Hancom will work with BGF Co. on a groupwide AI transition (AX), including converting companywide data into knowledge assets that can be used for AI search and responses.
Hancom said on the 11th that it has organized structured and unstructured data accumulated inside BGF Co. into a format that AI can read and process, centered on its AI data processing solution Hancom Data Loader and its knowledge search solution Hancompedia. The project builds an AI knowledge search system that helps intelligently handle internal data such as companywide bulletin boards, work documents, and attachments owned by BGF Co.
Instead of large-scale systems integration (SI) development, the project applied Hancom's AI solutions to fit BGF Co.'s work environment. In particular, to prevent leaks of internal materials and meet security requirements, it was built as an on-premises environment operated on the client's internal infrastructure. It also reflected BGF Co.'s security policies, including user-specific access permissions, document viewing scopes, and data processing standards, and was designed to connect with domestic and overseas large language models (LLMs) based on work characteristics and performance needs.
The two companies' collaboration began with an AI proof of concept (PoC) that started in Aug. 2024. After analyzing BGF Co.'s work environment and data structure, Hancom refined and transformed information dispersed across internal bulletin boards, documents, and attachments with Hancom Data Loader and linked it to a Hancompedia-based retrieval-augmented generation (RAG) structure. After confirming practical applicability through the proof of concept, the two companies selected in-house knowledge search as the first application area.
The first phase centers on building a Hancompedia-based knowledge search system linked to companywide bulletin board data. Hancom processed existing board databases (DB) and document materials into a form suitable for a vector DB. It established an integrated search framework by including diverse data formats—such as HWP, PDF, XLSX, plain text, bulletin board bodies, and attachments—as analysis targets.
Based on the organized data, Hancompedia provides RAG search, natural language Q&A, AI agents, and deep search functions. After identifying the user's question intent, it comprehensively explores information scattered across multiple documents and boards to deliver answers aligned with the work context.
With this system, BGF Co. plans to integrate knowledge and information dispersed by organization into a single framework and to push ahead in earnest with group-level AX. Hancom also strengthened its position as a B2B AI partner that safely converts corporations' data into AI assets.
Based on the performance of the first-phase system operation, the two companies plan to improve search quality and response completeness. While expanding application areas to various domains within the group, they also plan to broaden use cases to document drafting, summarization, and Q&A by linking with work-support AI services such as Hancom Assistant.
Hancom CEO Kim Yeon-su said, "For corporations, safely converting and using internal data as AI assets has now emerged as a key task that determines corporate survival and business competitiveness," adding, "Building on our collaboration with BGF Co., Hancom will pioneer the enterprise AX market and establish a sovereign agentic OS that secures both Data Sovereignty and AI productivity as a new standard for enterprise AX."