Ranked First in Civil Servant Category at Busan City Data and AI Utilization Competition

Reduced Material Search and Cross-Checking from 6-7 Steps to 3-4

"Plans to Expand AI Use to Budget and Settlement Reviews, Ordinance Reviews, and More"

The Busan City Council, led by Council Speaker Kang Mugil, has seen success with a system that brings scattered legislative materials together in a concise format.


The Busan City Council announced on October 19 that it had taken first place in the civil servant category of the Busan City Data and AI Utilization Competition with a system that uses artificial intelligence (AI) to integrate and make use of legislative materials scattered across meeting minutes, budget and settlement documents, ordinances, and other sources.


At the final presentation review of the "3rd Busan Metropolitan City Data and AI Utilization Competition," held on October 16, the project "Building a RAG-Based Legislative and Policy Support System," presented by Kim Munjeong, a policy support officer with the Policy Support Division (PhD in Management and Policy), won the top prize.


The system combines council meeting minutes, Busan City work reports, budget and settlement documents, laws and ordinances, press releases, research reports, and other materials into a single knowledge base to find the information users need. It uses RAG (retrieval-augmented generation), a technology that enables AI to generate answers based on retrieved materials.

At the 3rd Busan Metropolitan City Data and AI Utilization Contest, Kim Moonjung, a policy support officer in the Busan City Council’s Policy Support Office (right), posed for the camera with City Council Speaker Kang Mugil after winning the top prize.

At the 3rd Busan Metropolitan City Data and AI Utilization Contest, Kim Moonjung, a policy support officer in the Busan City Council’s Policy Support Office (right), posed for the camera with City Council Speaker Kang Mugil after winning the top prize.

View original image

The key feature is that users can verify the sources behind the answers. To reduce so-called "hallucinations," in which AI presents incorrect information as fact, the system was designed to let users trace the original documents and sources used to generate each answer.


It also automatically links official project names and abbreviations with similar expressions, helping ensure that users do not miss materials related to the same project. By linking to the "Busan Policy Wiki," it allows policy-related knowledge to accumulate and carry over even when the staff responsible for it change.


The system has reduced the steps required to collect and cross-check materials. When it was piloted during the review of a supplementary budget at the 339th extraordinary session, it cut a manual process that previously took six to seven steps to three to four steps, including AI-powered integrated searches and evidence review.


Security and usability were also taken into account. The system was designed to connect legislative materials through APIs. Sensitive internal materials are primarily processed using an in-house AI, a local large language model (LLM), while connections to external AI systems are also possible when needed.



"This is an example of combining the policy data accumulated by the council with AI to improve the quality of legislative activities," said Busan City Council Speaker Kang Mugil. "We will expand the use of AI across all aspects of legislative work, including budget and settlement reviews, the enactment and revision of ordinances, and administrative audits."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing