The self-learning AI for seamless straight through processing.

Completely digitalized processes ensure a high level of cost efficiency, no question about it. However, an obstacle along the journey to complete digitalization is still media breaks in the inbox. Simple text recognition (OCR) was able  to achieve at least some success here. The problem: the poor data quality. Aisaac goes two steps further here and provides better and more and more precise records as self-learning AI.

Collect, understand, learn - Aisaac works in three steps that repeat in a loop:

1. Collect data
In the first step, Aisaac extracts the relevant text and metadata from a document. If a record is not clearly legible for the AI - for example, the scan is too blurry - correction and filter procedures will automatically be applied to improve the quality of the document.

2. Interpret and classify data
All images, text and metadata are added to the machine learning unit - the heart of Aisaac. The records are inspected according to previously defined or learned characteristics and classified correspondingly. A text with the keyword "Address change", for example, is added to the policy management system. A text that fulfills all criteria of a damage claim is added to the claims management system.

3. Learning from data
Aisaac detects patterns and similarities in documents and therefore learns independently. Furthermore, a sample is provided to the quality assurance department from adesso insurance solutions. If the data has to be corrected, the AI will learn from this as well.

Aisaac is distinguished by the following:

• Automatic extraction of structured data from unstructured texts
• Records are detected and classified, making the following possible: automatic transmission of structured data into further processing systems
• Best possible detection rates through a combination of machine learning procedures, for example, multi-layer neuronal networks, support vector machines, statistic learning procedures, learning procedures in connection with pattern matching
• Significant increase in straight throuagh processing rate, for example, when processing claims or invoices
• Division neutrality of AI permits company-wide use or usage in partial areas