• Project structures expert knowledge using data from 832,996 pieces of equipment and approximately 2,900 unstructured documents at GS Caltex
• Ontology connects experienced engineers’ expertise with equipment data, maintenance history, and technical documentation
• “Guide” chatbot to provide potential causes, inspection sequences, and supporting evidence for equipment issues

‍

Enhans, an enterprise agentic AI company, announced that it is working with GS Caltex on an R&D project to develop an AI model that captures and structures the tacit knowledge accumulated by experienced professionals in manufacturing operations.

As experienced professionals retire and a new generation enters the workforce, manufacturers face the risk of losing years of accumulated judgment and problem-solving expertise. When equipment behaves abnormally, knowing what to inspect first or which potential cause to prioritize based on equipment conditions and operating context often depends on knowledge that cannot be fully captured in manuals alone.

Through the project, Enhans will connect GS Caltex’s structured equipment data, unstructured documents, and maintenance history with decision criteria and exception-handling expertise gathered through interviews with field engineers. Using data covering 832,996 pieces of equipment and approximately 2,900 unstructured documents, the project will structure relationships among equipment, components, abnormal conditions, potential causes, and corrective actions.

In particular, Enhans will interview engineers across equipment, mechanical, electrical, and instrumentation disciplines to understand how they assess equipment conditions, narrow down potential causes, and account for exceptional circumstances. These insights will be incorporated into the design of an ontology that structures relationships across disparate data and knowledge, allowing relevant information and supporting evidence to be connected within a shared operational context.

The resulting knowledge will be made accessible through a “Guide” chatbot designed for use in the field. When personnel ask questions in natural language about equipment abnormalities or maintenance situations, the chatbot will suggest potential causes and inspection sequences while providing access to the underlying source materials, including previous maintenance records and technical standards.

The Guide chatbot is not designed to make final decisions on behalf of field personnel. Instead, it will surface the information and evidence required for decision-making, enabling professionals to review the actual equipment conditions and relevant source materials before making the final judgment.

The project is part of the Korean Ministry of Trade, Industry and Energy’s 2026 AI Model Development Program Based on Manufacturing Tacit Knowledge. Enhans serves as the lead R&D organization overseeing the overall development, with GS Caltex participating as the industry partner. OpenGraph Labs is participating as the unstructured document parsing partner, while Chonnam National University and GIST are participating as joint research institutions.

The development environment will be deployed on GS Caltex’s on-premises infrastructure, with the system designed to ensure that the data used for the project remains within GS Caltex’s internal environment.

“Preserving expert experience should go beyond documenting what people know. The goal is to transform that experience into knowledge that an enterprise can continuously use,” said Seunghyun Lee, CEO of Enhans. “We aim to build a system that helps field professionals access the information and supporting evidence they need to make informed decisions and resolve problems more efficiently.”

‍