Practical AI Methods in Claim Handling and Reserve Predictions

Announcement from the EAA organiser:

In this web session, we will introduce the methods and approaches from machine learning (ML) and artificial intelligence (AI) that support quantification, causes, and data outputs from for claims’ handling files and its unstructured information. We will focus on the creation of data from the information set that is contained in the unstructured documents through methods from optical character recognition (OCR), natural language processing (NLP), and classification and prediction models. The goal is to create sufficient new data sets that support the analysis, automated auditing, and machine-support claims handling. We will show the ideas, approaches but also practical examples and software tools. The goal is to understand how unstructured information of claims’ files is transformed to machine readable and processable data and what can be done with this new rich data set. 

The web session is intended for claims handler, actuaries and other insurance professionals that are interested in application of state-of-the-art AI and ML methods to claim handling and claim auditing. No extensive or theoretical knowledge of machine learning is required.

Technical Requirements: Please check with your IT department if your firewall and computer settings support web session participation (the programme Zoom will be used for this online training). Please also make sure that you are joining the web session with a stable internet connection.

In this web session, we will show the practical application of modern AI frame works from the field of natural language processing and related machine learning techniques to claim handling in motor and other P&C insurance. Today, claim handling is mostly still a manual processing where the claim handlers read and understand the documents describing a claim, make an assessment, require new information and set a reserve or conduct a transaction. At DGTAL AG, we have developed an industrial approach to digitize all documents (letters, emails, accident reports, legal documents, court rulings, etc.) to machine readable text, combine the text in a time series with the structured transaction data (reserves, payments, actions, etc.) and create a much wider data set consisting of unstructured information (documents) and structured insurance data. The unstructured information gets further enriched by natural language processing models [NLP] to create a classification of documents (letter from a lawyer, medical assessment, bill form car repair shop) and the topics (court action, a severe medical condition, dispute of liability) the document is describing. By transforming the unstructured information into analysable and modellable data, we unearth 95% of data of an insurance claim file. This models the causality of claim that is contained in the documents and sets this in relation to reserves, payouts and total damage values. We will show a software demonstration how documents are digitized, modelled and understood. Moreover, we will show how large NLP foundation model create structure data in the context and semantic of document such as the summary of a medical diagnosis and assigned ICD10 codes. Last but not least, we will discuss reserve predictions including a model that predicts severe claims from the first few hundred words of a claim file.



Click here to make a reservation: Your early-bird registration fee is € 100.00 plus 19% VAT for bookings by 13 December 2022. After this date, the fee will be € 140.00 plus 19% VAT.