As AI becomes an increasingly common part of student life, professor Stephen Friedman has been grappling with a challenge: how can instructors encourage deep learning when technology makes it easier than ever to offload parts of the thinking process?
Across his teaching, Friedman has observed that as AI has made it possible to complete many cognitive tasks more quickly and easily, students have naturally integrated it into their learning experiences. They use it to brainstorm ideas, summarize information and assist with assignments. Friedman says this has become known as "cognitive offloading," a term used to describe shifting mental work to external tools. In the context of AI, it refers to relying on technology to do thinking that students could be doing themselves.
Friedman is not opposed to AI-enabled learning. In fact, he encourages transparency around its use in his courses and views it as a tool students will encounter in their professional lives.
To ensure cognitive offloading doesn't come at the expense of deeper learning, he began exploring how AI could be used to create richer learning experiences that require students to think critically, demonstrate their knowledge and take greater ownership of their work.
That idea led him to rework a presentation component of his MBA Change Management course. Student teams would still be asked to present, but rather than placing the emphasis on polished delivery, Friedman shifted the focus to learners’ ability to apply their understanding in the moment.
"What can I do to create the opportunity for just enough friction to get them to up their game?" he asked himself.
To do that, presentations would be followed by an extended question-and-answer session where students are expected to explain their research, defend their recommendations and answer questions about the sources and evidence in their project.
The presentations would also be transcribed in real time and fed to a custom AI agent Friedman created to generate live follow-up questions, requiring students to think on their feet to defend their ideas.
The technology also helped address a practical challenge. During presentations, Friedman often found himself trying to listen, assess participation, take notes and formulate meaningful follow-up questions all at once. By generating potential questions from the transcript, the AI tool allowed him to focus more closely on what students were saying while still supporting a richer discussion.
The result was an assessment that placed less emphasis on delivery and more on demonstrating understanding in the moment.
To build the system, Friedman "fed" the agent course readings, assignment instructions, grading rubrics and other materials. He then spent several weeks testing and refining the tool before the course began, experimenting with different instructions and examples to improve the quality of the questions it produced.
The results were encouraging. Friedman observed stronger preparation, greater command of project topics, greater confidence during discussions and more balanced participation among group members.
Students enjoyed the process, Friedman says, and he found they were challenged to think through their ideas in real time and demonstrate what they knew and how they arrived at their conclusions.
As for the impact on Friedman's role as an instructor, the AI tool didn't replaced him; instead, it supported his teaching while leaving key decisions in human hands.
Friedman reviewed the agent's AI-generated questions, refining or rejecting them before shareing with students. The result was an synergy that combined AI's ability to rapidly analyze information with the judgment, experience and flexibility of a human instructor.
"The question for me was never whether the AI was better than me or worse than me," he says. "It was: can I use it to make me better?"
There was another benefit, too – Friedman no longer needed to divide his attention between listening, observing, taking notes and preparing follow-ups. Instead, he could focus more fully on the students and their presentations.
"I could be really present for the students," he says.
That experience challenged a common assumption Friedman sees in conversations about AI in education. While many fear technology might reduce human interaction, he found the opposite to be true. Class members became more connected to their own thoughts and more engaged with their work, while the technology gave him more time to interact meaningfully with students.
The classroom exercise was not a one-off experiment. After using the approach in his MBA course for a second consecutive summer, Friedman is now adapting it for an undergraduate business course and continuing to refine the AI agent so it can be used in different teaching contexts.
Through his work at the Teaching Innovation Studio, he is also encouraging other faculty members to experiment with practical ways AI can support learning.
For Friedman, the technology's potential goes beyond improving the classroom experience. It can also help strengthen the connection between students and instructors.
"I want to use it in a way that enhances the classroom experience," he says. "Let's use technology to be more human."