Korea Deep Learning has supplied its document AI solution, 'DEEP Agent' to a major domestic insurance company, automating the classification and verification processes following document recognition. In pre-verification using insurance claim documents, the rate at which the entire procedure was completed without human intervention rose from 28% to 91%. The insurance company previously processed documents by first recognizing them using OCR, and then having external reviewers compare documents with uncertain results or missing items against the original text. As the volume of claims increased, dozens of additional reviewers had to be deployed, incurring a cost of approximately 2,500 to 3,000 won per document. Deep Agent groups multiple documents into a single insurance claim for processing, verifies that all required documents have been submitted, and forwards valid cases to the subsequent review stage. Cases where documents are missing, information does not match between documents, or low-reliability results requiring verification of the original text are classified separately as exceptions. As a result of pre-verification, the number of documents requiring re-review by humans decreased by 87% compared to the previous level. Korea Deep Learning has also advanced DeepAgent's core engine by reflecting error types identified during the insurance document processing process. 'DEEP OCR' recognizes key information, such as numbers and amounts, from tilted or low-quality receipts and medical records. 'DEEP Parser' restores the structure of complex tables containing merged cells and medical expense records spanning multiple pages, reducing errors where items and