"I earn 100 million won a year; how much will my overdraft limit be?"
When this question is posed to a bank chatbot versus ChatGPT, the responses differ significantly. The bank chatbot typically suggests product information or connects the user to a consultant, while ChatGPT asks additional questions about credit scores, employment duration, and existing loans, providing a range for the expected limit.
As generative AI technologies like ChatGPT and Gemini become more widespread, user expectations for financial consulting services are rising. However, evaluations indicate that the practical application of bank chatbots remains limited.
According to the financial sector on July 15, major banks are implementing artificial intelligence (AI) in customer service, but their chatbots primarily handle standardized tasks such as product information, procedural guidance, and interest rate inquiries. For personalized questions, like potential mortgage limits or overdraft amounts, chatbots often provide product descriptions or recommend in-person consultations instead of specific answers.
This limitation arises from the differing types of AI technology. Generative AI operates on large language models (LLMs) trained on vast amounts of publicly available data. It synthesizes information from news reports, government documents, and public financial data to understand user intent and formulate responses through follow-up questions.
In contrast, banks typically utilize machine learning technology, which relies on pre-registered scenarios based on internal operational guidelines and financial product information. Additionally, regulations such as data separation rules, consumer protection laws, and AI fundamental laws restrict the use of internal data for training external AI models.
However, placing blind trust in ChatGPT's information can be dangerous. Information like actual loan limits or interest rates is determined by internal bank assessments and can only be approximated, with the risk of presenting incorrect information due to hallucinations. Unlike banks, generative AI does not bear the responsibility for regulatory obligations or suitability principles.
The financial sector anticipates that if data separation regulations are relaxed and AI technology advances, the application of generative AI in banks will gradually expand. There is hope that a financial AI with natural conversational abilities comparable to ChatGPT can be developed while safely utilizing internal data, significantly alleviating current limitations.
A representative from a major bank stated, "The competitiveness of bank AI lies not in the variety of information provided but in the accuracy and reliability of that information," adding that efforts are underway to enhance services, including the development of generative AI-based consulting services to improve chatbot limitations.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.