KB Kookmin Bank has partnered with Snowflake to embark on data-centric AX (AI transformation). The bank is shifting from AI limited to chatbot-level functions to AI capable of performing actual financial operations. To achieve this, KB is restructuring its data foundation so that AI can carry out tasks based on trustworthy data and plans to expand the number of AI agents across the entire group from approximately 100 currently to around 300 by next year.
Lee Kyung-jong, Senior Executive and Head of the Financial AI2 Center at KB Kookmin Bank and KB Financial Group, stated on the 27th at 'Snowflake World Tour Seoul' held at COEX in Gangnam-gu, Seoul: "While model performance is important for the competitiveness of AI agents, true strength comes from a data foundation that enables agents to act with confidence." He added, "The more critical the data is as the basis for action, the greater the importance of accuracy and reliability."
KB Financial Group initiated the development of an agentic AI platform using generative AI in 2024 and officially launched it last April. The group currently operates over 100 AI agents, with more than 8,000 actual users.
Following its initial application to front-office areas such as asset management and corporate finance—key customer touchpoints last year—the group is now expanding into middle- and back-office functions as well as IT development this year. It is developing risk management, credit review, and legal support agents while also building a system in which AI supports the entire IT development process from planning to development and testing.
The core challenge for enabling AI to perform actual work lies in data. Existing data is scattered across multiple platforms, requiring separate development each time new data is integrated. Additionally, unstructured documents often lack the context necessary for AI to grasp business-related meanings.
This is why KB Kookmin Bank has partnered with Snowflake on data-centric AX (AI transformation). By connecting fragmented data and utilizing 'semantic views,' the bank assigns business context and meaning to data. The goal is to enable immediate access to data needed by AI, reduce repetitive integration tasks, and accelerate both the development and deployment of agents while improving accuracy. The system also allows business units to directly manage data quality and access rights, sharing verified data through an internal data marketplace for reuse by other AI agents.
The center head noted, "Not only GPT and Gemini but various open-source models are becoming increasingly standardized." He emphasized, "Differentiated competitiveness stems not from the models themselves but from the quality and utilization of data held by each enterprise."
KB's ultimate goal is for AI to go beyond merely drafting reports—to directly input results into business systems and independently plan and complete multi-step tasks. The group also plans to expand into agentic AI services targeting customers in the future. The center head stressed, "The AI data framework being built by KB represents a long-term strategic asset that competitors cannot easily replicate."
in TECH
