Digital Case Assurance – Digitalised insurance cases

What was the goal of the project?

The goal of the project was to advance the digitalisation of insurance cases. Insurance companies manage complex files in various formats, including handwritten and scanned documents. The project aimed to enable true digitalisation that goes beyond merely scanning files. Approximately 750,000 individual documents (e.g. diagnoses, medical reports, expert opinions) were analysed and automatically processed using heuristics as well as AI and machine learning models. The objective was to achieve valid results for case processing and to reduce the significant backlog in the handling of insurance cases.

Project duration: 01.01.2024 to 31.12.2025

What were the three biggest risks?

• A central risk of this project was that incorrectly recognised data could lead to the creation of inaccurate expert opinions. The project implementation was based on existing OCR software, which was to be further developed using a heuristic model. There was a technical risk that the quality of the underlying open-source OCR software would not be sufficient.

• Building the technical infrastructure, which had to handle hundreds of transformation paths, and the use of AI and machine learning models were complex and novel, bringing methodological and technical uncertainties.

• There was a risk that the large volume of documents (750,000) would not provide the desired quality to develop reliable heuristics and supply AI models with correct, high-quality data.

What was the result of the project?

The project team developed a heuristic model that was iteratively refined to achieve valid results. The digitalised and structured data serve as the basis for a RAG system that can answer queries from experts within seconds. The results are provided in the insurance-specific data format (CAMT.086 – XML format). This enabled true digitalisation, allowing for the efficient reduction of large backlogs in case processing.