offering at an existing company.
Both classes were explicitly positioned as GenAI “friendly,” which is still a novelty for many students. I gave responsible use guidelines for GenAI as part of the syllabus and discussed appropriate uses over the course of the term. You may see a skeleton example of my guidelines here: https://medium.com/@bruceclarkprof/generative-ai-classroom-guidelines-an-example-ccfe462ecfec.
I’ll reflect on each of these classes in turn and then make some more general observations at the end of this article.
Undergraduate Class Experience
In my undergraduate class, I repeated my structure from Fall 2024, when I assigned two formal graded GenAI exercises. These exercises accounted for a total of 10% of the final grade in the class.
The first was a “compare the summaries” exercise, where I produce two GenAI summaries based on a case we have discussed. I ask students to compare the two summaries and tell me which summary is better and why.
The second exercise asks students to have a conversation with a Large Language Model (LLM) such as GPT or Claude that extends an in-class discussion we have already had.
The order of assignments is purposeful in that I want students to work with outputs before I have them generate their own outputs — though some have experience with LLMs prior to my class. On their final projects, I give them latitude to use GenAI with disclosure.
Birch Computer Exercise
My “compare the summaries” exercise combined an individual quantitative exercise (2% of final grade) with a GenAI exercise (4% of final grade). Students were required to complete a series of calculations before class regarding a market entry case I had written. The goal was to help Birch, a fictional company, analyze entry modes for coming into the United States. In class, we discussed what we would do with those numbers and what else we might want to know to help Birch make decisions.
After class, I released two summaries, one produced using the most recent version of GPT and one produced using the most recent version of Claude. Each LLM was given the assignment and the following three prompts:
1. Perform the calculations in the attached assignment. Present in a format that matches the tables in the assignment, filling in every blank cell. Be sure to follow the number formatting guidelines on the first page
2. Based on your calculations, which of the three markets should Birch choose in entering the US. Explain the logic behind your recommendation.
3. Suggest three types of information beyond your calculations that would help Birch make decisions about entering the US market.
Students were given the following assignment:
Write a short (< 1 page) essay identifying which of the two analyses is better and why. Criteria you should consider:
· Are the Birch calculations correct? Did the GenAI program follow the number and formatting rules given in the assignment? (If you had trouble on the assignment, view the class recording where I shared screen.)
· Is the argument for pursuing a given market in 2005 plausible?
· Are the additional types of information the GenAI suggests plausible?
· Across the analysis, is the reasoning the GenAI presents compelling?
I graded these in terms of how well students argued for their choice using the four criteria.
Creativity AI Exercise
The second exercise was a “build upon class” exercise. In class, I put students into teams to apply the SCAMPER framework (https://thedecisionlab.com/reference-guide/philosophy/scamper ) to an object I gave each team, e.g., a stress ball, an apron. Each team had to come up with a product concept, a target market, and a positioning statement (a statement of competitive value the product provides) which they then presented informally to the class.
I then had each team describe its idea to an LLM and ask it to suggest creative target markets. Teams reported out on what they got back. I then asked teams to have their LLM write a positioning statement for the target market they liked best. Students reported out and we had a general discussion of the exercise.
After class, I gave individual students the following assignment:
Exercise:
1. Open an LLM and start a new chat. I do not care which one you use, but it probably makes sense to use one of the major packages such as GPT, Claude, Copilot, Gemini, or Perplexity.
2. As an individual — this is an individual assignment — take a product concept that you and your teammates worked with in class and describe it to the LLM. You may upload the file your teammates shared to Canvas if you wish, but this is not required.
3. Ask the LLM to summarize your product concept so that you can check that it understands it. If you need to clarify, do so.
4. Converse with the LLM to develop your concept further. You may find the ideas in my article useful, but I encourage you to be creative here. You could, for example, explore one or more of the 4Ps of marketing as a topic for conversation. You could ask the LLM to suggest implementation issues, pros and cons, or extensions of your concept as a topic.
