Smart Meetings – AI-Powered Meeting Minutes Pilot Project Launched in the KIT Senate
Meetings and discussions are part of everyday work in many divisions of KIT—as is the subsequent preparation of meeting minutes. Documenting these meetings, especially when they involve detailed minutes, is very time-consuming. Modern AI tools open up new possibilities for supporting such processes and relieving employees of time-consuming tasks.
In the GenAI∂KIT project, we are therefore exploring various ways to apply AI in everyday university life. The focus is not only on what is technically possible, but above all on where AI offers real added value, what limitations exist, and what conditions are necessary for its responsible use.
A first use case is the “Smart Meetings” pilot project for AI-assisted minute-taking in the KIT Senate. The Senate is particularly well-suited for this: On the one hand, it produces detailed meeting minutes; on the other hand, high-quality manually created minutes are available that can serve as a basis for comparison during the evaluation.
How AI-Assisted Minute-Taking Works
An important principle of the project is data sovereignty: The technical implementation is carried out exclusively using KIT’s local systems. No data is transferred to external cloud services. Audio processing takes place within the KIT network and is used solely for transcribing the spoken word. The audio data is not stored.
The process is as follows:
- Participants are transparently informed about the AI-assisted note-taking process.
- During the session, the spoken word is automatically transcribed.
- Based on the transcript, an AI model generates a first draft of the minutes.
- The designated minute-takers review, correct, and finalize the result.
The AI thus assists in creating an initial draft. Responsibility for the content, quality, and approval of the minutes remains with the designated individuals at all times.
Gaining Experience and Exploring Possibilities
During the pilot phase, a regular manual set of minutes will continue to be produced in parallel. This allows for a direct comparison of the results and an evaluation of various aspects—such as quality, completeness, the effort required to process the data, and actual benefit.
The insights gained are intended to help realistically assess the use of AI-supported solutions at KIT and to derive requirements for potential future applications.
Currently, the project is limited to the KIT Senate. Whether and in what form the technology can also be used in other divisions in the future depends on the results of the evaluation and the experience gained from it. The pilot project thus lays an important foundation for better understanding the opportunities and limitations of AI-supported assistance in everyday work.