The Research Promotion Team builds an in-house AI-powered administration system to drive efforts toward transforming Kyung Hee into an AI-Native University
As Kyung Hee University declares its transition into an AI-Native University, innovations in education, research, and administration are emerging across the institution. In March, the university unveiled “ChatKHU,” an AI platform that allows any member of the Kyung Hee community to access the latest generative Large Language Models (LLMs). This platform establishes a unified environment for leveraging cutting-edge LLMs from global AI leaders such as OpenAI and Google. Furthermore, the “Kyung Hee AI Committee," a strategic control tower reporting directly to the university president, officially commenced operations in April. This initiative aims to reshape AI not merely as a supplementary tool for university operations, but as the fundamental infrastructure of the institution.
Analyzing 1,800+ Emails to Deliver Results from Day One
At the administrative frontline, AI is driving tangible changes that support this overarching vision. The Research Development Team at the Office of Research has built an in-house AI-Based Research Administration System. The team created this system to address repetitive inquiries regarding research project regulations, procedures, and forms, while also allowing researchers to verify information during holidays and outside regular working hours. Gi-Jun Jin, a manager from the Research Development Team and the architect of the system, explained, “We wanted to introduce a useful service for researchers while simultaneously reducing the administrative burden caused by repetitive tasks. Prior to full-scale development, we systematically reviewed multi-year inquiry cases to secure the necessary data for the system”
The Research Development Team had already been recording and managing faculty inquiries separately. To complement this, they conducted a comprehensive review of approximately 1,800 selected research-related emails accumulated over the same period to secure raw data. They designed a multi-layer classification system with five fields—Research Project Type, Inquiry Topic, Inquiry Content, Answer, and Regulatory Basis—and refined the data into an optimized format for AI training through data layering. To ensure the reliability of the answers, they appended the relevant official regulations to each response.
The AI-based research administration system was officially made available to the university community through ChatKHU, marking the first time an administrative service achieved official certification within the platform. The system drew an immediate response following its launch. During its first month of operation, it recorded 847 consultation sessions, averaging 24 sessions per day, with users actively utilizing it as a practical consultation tool by asking consecutive follow-up questions within each session. Furthermore, despite being trained on Korean-language data, the system successfully processes and responds to inquiries made in English, thereby significantly improving accessibility to international faculty members.
Faculty members expressed high satisfaction with the system. A professor from the College of Sciences shared, “Previously, verifying the criteria or obligations for using institutional research funds required a significant amount of time spent searching through past emails and attachments. Now, this system provides immediate answers, heavily reducing my administrative burden. If the university expands this service across all administrative sectors, the community will experience even greater practical utility.”
Streamlining Repetitive Tasks to Lower Administrative Burden
Based on their experience of developing the system in-house, the Research Development Team shared the critical importance of daily data management. Jin noted, “Because we consistently recorded multi-year on-site inquiry cases, we were able to build this system within a short period. If we systematically structure repetitive, routine matters, we can leverage AI to significantly alleviate the burden of repetitive administrative work.”
The AI-based research administration system is not a generic, off-the-shelf solution; rather, it is a customized system designed specifically around actual on-site research administration cases and internal institutional regulations. Moving forward, the platform features a self-evolving structure where response quality will naturally improve as more operational data accumulates, promising to become a major asset to research administration. Currently the system primarily handles inquiries regarding institutional research funds, and the team plans to update it systematically whenever internal regulations change or new data becomes available.
Administrative innovation driven by AI extends beyond the Office of Research. The Office of Educational Innovation and Planning has launched the development of an AI Agent designed to review documents and budgets for the University Innovation Support Project. The office aims to implement this tool starting with the evaluation of final performance reports in the second half of the year. Their goal is to standardize evaluation criteria based on the national guidelines of the University Innovation Support Project while simultaneously innovating administrative operations.