As artificial intelligence (AI) search agents replace existing search box systems, it has been observed that major domestic and international institutions are also shifting their information provision methods towards generative AI.
A report by the National Information Society Agency (NIA) on Dec. 2 titled 'Changes in the search paradigm and the rise of AI search agents' contains these survey results. Recently, it has been shown that major public institutions in Korea are providing AI-based search services that go beyond simple response-type chatbots on their websites.
Since December of last year, the Legislation Office has been providing an intelligent legislation search system that performs repetitive learning on approximately 750,000 knowledge databases related to 5,218 laws and about 20,000 Q&A data related to legislation.
The Legislation Office noted that starting next year, it plans to enhance its AI legislation information service by organizing not only legal texts but also various information such as legislative background and intent, related case law, and interpretative examples to suit user convenience.
The Ministry of Food and Drug Safety has been providing a search service that learns 1,231 grievance guides and 226 guidelines for public officials based on generative AI and search-augmented generation (RAG) technology since March of this year.
The Financial Services Commission facilitated the activation of AI-based services in the financial sector by granting special exceptions for network separation regulations to major banks in November of last year.
In international cases, it is known that the Utah State Tax Commission in the United States trains its AI chatbot on its own data-based RAG system by incorporating past call content from tax call center agents, training materials, tax documents, and the latest information to provide accurate answers.
However, it is currently only used internally for responding to grievances from staff. The expansion of services targeting citizens is still under review.
The report stated that 'although information provision centered around AI is rapidly spreading both domestically and internationally, while Korea focuses on information search and response, abroad, the focus is being advanced beyond information provision towards decision-making support and actual operational execution.'
The report further pointed out that 'due to restrictions such as privacy protection and security, it is difficult for domestic public institutions to fully open data to the outside or link with other platforms. However, relying solely on internal data has limitations in encompassing unstructured data such as grievance cases and Q&A written in natural language.'
To overcome these limitations, it suggested that internal data should be converted into a vector database and, based on this, a proprietary RAG search system should be established, followed by a phased transition to an open Model Context Protocol (MCP)-based search environment structure that is currently spreading mainly overseas.