"Recommend a card with KTX discount benefits."
"I've gathered cards that offer various benefits related to telecommunication fees." (sends link for 'KT Family Satisfaction DC Shinhan Card')
This was the response from Shinhan Card's artificial intelligence chatbot when asked to recommend a card offering transportation discount benefits. When the question was rephrased to "recommend a card with the largest public transportation cashback benefits," the chatbot instead suggested a "card for customers applying for livelihood recovery consumption coupons."
Banks and card companies are expanding AI consultation services including callbots and chatbots, but actual consultation accuracy is falling short of expectations. As inefficient consultations—dubbed "AI runaround"—increase, experts point to the need for collaborative models between AI and human consultants.
According to a review of AI consultation cases at major financial companies on the 9th, simple keyword-based inquiries such as account limit checks and transaction history were processed, but complex questions were not properly addressed.
When Shinhan Card's AI consultant was asked to summarize the benefits of a credit card currently in use, it provided a link to the card company's homepage instead of an explanation. NH Bank's AI consultant similarly transferred the call to a human agent when asked about savings rate comparisons or automatic transfer setup methods.
Errors were also identified in AI chatbots embedded in mobile applications. When Woori Bank's AI loan consultant was asked about application requirements for the "Woori WON Jeonse Loan," the chatbot explained requirements for first-time homebuyers. When asked "Is the jeonse loan only available for first-time buyers?" the AI consultant answered "Yes." However, the actual application requirements for Woori WON Jeonse Loan contain no such condition.
Product recommendation functions were also limited. When KB Kookmin Bank's chatbot was asked to "identify the savings account with the highest interest rate," it provided information on various savings products without specifying the highest-rate product. Woori Card's chatbot also asked for clarification when requested to recommend cards with low annual fees. A KB Kookmin Bank representative explained, "Because preferential interest rate conditions differ for each customer, we operate by providing general savings product information rather than definitively presenting the highest-rate product."
Low accuracy in AI consultation services also affects user satisfaction. According to a Deloitte survey of 2,027 U.S. bank customers conducted in January last year, 74% of respondents preferred human consultants over AI chatbots even for simple tasks. Chatbot usage purposes were mostly simple tasks including resolving technical issues (60%), account inquiries (53%), and card payment processing (29%).
Within the financial industry, assessments indicate that current AI consultation services remain at the level of handling simple inquiries. A financial industry official said, "AI consultation often operates based on fixed keywords or scenarios, making it difficult to address complex product inquiries," adding, "The direction is for AI to handle simple tasks while human consultants focus on complex consultations."