5. After your initial submission, use at least three additional prompts to explore your product concept with the AI.
Deliverable 1:
Copy your entire chat, including your prompts, into a document that you submit as a word doc or pdf. Indicate which LLM you used at the top of this doc. Submit this to the assignment link.
Deliverable 2:
Write a reflection of no more than one page, single-spaced with spaces between the paragraphs. Address the following points:
· Why did you choose to explore the topic or topics around your product concept that you did?
· What was good and bad about your LLMs answers?
· How was this better, worse, or different, from working in class on product concepts?
· How, if at all, would this experience affect how you use an LLM in the future?
Think about one paragraph for each topic.
I graded this about 25% on Deliverable 1 and 75% on Deliverable 2. Deliverable 2 grading was similar to the Birch exercise. For Deliverable 1, I was not interested so much in what the LLMs produced — I skimmed that — but the quality of thinking students used. Did the prompts show command of marketing concepts, were the prompts clear, how well did one prompt follow up on the previous LLM answer, etc.
Student Feedback
At the end of the term I ran an anonymous AI feedback survey with students. Twenty-seven out of thirty-seven responded.
Allowing that there is likely some bias in the results — will you tell the professor his exercise stank? — student feedback was positive on the two exercises. In each case, 74% of respondents indicated the exercise was useful or very useful. Overall the creativity exercise was directionally higher, with more very usefuls and no responses less than neutral.
Beyond the two exercises, 24 students indicated the emphasis on GenAI in our class was “about right,” compared with one “too little” and two “too much”. This is consistent with my fall classes.
Seventy-three percent of respondents indicated that they had used GenAI in their group projects and 58% to improve their writing, both of which were allowed uses in the class. A majority of students indicated that both they and their groups benefited from using GenAI. Over 70% of respondents indicated both their attitude and likely future usage of GenAI were improved.
Over half of students also reported using GenAI to help them understand concepts or readings. Uploading copyrighted materials was not an approved usage. That said, this will be particularly tempting for students for whom English is a second language or for whom reading is difficult.
Whether this is good or bad depends on your view of education. Helping people who are smart in their native language learn in a second language is arguably a good goal; otherwise they are penalized for the language of instruction rather than their capabilities. At the same time, there is concern that students who do not put in the work will never get better at close reading.
Graduate Class Experience
As I had had all of my grad class students in a prior class, I decided to operate more experimentally with GenAI in my spring course. While I did use GenAI in a handful of assignments, these were not formally graded independent exercises. Rather, I asked students to experiment with incorporating GenAI in their workflows, as this is likely how they would be using it at work. As long as they followed the syllabus guidelines for appropriate use, they could do this as they wished. The only formal graded GenAI-related deliverable was a reflection at the end of the course. I used GenAI without a formal grade in three class sessions.
Workshop Day Exercise
About halfway through the class I held a “Workshop Day” to review team progress on their marketing plan project. I asked project teams for a milestone deliverable the night before this class. This consisted of a set of notes on several topics in their plan that I would review and give written feedback on prior to class. Then in class I would meet with each team to discuss their notes.
In addition to this, I asked them to have an LLM give them feedback on their notes as follows:
· Upload the Marketing Plan Project assignment file to an LLM of your choice.
· Ask the LLM to summarize the assignment so that you are sure it understands the project.
· Upload your notes to the same thread with your LLM.
· Explain that these notes are on only sections of the project as indicated. Ask your LLM to give you ideas and feedback on your notes for these sections.
· Copy the entire chat into a pdf document and submit this to the same Workshop Assignment link in the Communications Module.
I told students not to do anything about the LLM feedback, but that I would review it and discuss with them in class. As this was an experiment, I wanted to see the LLM feedback first. If this worked, it would be a way of students getting feedback from a second source in addition to me. In terms of process, I completed my own written feedback before looking at the LLM feedback and marking it up.
This turned out to be very interesting. I would definitely do it again.
