KakaoBank Presents Paper at the Largest AI International Conference in Finance Sector... "Expecting Customer Center Productivity"

KakaoBank presented its research achievements at the largest international artificial intelligence (AI) conference in the financial sector. The presentation focused on technologies that can enhance AI-based contact centers (AICC).

On the 27th, KakaoBank introduced a generative AI model that summarizes customer center consultation records at the 'International Conference on AI in Finance (ICAIF),' the largest international AI conference in the financial field.

ICAIF is the largest AI international conference in the financial sector. It is attended by employees and researchers from major global institutions such as JP Morgan, Morgan Stanley, US Bank, and BlackRock. The conference covers various topics related to AI in finance, including language models, risk management, credit evaluation, explainable AI, investment, and fraud detection.

This is the first time a domestic bank has presented research results at this conference. Notably, out of approximately 200 papers submitted to the conference, only 53 papers, representing the top 27%, are given the opportunity for oral presentation at the New York conference. KakaoBank successfully passed this evaluation.

KakaoBank independently developed a generative AI model that summarizes customer center consultation content using a fine-tuning method. Fine-tuning is a technique that tunes a pre-trained language model by further training it with domain-specific data to achieve the best performance in the desired business area.

The greatest significance of this model is its ability to generate accurate summaries in the customer consultation domain within the financial sector. While models for summarizing knowledge or everyday conversations existed, generating proper summaries in the financial field had been challenging.

To achieve this, KakaoBank retrained a Korean language model with about 5 billion parameters using financial sector data. Generating summaries of customer consultations in finance requires knowledge of financial products, the ability to understand customer inquiries and find solutions, and customer service skills.

KakaoBank expects that applying this model will reduce the time agents spend summarizing consultation records, thereby improving work productivity. Additionally, it anticipates that the quality of consultation summaries will be standardized regardless of the agent’s level of experience.

This content was produced with the assistance of AI translation services.