the increasing number of customer
communications via email, which contained supporting information for insurance claims including attachments in various file formats, such as court judgements and police reports. Its old solution used OCR to read and categorize over 2000 customer emails daily which needed to be routed to 40 different categories. Only 30% to 40% were being categorized correctly. With the growing volume of inbound emails, the insurer searched for a new solution.
inSTREAM now receives the emails and containing claims information and attachments, understanding the meaning and intent, extracting the relevant information and categorising it, finally uploading it into the insurer’s business system. When inSTREAM is unsure of the meaning of an email or attachment it refers it to a member of the customer service team and learns from the actions and decision they make for future reference.
Within seven to eight weeks of operation, 68% of emails were correctly categorized. The insurer can analyse the data related to incoming claims, which helps it to process claims quickly — a value-add for their insured customers. The system "learns" from the email exceptions, enabling an increase in the number of emails it can categorize. Process costs were reduced by 70%.