@article{ART003374986},
author={Keun-Ho Lee},
title={A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System},
journal={Journal of Internet of Things and Convergence},
issn={2466-0078},
year={2026},
volume={12},
number={4},
pages={11}
TY - JOUR
AU - Keun-Ho Lee
TI - A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System
JO - Journal of Internet of Things and Convergence
PY - 2026
VL - 12
IS - 4
PB - The Korea Internet of Things Society
SP - 11
EP -
SN - 2466-0078
AB - With the rapid advancement of Generative AI, many enterprises are adopting Large Language Models(LLMs) for internal operations; however, cloud-based LLMs are difficult to deploy directly in security-sensitive manufacturing environments due to the risk of data leakage and the hallucination problem. To address the fragmentation of in-house knowledge, repetitive document-drafting tasks, and delayed onboarding of new employees caused by personalized tacit knowledge, this study designed and implemented an on-premise LLM Retrieval-Augmented Generation(RAG) system for enterprise document task automation through an industry-academia project conducted with Dujeong Tech. The proposed system improves retrieval accuracy through a RAG pipeline composed of query expansion, parent-child chunking, and two-stage reranking, and adopts an extensible database schema based on a document_type column so that heterogeneous document formats-such as quotations, transaction statements, and weekly reports-can be accommodated without structural changes to the system. In a retrieval performance evaluation conducted by four implementers, the best-performing implementer achieved a Recall@3 of 96.0% and an MRR@3 of 92.7%, and measurements confirmed that the local LLM, embedding, and reranking models can be operated together on GPUs ranging from a minimum of 4GB (GTX 1650) to 12GB or higher (RTX 5070) for multi-user environments. This study presents an empirical case demonstrating that an on-premise RAG system can automate internal knowledge retrieval and repetitive document tasks for small and medium-sized manufacturing enterprises without the risk of data exposure.
KW - Retrieval-Augmented Generation;Large Language Model;Enterprise Document Automation;Hybrid Search;On-premise
DO -
UR -
ER -
Keun-Ho Lee. (2026). A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System. Journal of Internet of Things and Convergence, 12(4), 11.
Keun-Ho Lee. 2026, "A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System", Journal of Internet of Things and Convergence, vol.12, no.4 11.
Keun-Ho Lee "A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System" Journal of Internet of Things and Convergence 12.4 11 (2026) : 11.
Keun-Ho Lee. A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System. 2026; 12(4), 11.
Keun-Ho Lee. "A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System" Journal of Internet of Things and Convergence 12, no.4 (2026) : 11.
Keun-Ho Lee. A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System. Journal of Internet of Things and Convergence, 12(4), 11.
Keun-Ho Lee. A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System. Journal of Internet of Things and Convergence. 2026; 12(4) 11.
Keun-Ho Lee. A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System. 2026; 12(4), 11.
Keun-Ho Lee. "A Study on Enterprise Document Task Automation for Small and Medium-sized Manufacturing Enterprises Using an LLM-based RAG System" Journal of Internet of Things and Convergence 12, no.4 (2026) : 11.
