The Oxbridge tutorial system, dating to the 1800s, is a teaching method built on small-group sessions where students articulate their ideas, defend them and challenge the thinking of others. It became foundational at the University of Oxford and the University of Cambridge, shaping generations of scholars through dialogue rather than lecture. Now, that model is getting a modern upgrade in American higher education, powered by AI.
At Columbia Business School, one professor saw the shift coming — and the risks with it. In the fall of 2022, classroom discussions began to feel increasingly uniform, as if students were drawing from the same well of ideas. Professor Dan Wang suspected the cause was straightforward: students were relying on chatbots such as ChatGPT, copying and pasting answers that flattened the diversity of thought.
Instead of banning the technology, Wang decided to rework it. He created an AI application called CAiSEY, designed not to hand students answers, but to challenge them. The platform, short for Classroom Artificial Intelligence Studio for Engaging You, is a conversational tutor, prompting students to think more critically and engage more deeply with course material.
CAiSEY is no pushover. The voice-powered tool, developed by Professor Wang alongside Columbia Business School alumni Jill Cohen, Class of 2020, and Johnny Lee, Class of 2023, counters fallacies, questions assumptions and offers constructive criticism. It is designed to simulate the give-and-take of a rigorous academic discussion, not simply replicate information.
The model has quickly gained traction. What began as a prototype in Wang’s Technology Strategy course has expanded into a broader platform used by about 3,000 students across 10 business schools globally, according to Columbia Business School. In each case, the system is tailored to specific courses, adapting to the subject matter and teaching goals.
Beyond delivering content, CAiSEY is intended to sharpen students’ ability to debate and communicate. It allows them to rehearse conversations, explore different viewpoints and prepare for the kinds of questions they are likely to encounter in class.
“My first hunch was that everyone would just use AI to cheat,” Wang said in a statement posted on the business school’s website. He is the Lambert Family Professor of Social Enterprise and co-director of the Tamer Institute for Social Enterprise and Climate Change.
“It turns out that I wasn’t wrong about that, but I also realized that was a naive way of looking at AI, because it’s really incumbent upon instructors and the entire educational system to think about new modes of integrating technologies in order to enhance learning, as opposed to just finding ways to make our classrooms invulnerable to cheating.”
According to Columbia Business School, the idea behind CAiSEY hinges on a simple but powerful shift: turning AI into a conversational partner rather than a shortcut. Wang found that voice-based interactions, introduced after early text-based prototypes in 2023, created a more personalized and dynamic learning experience. Each exchange is unique, unfolding in real time as students respond to prompts and refine their thinking.
The platform’s development accelerated in 2024 with the arrival of voice-based application programming interfaces (APIs), allowing the team to move beyond text and into spoken dialogue. After multiple iterations, a beta version launched in 2025, reaching nearly 300 students. On average, students spent about 22 minutes per assignment, with some engaging for as long as 45 minutes.
More than 93% of students described the experience as “great” and said they wanted more, citing improved preparation for class discussions. Professor Wang’s research also found that voice-based conversations encouraged students to explore a broader range of topics, leading to more varied and creative insights than text-based interactions.
That preparation carried into the classroom. Students arrived more ready to debate, more open to opposing views and more willing to engage in extended discussions. The result was a noticeable shift: conversations became richer, more sustained and more civil.
The approach has also resonated with students who often struggle in traditional academic settings. According to Wang, roughly 20% of U.S. adults are dyslexic or dysgraphic, conditions that can make reading and writing difficult. For those students, a voice-based system offers an alternative path to learning.
“I cannot tell you the number of students who wrote to me during the spring semester who said, ‘Finally, this is the first time in my educational career, 20 years of school, in which there has been something that has accommodated my learning style. I learn better by listening. I learn better by speaking and engaging,’” Professor Wang said.
Students who speak English as a second language have also reported benefits. Practicing conversations with an AI partner allows them to test ideas and refine language without the pressure of a live classroom, where hesitation can be more pronounced.
Professor Wang believes the implications extend beyond business schools. Interest in CAiSEY has come from a wide range of educational and professional settings, suggesting that AI-driven conversation could reshape how people learn and train across disciplines.
“One of the joys of developing CAiSEY is that we’ve gotten a lot of inbound interest from folks,” he said. “They email and say, ‘can I use this?’ And there are different settings that I wouldn’t have imagined, and I can’t tell them whether it would be effective or not. I can only tell them that you’ve got to try it out and see.”
CAiSEY is not alone in this emerging category of instructional AI. Packback, a company that began as a textbook rental startup at Illinois State University, has evolved into a platform focused on discussion and writing. After gaining national attention in 2014 with an appearance on “Shark Tank” and an investment from Mark Cuban, the company pivoted toward software designed to foster critical thinking.
Its tools, including Packback Questions and Packback Writing, use AI to guide student discussions and provide real-time feedback on writing. The goal, the company says, is not to replace student effort but to deepen it.
“Most generative AI tools replace student effort short-circuiting the learning process,” the company said in a statement on its website. “Packback is different, our instructional AI is purpose-built for education, serving as a tutor and coach that guides students through writing and strengthens critical thinking by cultivating authentic learning and originality at the source. Packback prevents academic dishonesty before it happens.”
The rise of these tools reflects a broader shift in how educators are approaching artificial intelligence — less as a threat to academic integrity and more as a potential partner in learning. Instead of focusing solely on detection and prevention, institutions are beginning to explore how AI can be integrated into the educational process itself.
Other platforms are taking a different approach, emphasizing structured study rather than live interaction. Tools like Mindgrasp allow students to upload course materials and generate comprehensive study sessions, complete with notes, flashcards, quizzes and summaries. The system can turn lengthy lectures into condensed guides, offering a more efficient way to review complex information.
Penseum operates in a similar space, converting PDFs, lecture notes, videos and presentations into organized study materials. These tools extend the reach of AI beyond conversation, helping students process and retain information in ways that go beyond simply retrieving answers.
