KI PoC with the ITZBund for the Federal Finance Administration

Having an overview of invoices and financial transactions is something only the most experienced experts can do. Using data analytics, artificial intelligence and employee know-how, a team from sovanta AG, together with the ITZBund and a larger team of Data Scientists, developed two proof of concepts to show how the work in the area of federal budgetary procedures can be supported. Larissa Haas, Senior Data Scientist at sovanta, summarized the success story for us:

The development environment: SAP BTP, SAP DI, SAP HANA Cloud & SAP AI Core

The use cases were implemented on the SAP Business Technology Platform within 6 months. SAP Data Intelligence was used to process data, conduct experiments, train machine learning models and make them available for testing. The SAP HANA Cloud supported the team with storage capacity for countless data sets as well as computing power to predict features. In addition, the SAP AI Core was also tested to see what the perspective use of the ML model might look like in production use. Ultimately, we used SAP BTP’s Cloud Foundry instance to host the tools that used the ML results and release them for the subject matter experts to test. 

Our goal: Create a modern working environment in which many manual processing steps are eliminated.
Larissa Haas
Senior Data Scientist, sovanta AG

Use Case 1: Assignment of e-invoices to the correct recipient

One of the use cases dealt with invoices. When these are submitted, it is often not clear to which office or to which recipient the invoices are to be sent. The employees of the so-called “clearing office” who look at these unclear cases often have to carry out time-consuming research. This is where AI can help in the future. Based on past, correct assignments, AI learns which invoices belong to which billing office. A list of the top five possibilities is displayed for employees, from which the correct billing office can then be selected. In the future, complete automation with the help of artificial intelligence is also possible. 

Use Case 2: Detection of anomalies in financial transaction data

The second use case revolved around financial transaction data. Due to the very large amount of data and different levels, it is often very difficult for an employee to gain an overview of the data. The primary goal was first to statistically analyze several years of data and present it in a clear way for the cash audit. Another goal was to highlight conspicuous transactions that require further investigation. Within different areas, there are of course many checks in the pre-systems whether postings are correct, whether they have been released, or whether they comply with legal requirements. In addition to this, the final postings should also be searched for conspicuous transactions. 

Result: Two executable PoCs

Within 5 sprints and several workshops as well as a hackathon, the work packages were delivered by the team, to the greatest satisfaction of the client. The results of the 6-month project are two executable PoCs that have shown what artificial intelligence can do, as well as an architecture plan of how the various SAP technologies that were in use would be deployed with the greatest possible benefit in the future.




All employees in
the financial sector

Go Live


Open questions?

We are happy to talk to you about projects that are not only exciting, but can also be successfully implemented.

Larissa Haas
Senior Data Scientist

Your Contact

Larissa Haas works as a Data Scientist at sovanta AG. She focuses on Natural Language Processing and helps to make software more and more intelligent and to automate tedious tasks with the help of Artificial Intelligence.
AI & Data SAP BTP Artificial Intelligence