Indian Institute of Management Bangalore (IIM Bangalore) has adopted a new way of examining students. Instead of taking a traditional written exam, students studying the generative AI course at the institute were tasked with building their own AI chatbot. Students had to prepare answers to the questions asked in the exam with the help of this chatbot. 78 students of Business Analytics and General MBA optional courses participated in this assessment conducted in June-July 2026 session. The aim of this initiative was to understand how well students can use AI technology in practical terms.

A chatbot was to be created instead of a written exam

In this midterm assessment from IIM Bangalore, students had to create their own AI chatbot instead of writing answers with paper and pen. For this, they needed to give appropriate instructions to the chatbot and provide the necessary content related to the course. Students can also build and test different versions of their chatbot. This gave them an opportunity to understand how changes in instructions and available information affected the quality of responses. The exam looked at students’ memory along with their practical ability to use AI.

15 questions were asked in the exam

A total of 15 questions were included in this midterm exam. Students had to prepare answers to these questions with the help of their created chatbot. For this, it was necessary to write appropriate prompts, provide required information and improve the chatbot. In traditional exams, students write answers based on their knowledge, whereas in this experiment, the ability to use AI tools effectively was given importance. This gave the students experience in the practical use of technology.

The professor prepared a digital platform

The assessment was designed by Subhabrat Majumdar, professor of decision science at IIM Bangalore. For this he has also developed a separate digital platform. The platform was powered on Vercell and was built with the help of AI-based coding tools. Each student was given access to the platform separately. Here they can design their chatbot, test its capabilities and finalize it. This arrangement also gave students who are weak in coding an opportunity to participate in this activity.

No AI, humans checked student answers

In this exam, AI was used to prepare the answers, but the responsibility of final evaluation was not given to AI. Students’ final answers were checked by a teaching assistant. Thus the human role in evaluation remained intact. This experiment helped realize that AI-based learning can assess students’ technical understanding and subject-related knowledge in a new way.