[DigitalToday reporter Jin-ho Lee (이진호)] The Ministry of Science and ICT is expected to deploy artificial intelligence to respond to parliamentary queries as early as this year. It will first build a "document centralisation system" and raise administrative efficiency with AI services tailored to ministry work such as handling external inquiries and budget reviews.

According to a report titled "A study on strengthening productivity and efficiency in administration through the use of artificial intelligence," obtained by DigitalToday on Tuesday, the ministry said document centralisation and building a platform based on integrated search must come first to expand AI use internally. The report is the outcome of a policy research project commissioned by the ministry to build its artificial intelligence transformation (AX) promotion system.

The ministry is pushing a plan to apply AI agents to internal work through its "AI-based specialised administrative services (AI-NEXT)" project. The goal is to boost productivity and efficiency in administration by creating an intelligent work environment within the ministry. The ministry is also stepping up administrative AX by using an AI agent developed by an "AI Sapiens" team involving internal civil servants for global issue analysis.

The report diagnosed limits to AI-based administrative advancement under the current structure in which documents and data are dispersed within the ministry. It presented building a centralised system to integrate and manage all administrative documents as a top priority. It said an intelligent work support system linked to large language models (LLMs) should be 마련 built on that foundation.

Building the groundwork for the use of generative AI was also presented as a task. Retrieval-augmented generation (RAG) technology, which automates search, summarisation, drafting and verification using internal documents, is expected to be key.

Specifically, the document centralisation system focuses on managing all of an organisation’s data, while the RAG system restructures data from the document centralisation system into knowledge data that AI can use to generate answers. The report also presented a plan to link generation and inference functions for real-world work use with LLMs on a governmentwide common AI platform to ensure continuous service operation.

The report further set the ministry’s AI transformation goal as building an "intelligent AI agent platform." Under the structure, AI searches for required materials by task type, produces analysis results, and then accumulates them again as knowledge assets for reuse.

The measures presented in the report were also included in a request for proposals recently issued by the ministry for building AI-NEXT.

The ministry plans to build, within this year, AI agents specialised in 5 areas: support for responding to parliamentary and external inquiries; searches for MSIT budget deliberations; support for drafting major media trend reports; determining whether equipment is subject to conformity evaluation and supporting related work; and support for inspection work for radio station permits.

The agent for parliamentary query responses is expected to have AI identify and structure query forms received as photos or PDFs and use them to help determine the responsible department. For budget deliberation searches, it focuses on improving work efficiency by configuring deliberation materials into a vector database so natural-language searches can be conducted through generative AI.

An MSIT official said, "We plan to develop AI agents that can be applied to work through a process of centralising documents and turning them into assets." The official added, "The current plan is to complete the first-phase buildout (of specialised AI agents) by year-end."

The industry, in particular, sees the document centralisation system as a factor that will influence the level of public-sector AI use going forward. Accurately collecting scattered materials and centralising them is a prerequisite for improving AI agent accuracy. If MSIT, the lead ministry for AI, moves the service into the actual build phase, the pace of spreading an AI administration model across government organisations is expected to accelerate.