AI assisted Claim Document Recognition System Tokio Marine & Nichido Fire Insurance Co.
A unique and secure method to handle handwritten claims notice documents using cloud based AI OCR service. It reduced 50% of document input road as well as comply with privacy regulation.
Launch date: 2017
Keywords: Claims Management, Operational Excellence, Robotics/AI
A unique and secure method to read and process handwritten claims notice documents using cloud based AI OCR service. It reduced 50% of document input road as well as comply with privacy regulation.
Tokio Marine & Nichido(TMNF) receives 300,000 handwritten claims documents for Group Sports Insurance per year. The claims office staff input those documents manually. Reading handwritten “Kanji” (Chinese characters) is not an easy task. Kanji has complicated structure with a lot of ambiguities. Office staff are often forced to make inquiries to confirm what is actually written on the documents. It’s a cumbersome and time-consuming process which sometimes result in the delay of claims payment, especially after Typhoon or long holidays when the claim notices rush into TMNF’s offices. OCR for handwritten Kanji has been long lasting dream for TMNF, but Kanji is too complicated for machines to read especially if they are handwritten in free formats. Recently, a Japanese tech startup “AI-insights” developed a revolutional technology to analyze and read Kanji using Deep-Learning AI running on Amazon AWS. However, Japanese regulation limits financial institutions to send their customer data to cloud system. So, TMNF send scanned images of claims documents to NRI. NRI shreds the images into pieces, shuffle them to hide identities, then send it to AI Insights, where shredded images are recognized by AI. The recognized texts are sent back to NRI, who re-constructs the data and send back to TMNF. All of these processes are done automatically.