Enhance said on Wednesday it has begun a pilot project with IBK Industrial Bank to test ontology-based AI data analysis.

The core of the pilot is to structure the bank’s internal work knowledge and tacit know-how into a form that AI agents can use, and to verify whether frontline employees can retrieve and analyse needed data using only natural-language queries.

Enhance will build its in-house AI agent platform, AgentOS, on IBK Industrial Bank’s internal network. It will set up a semantic layer that links business meaning and relationships to the bank’s information-system data, and use an ontology to organise business concepts and relationships into a knowledge dictionary.

It will connect this to the database structure so AI can interpret questions by reflecting the bank’s business terms and decision criteria. It will also use text-to-SQL technology that converts natural language into SQL queries, and retrieval-augmented generation, or RAG.

The scope of verification includes not only data retrieval but also complex analysis such as cause analysis and drawing business implications. It will also check whether the system can present the reference data used for answers and the reasoning path together.

The pilot will run through November, and the two sides plan to quantitatively evaluate performance using a “golden set”, a benchmark dataset agreed in advance.

An IBK Industrial Bank official said, "We hope this pilot will become a starting point for realising AI-ready data by turning internal work knowledge and tacit know-how into assets and establishing a system for AI agents to use them."

Lee Seung-hyun (이승현), chief executive of Enhance, said, "In finance, it is important to build a meaning layer for data so AI can use domain knowledge such as business terms and decision criteria." He added, "Through this pilot, we will prove that it is possible to present analysis results and the grounds together even on a bank internal network, and expand frontline employees’ use of data."