KIWi-Pro stands for AI-based knowledge and process management and, as a research project, combines the recording of digital as well as manual process steps for holistic process documentation.
In the future, computer vision methods and innovative camera systems will be used to automatically record and classify manual process steps in order to map a company process as a whole.
The AI-based process view is intended to provide companies with unprecedentedly detailed documentation and analysis of real processes in order to preserve knowledge and pass it on in a structured manner.
Phone: +49 681 96777 629
In a nutshell
The aim of the research project is to develop a solution that takes a holistic view of business processes and automates them as far as possible. This includes both digital and analog steps of a process that must be recorded and documented. This should enable companies and employees to analyze and document overall processes in a cost-effective manner. In this way, the successful introduction of Desktop Activity Mining (DAM) by the August-Wilhelm Scheer Institute is being used and further developed.
Desktop recording software is used for the anonymous recording of digital process data. In contrast to classic process mining, this does not rely on the provision of data from specific IT systems, but records in detail all process-relevant actions performed on a PC. The analog process steps are captured by a video-based method (video recording). In the first step, detection and tracking algorithms are used to recognize persons and objects (e.g. packages, documents) in the video image and to track their movements. In the second step, these data have to be extended by the analysis of the interactions between the persons and objects and their activities have to be classified. For this purpose, domain-specific standard activities are defined, such as “making a phone call”, or “starting a machine”.
loss of knowledge
Employees, and in particular their knowledge and know-how, are the basis for the economic success of most companies. In a study conducted for the BMWi, a majority of the companies surveyed attested to the overriding importance of knowledge management for economic success.
In practice, however, companies struggle to build up the necessary know-how or keep it within the company. For example, German IT companies lose around eleven billion euros in revenue every year due to the loss of knowledge and expertise. The transfer of knowledge from “old” employees to “new” colleagues is still largely unstructured.
11 billion euros
The August-Wilhelm Scheer Institute is developing two basic components:
Desktop recording. For the recording of digital process data, an existing work of the August-Wilhelm Scheer Institute is used and further developed by us. Here, AI-based image recognition methods are used to recognize the respective process activities anonymously with the help of screenshots. Moreover, additional information about the process step must be captured from the screenshots via OCR and anonymized if necessary.
Process Discovery. The acquired process data from digital and analog sources must be combined and semantically enriched to enable a summary of a holistic process. In addition to taking temporal relationships into account, unsupervised learning methods from the field of clustering are to be developed and used for this purpose. Based on the available additional information of the activities (e.g. screenshots of the application, OCR from the video), these methods learn which activities have similarities and thus belong to the same business transaction. For example, different activities in an invoicing process may be linked because the same invoice number was identified in screenshots and in the video. In practice, there may also be different ways in which a process is executed by employees. Process mining methods are used to automatically analyze the different variants and create a model that describes and documents the process holistically.
The KIWi project is funded by the German Federal Ministry of Education and Research.