Automation for reinsurance

In the insurance industry, especially in the reinsurance sector, companies face significant challenges in terms of document processing. A central pain point in reinsurance is the manual processing of large volumes of scanned documents such as policies, applications, claims and correspondence, in which relevant data fields are often spread over hundreds of pages.

By implementing Konfuzio, both this pain point and other industry-specific challenges were resolved, as this successful case study describes in detail.

Initial situation

The day-to-day work at an established reinsurance company was characterized by manual processes for processing scanned insurance documents. The underwriters had to laboriously and time-consumingly process large volumes of PDF documents, which led to inefficiencies and high costs.


Underwriters spend a lot of time sifting through various pieces of information instead of concentrating on the actual task of risk assessment. This manual handling is time-consuming and inefficient, especially if there is also manual post-processing. The speed of document processing is therefore crucial to how quickly a contract is concluded with a client insurer. If the process is too inefficient, this leads to a competitive disadvantage, as insurers strive for prompt and reliable reinsurance. The central challenge therefore lies in a comprehensive increase in efficiency and streamlining of document processing. To achieve this, a fast and precise automation solution is required. This is where Konfuzio comes into play.


The primary objective was the Automation of the entire document processing. Through this initiative, the reinsurance company aimed to significantly increase efficiency, resulting in reduced document processing time and an increased number of processed documents per unit of time - the specific objective therefore involved a complete elimination of manual data extraction and maximum speed in process handling.

The customer's expectation was that the automation of manual processes and the use of intelligent technologies for data extraction and validation with Konfuzio would not only speed up operational processes, but also bring significant cost savings. By linking these technological improvements with the business processes, the reinsurance company intended to use internal resources more efficiently and strategically expand document management capacities.


The need for automation in document processing is a key issue for increasing efficiency and accuracy. By implementing the automation technologies offered by Konfuzio, the reinsurer can achieve significant improvements in processing speed and quality, resulting in substantial cost savings. The use of Konfuzio's intelligent technologies makes it possible to automatically extract and validate data from documents. This approach contributes significantly to the optimization of workflows within reinsurance and helps underwriters to operate more accurately and effectively.

Important functionalities of Konfuzio

  • Integration capability - The Konfuzio interface (API) enables seamless integration with the reinsurance company's existing workflow systems, ensuring efficient interaction and data transfer. The POST/documents API is used specifically for receiving and processing documents.
  • Automation functions - The IDP-engine and the Konfuzio AI are central to efficient data extraction and processing. In particular, the splitting AI, which plays an essential role in document splitting, takes up around 50 % of the total processing time and is therefore an extremely important component in achieving the goal.

Cloud, private cloud or on-premises use

The decision to use on-premises was made by the reinsurance company in order to retain full control over the processed data. In addition, the On-Premises solution This enables the customer to make customized adjustments and adapt the system environment precisely to internal requirements.

Hardware requirements

Specific hardware requirements had to be met in advance for the implementation of Konfuzio in the reinsurance company's on-premises environment.

  • Specific hardware - IT had to ensure that the hardware met the VM requirements of Konfuzio. Details can be found in the installation guide at Konfuzio On-Premises Installation Guide.
  • RAM and CPU - The configuration of RAM and CPU also had to meet the minimum requirements of Konfuzio. These requirements are described in detail at Confucius Hardware Requirements described.

Technical implementation

Once all the requirements had been clarified, the technical implementation of Konfuzio at the reinsurance company followed in 4 steps.

  • Step 1 - Software for workflow systems sends documents to the Konfuzio interface of the type REST API. This is used for the web-based receipt of documents.
  • Step 2 - Konfuzio's Intelligent Document Processing (IDP) system splits, categorizes and processes the documents received.
  • Step 3 - Konfuzio uses an RPA robot provided by the reinsurance company for data processing, which is an internal part of the infrastructure. The reason for this is that the reinsurer - which owns and operates the RPA robot - guarantees complete control over its functions. The RPA robot captures the extracted data from the most recently processed documents using the Konfuzio API.
  • Step 4 - The reinsurance company's RPA robot post-processes the extracted content and sends it to an external system for processing insurance data.
Underwriting Automation Reinsurance Reinsurance

Background information on the RPA robot

The RPA robot is used to automate recurring tasks by imitating human interactions with various software applications. Its range of applications extends from data input and extraction to filling out forms, creating reports and data processing. It uses user interfaces, web browsers and APIs to carry out defined action sequences and aims to make business processes more efficient. The robot is designed in such a way that even users without in-depth programming knowledge can use it, although a basic technical understanding of APIs, for example, is required. It is operated via a user-friendly drag-and-drop system that does not require any programming knowledge.

In this case study, the RPA robot automates the retrieval and modification of JSON data from Konfuzio's backend APIs and then transmits it to an external system, as shown visually in the graphic. From a developer's perspective, the introduction of this RPA robot into the underwriting automation process offers a significant advantage because: the need to develop a separate 'connector' for data transfer is eliminated.


After a thorough cost-benefit analysis, it turned out that the implementation of Konfuzio ideally met the client's specific requirements and that the client now enjoys a significant advantage over competing companies in the insurance sector by achieving the automation targets set. The decisive factors include a drastic reduction in processing time and a significant minimization of manual reworking.

A key unique selling point (USP) of Konfuzio is the exceptional speed of document processing:

Konfuzio is capable of processing 100,000 pages in just 12.5 hours with above-average quality.

In our specific use case, this means that Konfuzio processes around 40,000 pages in under five hours - from the moment the data is uploaded until it is fully extracted. With 128 GB RAM and 32 CPU cores, the reinsurer's production environment is optimally set up to maximize the efficiency and security of data processing in the context of on-premises use.

It should be emphasized that the AI-supported splitting of documents is the most complex component and takes up around 50 % of the total processing time. This is particularly relevant as an average customer document comprises around 200 pages. The specific information required from these pages is essential for risk assessment and rate setting, which ultimately determines the acceptance or rejection of insurance policies. This is where the OCR technology Konfuzio's software provides crucial services by enabling the precise and efficient extraction of the required data in the first place.

Charlotte Goetz Avatar

Latest articles