The search for a cross-system solution
The head of IT and his team were searching for a flexible software solution that could be used across the board. Their workflow system provider brought their attention to the possibility of solving the problem across all their systems using artificial intelligence. With its ability to extract and import even complex data from widely varying documents, DocuWare IDP perfectly met their requirements.
In their business, each client uses different, sometimes outdated time-tracking software, and not all software systems produce documents from which the pertinent data could be automatically extracted. This meant that invoicing clients for the billable hours required a considerable amount of time-consuming, manual data entry. The previous solution was to use rule-based programs to import the data, but these needed to be readjusted on a regular basis.
Training AI for complex documents
At the end of 2024, the first trial solution was tested with the help of the implementation partner. Some of the imported lists with time-tracking data were processed accurately right from the start; for others, the AI required more in-depth training. After the additional training, the AI was able to accurately extract data from the more challenging lists as well. Since March 2025, a continuous improvement process has been established, in which the AI receives targeted training on especially complex files.
Of the approximately 1,000 documents that are imported each month, about 200 present a particular challenge due to their complexity. These go through a workflow which integrates DocuWare IDP. When needed, a one-time manual check may be carried out for especially challenging cases—such as conversion of industrial minutes (100 seconds). After the data is successfully categorized, no manual checks are required for future processes. The system calculates a confidence value for each document. If this value is under a defined threshold, a visual check is automatically triggered in the workflow. Finally, the data is used to produce XML files which are transferred to the ERP system for billing.
Avoiding peak loads
The AI system has the potential to be a long-term solution to one of the company’s core problems—one that always pops up when payroll accounting is due and a large number of documents need to be processed in a short time frame. When this happens, the extensive need for manual data imports pushes the administrative department to its limits. But these peak loads can now be avoided. The automation makes it possible to drastically reduce the manual workload—and the goal is to eliminate it entirely.
The IT team already has identified additional use cases for AI-based document processing. These include extracting data from hand-written work documents from smaller companies without electronic time-tracking systems, as well as processing documents such as paper-based vacation requests.