At a basic level, LLMs were good at identifying where a team had simply left things out. One student went on to check their remaining individual assignments in the same way: have I done what I am supposed to here? This strikes me as a good use case for sometimes overwhelmed students working on multiple submissions across classes.
Feedback on the projects was generally helpful, but faltered in two areas. First, it suggested improvements that were beyond the scope of what had been due on the Workshop Day. Second, it identified ideas that were too complicated to execute in the scope of the project. Both of these could be useful in the real world, but I try to make sure students do a good job on a narrower range of subjects rather than a less good job on a broader range of subjects.
In sum, my feedback on the LLM tended to fall into three categories: “yes,” “no,” and “here’s how to build on/address this.”
Live Case Assignments
In two other class sessions, students prepared an exercise I call a “live case.” In these assignments, students are given a recent business case for background and are asked to do research to update it to the present day. Different students are assigned to different positions or roles. I suggest articles to start with, but students must go beyond these to prepare their positions.
Students were told to come to class with their own analysis which they would be discussing with teammates in class. Students were encouraged, but not required, to use an LLM with search capabilities as part of their preparation.
In class, all students were put into teams with other students who had been assigned the same role. Teams were asked to come up with recommendations for their role. In their teams, I asked students to use one LLM — with an LLM manager — to be the “sixth person” on their team for ideas and advice. We then discussed various questions and follow-ups around what the company in question should do.
Both classes unfolded well, but live it was hard for me to see how students used their LLM teammate. I learned more about that in student reflections and end-of-course feedback.
GenAI Reflection Assignment
For my one graded GenAI exercise, I asked them to reflect on their GenAI experience in the course as follows:
I would like you to reflect on your experience with Generative AI throughout the course. This should include both the role of GenAI in our in-class discussions (e.g., the Walmart-Amazon case, workshop LLM feedback) and your use of GenAI out of class (e.g., Marketing Plan project). Please address the following three questions:
- How, if at all, has your experience in this course changed your attitude towards GenAI? Explain why.
- How, if at all, has your experience in this course changed your use of GenAI? Explain why.
- Going forward, what is something you would experiment with using GenAI for? What is something you would not experiment with using GenAI for? Explain why.
Please submit no more than one single-spaced page of reflection with spaces between paragraphs.
I graded the reflections on how well they answered the three questions.
Student Feedback
My sample size on a post-course survey is too small to present numbers, but across the survey and reflections, I did receive rich qualitative feedback. Reflections were obviously not anonymous, so my previous caution on overly rosy AI answers remains. In that sense, I will give a shout out to the three brave reflections that essentially said of GenAI, “I don’t like this.” I’ll return to that point.
Broadly speaking, the qualitative comments here are similar to the quantitative results for the undergraduate class: attitudes mostly improved, and students used GenAI more over time with plans to continue doing this in the future.
In terms of specifics, the formal class exercises (workshops and live cases) all received positive comments. Students noted idea generation, feedback, and visualizations as strengths of LLMs. Using LLMs as a thinking partner came up frequently. Students appear to have become more strategic in their use over time. One student explicitly developed a portfolio approach, adopting different tools for different tasks. Multiple students also felt their prompting improved.
At the same time, students frequently stressed the importance of human review and appropriate use. Many students were wary of LLMs for personal or ethically-fraught content.
Points of disagreement across students included the accuracy of LLMs for calculations and LLM usefulness for complex tasks such as producing full strategies or campaigns. Some thought these were good uses, others were less impressed.
Representative positive comments included:
Using GenAI, I could test out different argument structures from the other side or have it help me organize thoughts when I was stuck. It became less of a “crutch” and more of a partner in my learning.
Overall, this course helped me develop a healthier, more balanced relationship with GenAI. I’ve learned to see it as a creative partner, not something to fear or avoid, but also not something to blindly trust.
As I noted earlier, there are doubters. A couple of students remarked on their fear for jobs. One student raised the issue of algorithmic bias, where more marginalized communities might be less well or poorly represented by GenAI. That student also noted GenAI’s people-pleaser personality also might lead to bias in the way it interacted with the user.
