AI-based analysis of industrial insurance contracts

MSIG uses kinisto for AI-based automated recognition of information in industrial insurance contracts

Challenge

The use of AI technology on contracts should enable the recognition of information such as individual coverage components across the entire portfolio, e.g. in order to be able to assess the profitability of individual coverages even more accurately.

Concept

Use of AI technology to extract information from complex, long contracts; training of AI models for data extraction and initial classification of information in a broader context (evaluation).

Result

Relevant sections in contracts are recognized automatically with > 95 % accuracy and the contents are correctly evaluated. This enables the automated analysis of the entire portfolio with regard to specific issues.

Dr. Alexander Kozan

Quote We now have the possibility to check our contracts for specific content in just a few hours - and this already includes the validation of specific cases by a person. The prediction quality in the pilot study clearly exceeded our expectations. The content relevant to us is recognized with an accuracy of > 95 % and already correctly classified by the deep learning technology.

Dr. Alexander Kozan
Head of Analytics & Reporting
Highlights
12 weeks from start of project to achievement of recognition rates
Analysis, classification and evaluation of a large number of customer contracts
Basis for (partially) automated data capture for incoming contracts

Information from complex contracts

MSIG Insurance Europe AG wants to set standards by digitizing knowledge work in the company. The first joint project with tetrel aimed to enable the rapid accessibility of information from contracts across the entire portfolio. The company has a large portfolio of industrial insurance contracts. The contracts describe exactly which causes of loss are covered including the exact amount covered - but the wording is not uniform. This coverage information is highly relevant, e.g., to be able to assess the profitability of individual sub-coverages even more precisely.

However, semantic text understanding is required to recognize specific information in the body text, such as the individual sub-coverages. A check by persons is not realistically feasible at all for this number of contracts.

The AI technology thus makes the analysis possible in the first place. The extraction of information from the complex, long contract texts of MSIG Insurance Europe AG can thus take place in very quickly.

The AI technology

The AI technology developed by tetrel enables the contextual reading of information from texts. All text modules contained in the document are recognized and their connotation and position are contextualized. Through training with a few sample contracts, the AI technology “learned” how to read out the information relevant to MSIG Insurance Europe AG and how to “assess” it. For this purpose, a database with typical formulations was built and Deep Learning models were trained with the tetrel development environment.

Prediction quality
> 99%
correctly identified general contract components
> 95%
correctly identified relevant contract clauses
> 95%
correct assessment of content

Future

For the future, MSIG Insurance Europe AG together with tetrel plans to transfer the results from this project to other contract elements and thus, among other things, to (partially) automate the data capture for incoming contracts. We look forward to further cooperation.

About MSIG Insurance Europe AG

  • Part of the MS&AD Group, market leader in Asia with approx. 40,000employees and 49.2 billion US dollars in sales
  • Insurance solutions and professional services for industrial customers

Talk to an expert
Contact