FinanceLow importance
Shinhan Bank presents in Milan a system that checks not only the answer but the path an AI agent took in a banking task
South Korea’s Shinhan Bank presented two papers at a workshop of the ICAIF 2026 conference in Milan on Sunday (11/15). One is K-BankEval, which evaluates language models on banking tasks with 5,000 questions in six areas, such as lending, deposits, foreign exchange, document retrieval, tool use, calculations and safety from a consumer protection standpoint. In a comparison of six models, even the best on general tests performed differently on the bank’s real tasks. The other is Agent Foundry, an internal platform that analyzes agents’ answers and tool execution records to check whether they followed the rules and whether the process was efficient, before taking them into real customer service.
Published in edition no. 33 of Noticiário do Joãozinho ·
Translated by Joãozinho AI from the original Brazilian Portuguese edition. Facts, numbers and sources are the same as in the original; quotes, video titles and source names stay in their original language. Read the original in Portuguese
Original source
https://en.fnnews.com/news/202611161400551053