Duzon Bizon to Supply 'ONE AI PE' for Seoul's 'AI Agent' Project
IT DAILY ·
✦ Resumen de IA
Duzon Bizon said it will supply its generative AI platform, "ONE AI Private Edition (PE)," for Seoul City's "Administrative Support AI Agent Development" project.
The project is about enabling AI agents to handle tasks such as document summarization, meeting room reservations, and the selection of proposal evaluation committee members.
Seoul City is pursuing the project based on Chatbot 2.0, which it built last year, and ONE AI PE connects administrative systems and AI through an MCP-based standard integration framework.
Duzon Bizon announced on the 2nd that it will supply its generative AI platform, "ONE AI Private Edition (PE)," for Seoul City's "Administrative Support AI Agent Development" project. Seoul City is pursuing this project to support the automation of administrative work, and Duzon Bizon will provide the AI platform to the contractor carrying out the project based on its own generative AI platform and serve as the core technical foundation for implementing the agent.
The tasks to be supported in this project include document summarization, meeting room reservations, and the selection of proposal evaluation committee members. Seoul City plans to support AI agents in carrying out these tasks.
The project scope includes building a common operating environment for the administrative support AI agent, pilot deployment of AI agents for each task, expanding AI infrastructure, and identifying demand for broader adoption. The key to the project is for AI to access administrative systems based on employees' natural-language instructions, process tasks, and deliver the results.
The goal of the project is to establish an execution-oriented agent framework. Duzon Bizon participated last year in Seoul City's generative AI administrative platform project, "Chatbot 2.0," and will be involved in Seoul City-related generative AI projects for a second consecutive year.
Seoul City will pursue this project based on Chatbot 2.0, which it built last year. Chatbot 2.0 has been operating in Seoul City's administrative network internal self-hosted large language model (LLM) environment based on ONE AI PE. It has provided answers with evidence and sources by training on 270 types of administrative work manuals and applying retrieval-augmented generation (RAG), and has supported administrative work through My Chatbot and AI templates.
This project expands such use cases by first building 3 types of agents and establishing common functions such as organization chart lookup and budget information lookup that can be used by multiple departments. The 3 agents to be built first are document summarization, meeting room reservation, and proposal evaluation committee member selection.
The document summarization agent will handle summarization of documents received through the work management system, documents produced in the work management system, and attached files, using customized methods tailored to each type. The meeting room reservation agent will handle reservations, changes, and cancellations through conversational instructions alone, and will also be able to suggest alternatives that meet the conditions. The proposal evaluation committee member selection agent will send sequential text messages to candidates to confirm attendance and, based on the results, secure the number of people needed to meet the composition rules.
ONE AI PE was presented as playing a role in providing the common foundation for agent operations. Administrative systems and AI are connected through a standard integration framework based on the Model Context Protocol (MCP), and this foundation also supports user permission checks and execution history management.
In addition, ONE AI PE was designed as a structure that is not tied to any specific AI model. Accordingly, even if the model is replaced, accumulated assets such as administrative system integration and task-specific agents can be preserved.
Sineaps Soft's document structure analysis solution, DocuAnalyzer, has also been installed. DocuAnalyzer is specialized for Hangeul (HWP) documents and images, helping improve the accuracy of administrative document recognition.
Along with operating this pilot service, Seoul City plans to investigate repetitive tasks in each department, identify additional automation targets, and establish an expansion plan. A Duzon Bizon official said the project has enabled the city to implement an agentic AI administrative model in which AI actually carries out work, and said the company will support Seoul City and the contractor in creating successful public AI agent cases based on its proven platform and technical support.
Source: IT DAILY · Yang Seung-gap
Original: https://www.itdaily.kr/news/articleView.html?idxno=241995
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
Ver originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.