Some of the most poignant comments came from the AI-hype resisters. (I see resisters in every class.) Two highlighted some of the same concerns the academic community has expressed about creating learning without friction:
My hesitancy comes not only from fears of cheating or hallucinations but also because when learning something new I enjoy taking the long way and find it much more satisfying to think critically and do my own research.
None of us have to be here getting an MBA. So if we’re forgoing a steady salary and career experience in order to take on more reading, writing, and likely debt, why would I outsource any part of the learning process to an algorithm? . . . The moments in which I have struggled, dug deep, and kept pushing to find a solution are the ones that stick with me the most.
General Discussion
Some concluding thoughts on all this.
GenAI can do the calculations. Even if it’s not perfect, it will be plenty good enough for a decent grade on a lot of assignments. You certainly want to check how it does on any of your assignments. Do this with one of the pro packages, which many if not all students will have access to as the LLM vendors give out free trials.
Reflections are a mixed bag. I like them, but I find them hard to grade. I have leaned on Mike Kentz’s “grade the chats” framework (https://mikekentz.substack.com/p/a-new-assessment-design-framework; all errors remain mine).
What I struggle with is that any reflection is influenced by the student’s baseline, and some students have higher baselines than others. A revelation that seems mediocre to me may still be a revelation to the student. So I’m reluctant to grade on quality of insight.
In that case, I tend to come back to “how well did the student follow the instructions.” That’s OK as far as it goes, but can also lead to mechanistic grading that discounts the insights of someone who has trouble following instructions. But then I am in the territory of what I described to one student as “vibe grading.”
I end up giving high grades. For all of the reflections discussed here, about 2/3 of students received some form of A.
Practice conversations with LLMs is good for students. I firmly believe that looking at the chats tells me something about how well students are thinking about the problem. Again, some students start at a higher baseline, but asking the right question and following up with the right continuation are both important skills in dealing with LLMs (and people). And some quiet students bloom when allowed to chat with an LLM. It’s also usually the case that the conversation is much more interesting than simply asking an LLM for an answer. In this sense, prompting is critical thinking.
Human in the Loop matters. This is certainly true for students. They must check and be responsible for their outputs. But it’s also true for faculty. If I had simply told grad students to act on the LLM Workshop feedback, that would have led all teams astray. It was important that I looked at the feedback and gave them guidance.
Guidelines are good, but only get you so far. My experience is that students really like having some kind of guideline so that they know what they can and cannot do. They’re as uncertain as faculty are about a lot of this stuff. That said, you can’t come up with a rule for every situation, and students will be marvelously inventive.
And some will violate the guidelines. For example, I asked students not to upload copyrighted materials to LLMs, but my impression across classes generally is that this is definitely happening. You’re going to have to decide how much of a policeman you want to be.
With that, I will wish all professors and students very good luck in the coming year. . .
Resources
· Ethan Mollick has become an important resource for educational change for business schools in this area, and he frequently reviews new developments in GenAI. His Substack is well worth a look if you are interested in more ideas and examples: https://www.oneusefulthing.org/.
· I will also recommend Jason Gulya (Substack: https://higherai.substack.com/ ) and Mike Kentz (Substack: https://mikekentz.substack.com/ ) for thoughtful ideas on integrating AI in the classroom.
Bruce Clark is an Associate Professor of Marketing at the D’Amore-McKim School of Business at Northeastern University where he has been a teaching mentor across online and on-ground classes. He is a member of the DMSB AI Teaching Group. He specializes in managerial decision-making, especially regarding marketing and branding strategy, but at present is deeply engaged with GenAI in business and higher ed. You can find him posting and discussing regarding these topics on LinkedIn at https://www.linkedin.com/in/bruceclarkprof/.
Disclosure: I used Claude as a thinking partner for the graduate experience part of this article, helping me make sense of qualitative themes. All writing and conclusions are mine.
