Smart Underwriting System: An Intelligent Decision Support System For Insurance Approval & Risk Assessment

Over the past years, insurance industry poses many challenges, one of them is maintaining the data either in legacy systems or in paper files for underwriting transaction. Most of the insurance companies are automating their data collection process. Traditionally, information of the client (such as personal details, medical records etc.) who needs insurance is sent to the underwriter through an email and after proper analysis, underwriter sends the quickQuote back to the agent based on his intuition and experience. Generally, quickQuote consists of insurance approval conditions and insurance plan name. Due to enormous amount of diseases and medicines, complexity in underwriting process has been increased. In a nutshell, an improved and optimized way of underwriting process is required. Introducing Artificial Intelligence can help to transform the traditional underwriting process to smart one. Usually data given to the underwriter is in unstructured format. Using Natural Language Processing and by training numerous statistical machine learning classifiers over the unstructured texts, important features were extracted out from unstructured emails. Main challenge is to exploit the information embedded in emails using automated tools, because of noisiness, uncleaned and unstructured data. Based on the features extracted, a model was trained and tested for unseen mails to get the proper insurance plan name and advice. This data was drafted to a template and sent back to the agent through an automated email reply. Main Objective of the project is to handle dynamic situations efficiently and to automate the underwriting task.