Embracing AI for Enhanced Banking Efficiency: A Case Study of Fukuoka Financial Group

Title: Embracing AI for Enhanced Banking Efficiency: A Case Study of Fukuoka Financial Group

In an era marked by demographic shifts such as aging populations and increasing urbanization, the quest for more productive ways of working has become imperative. Fukuoka Financial Group (FFG), operating primarily in the Kyushu region with Fukuoka, Kumamoto, and Nagasaki as its main business areas, is leading the charge toward sustainable community contributions and improved work practices through the strategic utilization of artificial intelligence (AI). This essay explores FFG's AI initiatives, with a focus on the groundbreaking experiment of "Loan Proposal Document Generation AI" conducted in collaboration with their DX co-creation partner, IBM.
FFG recognizes the potential of AI, particularly in the realm of Generative AI, and has embarked on a journey to integrate AI into various facets of their operations, spanning front-end banking services, middle-office tasks, back-office functions, and even marketing. However, realizing that tackling everything independently is a formidable task, FFG has entered into a strategic partnership with IBM, focusing on areas such as banking applications, business portals for corporate and individual clients, and IT architecture.
The focal point of their AI exploration lies in making operations more convenient and effective for their employees, starting with the practical and impactful use of AI in branch operations and headquarters. The collaboration with IBM began in earnest in April 2023, and since then, FFG has undertaken two key verification steps, aptly named Day1 and Day2.
Day1 involved the validation of the value of "Conversation Response AI" using Generative AI. Approximately 100 headquarters employees were provided access to ChatGPT within a secure internal environment to freely explore its application in banking operations. The positive results led to its widespread adoption across almost all employees.
In Day2, FFG delved into an experiment involving the integration of ChatGPT into specific internal processes, exploring its application as "Knowledge Mining AI" to enhance productivity and efficiency in middle-office tasks. One notable application was the experiment with "Regulation Compilation Response AI," where regulatory rules were trained into the generative AI to provide accurate responses to queries.
Furthermore, in Day2, a pivotal experiment unfolded in the core banking function – the creation of loan proposal documents. Collaborating closely with IBM, FFG sought to assist staff in the loan approval process by developing a "Loan Proposal Document Generation AI."
The Loan Proposal Document Generation AI aimed to streamline the creation of loan proposal documents, a critical part of the loan approval process. The typical process involves several steps, including information gathering, internal discussions, case registration, attachment creation, opinion input, and document submission. The time-consuming nature of the opinion input, where a responsible officer manually enters details about the loan request, prompted FFG to explore how AI could significantly reduce this workload.
In the experiment, a prototype was developed to allow staff to input the name of the loan applicant. The AI would then retrieve data from CRM systems, Excel sheets, and other sources to generate prompts for the staff, incorporating information such as financial statements, profit and loss statements, and collateral evaluations. The AI also utilized a technique called "grounding," where it referred to similar loan proposal documents from the past, enhancing the accuracy of its generated responses.
One innovative feature of the prototype was the ability to stream the generation process of the opinion section in real-time. This allowed staff to observe how the AI formulated responses and provided an opportunity for them to suggest modifications while considering specific numerical information for evidential support.
The prototype aimed to generate a first version in about one minute, followed by a second version incorporating staff's requested modifications in another minute. After these AI-generated versions, staff spent approximately 10 minutes manually refining the document. The entire process aimed to reduce the time required for opinion input in the loan proposal document to around 12 minutes, significantly shorter than the 30 minutes to 1 hour typically spent on such tasks.
An evaluation of the AI-generated loan proposal documents was conducted, comparing them to versions created manually by staff. The results revealed that the AI-assisted version, despite being shorter in time, maintained high quality and comprehensiveness. Notably, the AI-assisted version demonstrated a 35% reduction in time compared to the manual creation process.
Despite the ongoing nature of the experiment, the preliminary outcomes indicate significant benefits. FFG's Vice Head of DX Promotion, Taro Takeju, emphasizes the multifaceted advantages, stating that while some internal concerns surfaced about potentially diminishing the role of banking staff, the efficiency gains would translate into improved communication with customers and faster decision-making in the approval process.
Looking ahead, FFG envisions a three-step approach to AI utilization: first focusing on internal processes to enhance efficiency, then incorporating AI into customer-facing services for improved service and efficiency, and finally envisioning a scenario where AI autonomously handles certain tasks. The plan aligns with their commitment to agile progress, avoiding rigid long-term goals in the fast-evolving landscape of AI technology.
Furthermore, FFG acknowledges the importance of "unique AI" driven by their proprietary information to differentiate services. They stress the significance of effective data management, not only relying on Generative AI but also leveraging diverse AI models based on specific needs. FFG envisions deploying AI-driven sales and marketing strategies that combine these elements for a competitive edge.

In conclusion, the collaborative effort between Fukuoka Financial Group and IBM in the realm of AI is exemplified by their Loan Proposal Document Generation AI experiment. The ongoing success of this venture serves as a testament to the commitment of both entities to explore the full potential of AI for enhancing banking operations, while maintaining a keen focus on responsible and strategic AI governance. As AI becomes an integral part of the financial landscape, FFG's dedication to balancing innovation with safeguards underscores their commitment to staying at the forefront of technological evolution.