A bid notice that drew the attention of Korea's system integration (SI) industry was posted in March. It was a project to build AI agents for Woori Bank's AI transformation, or AX. The core of the project was the large-scale introduction of AI agents across the bank's overall operations. Woori Bank is also reviewing plans to purchase more than 200 graphics processing units (GPUs) to push forward with AX. Although Samsung SDS was selected as the preferred bidder for the AI agent project, the industry remains keenly focused on the project's progress. The success or failure of this project could rapidly spread the AX movement to other commercial banks, as well as the securities, insurance, and card industries.
Within the financial industry, AX is viewed as fundamentally different in nature from the metaverse and non-fungible token (NFT) booms that previously heated up the market. The view is that AI can boost work productivity and efficiency, and even transform customer touchpoints through new services. SI companies are also strategizing to seize this opportunity by leveraging AX capabilities tailored to the characteristics of the financial sector.
Banks Accelerate Employee Workflow Improvements
Within the financial industry, financial holding companies centered on banks are the most active in pursuing AX. Financial firms tend to be conservative in adopting new technologies, given that the industry provides services through money. Customer trust can be shaken if security and stability are not guaranteed. Nevertheless, banks are accelerating their AX push because there is a growing consensus that AI adoption is unavoidable for the survival of financial firms.
The area banks are focusing on first in their AX process is employee workflow innovation. Cho Sung-woo, executive vice president of LG CNS's Financial AX1 Business Division, said, "The areas where AI is driving innovation in financial institutions are largely customer experience and employee work productivity, but the recent atmosphere is to focus first on productivity innovation." He added, "Because finance is a number-based service, even very low levels of hallucination can be problematic." This is also why customers are not yet feeling significant changes despite banks actively pursuing AX.
This trend is also evident in Woori Bank's AI agent project, for which Samsung SDS was selected as the preferred bidder. Woori Bank plans to introduce AI agents in 29 major operations, including customer relationship management (CRM), corporate lending, asset management, internal controls, customer consultation, and work automation, to enhance employee work efficiency. Hwang Soo-young, vice president of Samsung SDS's financial division, explained, "Bank employees can use AI agents when creating customer asset management analysis reports or analyzing companies." He added, "Private banking (PB) customers and corporate clients will feel that services are provided more quickly and accurately through bank employees."
One of the features that financial firms have commonly demanded in recent AX projects is a "code assistant." A code assistant is a feature that uses AI in the process of interpreting existing code or writing new code. It is used to analyze C language and Java code written by developers and to generate new code. Using this can not only shorten development periods but also produce standardized code that does not depend on individual developers' tendencies or preferences.
Fraudulent transaction detection is also cited as a work area where AI effects are anticipated. Cho said, "While conventional fraud detection only filtered transactions matching pre-set conditions, using AI agents can detect transactions that differ from the norm or suspicious behavior even if not pre-defined." He added, "AI also has the characteristic of continuing to operate even when humans are resting."
Beyond banks, insurance and card companies are also pushing for AI adoption by service. Insurance companies can utilize AI for underwriting design and customer consultation automation. In the card industry, there are moves centered on Visa and Mastercard to test and verify "agent commerce" features that automate even payments.
Network Separation and Human-in-the-Loop Considerations
The financial sector's AI transformation must be carried out within the scope of compliance with network separation regulations. Domestic financial firms must operate by physically or logically separating their internal business networks from external internet networks. This is a measure to prevent hacking and information leaks and to protect financial data. However, this regulation acts as a constraint on financial firms utilizing the latest AI services or cloud-based services developed externally.
This is why SI companies build AX projects for financial firms in an on-premises manner. On-premises is a method of installing and operating AI systems and software on a company's internal servers or data centers, rather than on external clouds. Increasingly, generative AI models including ChatGPT are also being built in on-premises form to fit financial firm environments. Cho said, "Even after building in on-premises mode, AI models continue to evolve, so continuous upgrades are necessary." He added, "Considerable investment must follow accordingly." Hwang also explained, "If a separate research and development network is installed, bank employees can utilize vibe coding or ChatGPT within that network." He added, "Some internet-only banks have already installed and are operating separate research and development networks."
Recently, as Anthropic's cybersecurity-specialized AI model "Claude Mythos Preview" has been highlighted for its ability to find vulnerabilities that humans have not yet recognized, concerns are growing that AI could amplify security threats. Financial authorities are pushing to ease network separation regulations for some areas, but the general industry view is that, given the nature of the financial industry, network separation regulations are difficult to fully resolve.
For financial firms' AX to produce results, the unique characteristics and expertise of each business area must be precisely reflected in AI systems. This means that the work of converting internal regulations, product manuals, customer consultation records, research materials, and various manuals into databases and structuring them so that AI can read and understand them must come first. Cho said, "In the financial sector, explainable AI and human in the loop are important." He added, "We must be able to explain to customers why AI produced such results, and we must also leave evidence of the moments when AI made judgments."
Banking Operations Also Move to 'Zero UI'
The financial and SI industries are also paying attention to next-generation IT network construction projects alongside AX projects. Most banks have been using Unix as their main IT system for 15 to 20 years. As the operating period has lengthened, technical support has decreased, and securing developers who can handle C language, the system's foundational language, has also become increasingly difficult. This is why banks are pushing to transition to Linux-based IT networks. The industry expects an order market of about 2 trillion won to open over the next one to two years.
Demand for AI technology is also growing in this process. Hwang said, "There are many requests to include AI code assistants while building next-generation IT networks." He added, "We also receive requests not only for writing new code but also for converting C language to Java." LG CNS also introduced AI solutions in January this year while pursuing a next-generation system construction project with Nonghyup Bank. AI is utilized in tasks such as analyzing existing programs to produce deliverables.
The industry expects that as AI technology advances, the form of financial services will also rapidly change. A representative example is KakaoBank's launch of an "AI transfer" service, in which AI handles transfers on behalf of customers when they request transfers as if conversing in everyday language. Cho said, "Zero UI services such as AI transfers can also be implemented in the financial sector." He projected, "Within the next five years, AI-based financial services will be standardized at a higher level."
