Korea Deep Learning Says IBK PoC Cut Financial Document Processing Time by 88%

Korea Deep Learning (KDL), a South Korean company founded in 2019 that serves more than 80 companies and institutions, announced on November 10 the results of a proof of concept with Industrial Bank of Korea (IBK). The company said its vision-language model agent, "DEEP Agent for Finance," automated documents from voice phishing relief cases, combining document recognition, information extraction, and crime-type inference.

Based on the project's evaluation criteria, KDL reported 100 percent accuracy on printed text and document classification, above 97 percent on handwritten text, and an 88 percent reduction in processing time. The system runs on IBK's closed network, with employee review applied to cases requiring additional verification. All figures are self-reported; independent analysts at Five Takes noted the absence of an external auditor.

KDL said a proof of concept with Shinhan Vietnam Finance has been completed, with deployments running in Vietnam and Taiwan. The company said it is in PoC discussions with nine Japanese financial institutions and plans local sales channels in 2027, followed by expansion into insurance and securities.