Background
Our client is a Belgian multi-channel bank-insurer, employing 41,000 and serving 12 million customers worldwide. Within the insurance sector, the European Accident Statement (EAS) is a complex handwritten document that is filled out by drivers in the event of a road accident.
Challenge
Upon receipt, claim experts manually process the EAS to establish drivers' responsibilities and to settle claim payments. Our client wanted to improve efficiency and automate the processing of the handwritten sketches on the EAS forms.
Approach
- Assess feasibility of the initiative via the Keyrus Innovation Lab to take a go/no-go decision for the project. - Obtain sufficient EAS examples and identify key objects on the sketch (like cars, letters, arrows, etc.). - Improve drawing quality to increase image classification reliability. - Build and test the advanced analytical model components and algorithms (deep learning algorithms). - Classify sketches according to the RDR convention.
Key results
Benefits
Using artificial intelligence to automate sketch classification on the EAS has significantly decreased personnel costs by automating time consuming tasks like analyzing handwriting drawings. The project also decreased claim payments by reducing potential errors and misinterpretations, and eliminated unnecessary time spent in the cumbersome manual process.
Technology partners
Python
Share this key play
