Outstanding recognition rates, lightning-fast case processing – the Cognitive Classifier at IDEAL
The IDEAL Insurance Group is revolutionizing its input management through cutting-edge AI technologies from Insiders. With the introduction of the Cognitive Classifier and a deep learning-based OCR, the insurance company achieves a new level of quality. The conclusive processing of straightforward issues is now possible directly in the inbox – a quantum leap in efficiency.
Based in Berlin, the IDEAL Insurance Group offers its customers various products for securing biometric risks and has established itself as an innovator in the insurance industry with new concepts such as private care annuity or the first completely digital insurance account „IDEAL UniversalLife.“ The company always prioritizes the interests of its customers.
Since 2004, IDEAL has successfully utilized smart INVOICE for processing incoming invoices and later introduced smart FIX for automated incoming mail processing. IDEAL processes more than 500,000 documents annually, totaling around one million pages. The document intake is highly diverse, including handwritten letters, photographed documents, faxes, emails with various attachments in different file formats, or embedded in the message, medical records, and official documents.
To minimize manual efforts, extract all relevant data accurately, and promote automated processing with a focus on a high dark processing rate, IDEAL was in search of an advanced, intelligent solution. This solution needed to handle poorly readable handwritten receipts, low image quality in photographed documents, or the combination of text and images on documents without any issues. A particular challenge was the desire to conclusively handle simple tasks – such as address changes or changes in bank details – already at the first contact with the document.
A clear case for the new Cognitive Classifier from Insiders! The Cognitive Classifier combines image-based classification with optimized text-based classification through advanced technologies such as Natural Language Processing (NLP) and Deep Learning. With the recognition of visual features, the Cognitive Classifier can now automate the recognition and processing of items such as certificate rolls with seals, notarial documents, or photo attachments. Additionally, the Universal Integration Layer for OCR (ILFO) allows the integration of various OCR components to address specific use cases. IDEAL also introduced a deep learning-based OCR along with the ILFO.
„Thanks to the Cognitive Classifier and ILFO, we have been able to crack the tough nuts in input management and make a significant leap in process automation. It’s great to be able to implement new AI technologies in close collaboration with the manufacturer,“ says Carsten Dickau, Business Analyst at the IDEAL Insurance Group.
Initially, the combination of the Cognitive Classifier and deep learning-based OCR was tested on the particularly challenging document class ‚Death Certificate.‘ This document class frequently appears in life insurance and serves as an example for all certificates whose extraction is particularly challenging with a classic OCR due to printed watermarks. In the test, a significant quality leap was achieved: 97% of death certificates could be correctly automated, and in only 3 percent of cases, a verifier had to intervene.
This result thrilled IDEAL so much that the Cognitive Classifier and optimized OCR were introduced for all document classes of the insurer. By combining image and text information, the Cognitive Classifier excels even with previously hard-to-classify document classes, increasing recognition rates by another approximately 15 percent to nearly 100 percent. The alternative OCR connected via ILFO also contributes to improved recognition in the case of qualitatively poor photos, faxes, or scans, as well as handwritten documents, significantly enhancing the higher level of automation.
With these innovative AI components, IDEAL can now process information that was previously difficult or impossible to extract. The conclusive processing of simple tasks in input management has also been implemented.
The IDEAL example demonstrates that providing a platform where various AI technologies can collaborate consistently brings process improvements for customers through technological advancements.