Bar-Ilan University Pilots AI Agent to Help Students Catch Up on Missed Lectures

A study tool grounded in course materials

In February, Bar-Ilan University (BIU) began piloting the agent on Moodle. It answers natural language queries and turns the large volume of course content into learning support. According to Microsoft, it is powered by Azure AI Services and was developed by KPMG and OpenApp using Microsoft Copilot Studio.

The agent is grounded exclusively in material uploaded by BIU's teaching staff. That includes recorded lectures, transcripts, presentations and articles. Students can ask it to focus on a topic mentioned in a lecture, create summaries and test themselves with the quiz feature. It operates in the university's secure Microsoft Azure environment and is available in Hebrew, Arabic and English.

BIU is one of Israel's leading public research universities, with over 20,000 students in Ramat Gan, near Tel Aviv. The agent will be rolled out to another 100 courses in the first semester of the 2026/27 academic year, which began in October.

A student's view: from catching up to active study

Yarden Shay, who studies criminology and psychology, missed some classes in her first semester last year. The lectures covered legal responsibility, one of 22 modules in the first year of her degree. Borrowing notes while following six to eight hours of new lectures daily was hard. "It's like a gap that is growing through time," she says.

Today she no longer spends hours tracking down classmates for notes. She also uses the agent, represented by a blue robotic owl, as a study aid. It tests her with multiple-choice and open-ended questions and tells her whether an answer is right or wrong, and why. She now studies about three hours a day after lectures, compared with five to six hours before. She says the tool has made her feel more self-reliant, and other students have asked for the summaries she made with its help.

A lecturer's view: teaching the bot what to teach

Eliyahu Biton, a practicing lawyer and BIU lecturer since 2017, helped test the agent before launch. He runs online healthcare law and ethics courses for medical imaging students and overhauled his materials for the pilot. That meant feeding the agent transcripts of all his video lectures and expanding the bullet points on his slides with research papers and recommended reading. He then tested its responses until he was satisfied with their accuracy.

"If I want it to be my assistant, I need to teach it," he says. He reports fewer routine and administrative questions and says students are less nervous about exams. Instead of only uploading old exam questions, he now asks students to have the bot generate practice exams.

At the finals in August, the average score among his roughly 50 students was three percentage points higher than in previous years. Biton concedes this could also be due to a better-than-average cohort.

What it means for teaching

Biton still holds 90-minute sessions with his students four to six times a year for deeper discussion of a course topic. He says his role is shifting from pure teaching toward mentoring. "Knowledge is no longer the problem," he says, adding that he can do better than a chatbot at guiding students on how to learn and approach real-world challenges.

Shay sees the agent as an add-on. "It's something that goes with the lecturer, not instead of the lecturer," she says. The BIU example suggests that the quality of such tools depends on lecturers preparing and checking the materials the agent draws on.