Digitally supporting master data maintenance in agricultural trade

Training sessions for artificial intelligence at AGRAVIS

14th September 2023

AGRAVIS Raiffeisen AG is a modern agricultural trading company in the segments of agricultural products, animal nutrition, plant cultivation and agricultural technology. It also operates in the segments energy and Raiffeisen stores, including building materials outlets, as well as in project construction. As one of five main agricultural cooperatives in Germany, AGRAVIS, headquartered in Münster and employing around 6,600 people, generates a turnover of approx. 9.4 billion € per year.

The DOCK program initiated by AGRAVIS in 2019 focuses on the group-wide introduction of SAP S/4HANA to support the company's core merchandise management processes – with consenso as the implementation partner.

An essential part of the program focuses on master data. First of all, step 1 covered the implementation of SAP MDG for business partners (organisations or persons with business partner roles such as customer, vendor, supplier, sold-to party or ship-to party). Step 2 is the much more complex project: The implementation of SAP MDG-M (material) to prepare and secure the ERP-relevant article master data.

The complexity is easy to explain, since the sheer unbelievable quantity of about 125 thousand articles in the Raiffeisen store alone and more than 14.6 million articles in the AGRAVIS Technik environment (plus machine data) have to be migrated from the legacy systems into SAP MDG-M to enable future article maintenance for AGRAVIS as a whole.

To ensure this article maintenance with as little effort as possible, AGRAVIS employees' manual work in master data management will be supported by artificial intelligence in the future.

First use cases for the application of AI were quickly found:

  • Material group assignment
  • Identification of deposit articles to ensure that these are created in SAP with the correct material type.
  • Identification of articles requiring serial numbers, such as power saws, so that the appropriate indicator is set on these articles.
  • Proposal of the base unit of measure based on various article characteristics
  • Weight prediction based on articles with comparable characteristics

With regard to the technical implementation, we rely on an architecture fully integrated in SAP. The data basis for all algorithmic evaluations is the SAP HANA database. Classification and regression algorithms that have been tailored to the specific requirements of the respective use case are executed directly on HANA using SAP Predictive Analysis Library (PAL). Even with the huge number of articles, HANA as an in-memory database ensures high-performance calculation. Intelligent Lifecycle Scenario Management (ISLM) is used for the creation, training and scheduling of model versions. ISLM is a framework in the browser-based Fiori environment managing the algorithms implemented with PAL centrally.

Important to know: Similar to humans, the AI first needs to learn which patterns are present in sample data in order to make a statement about a new data set to be classified on this basis.

For this purpose, a working group consisting of AGRAVIS and consenso employees repeatedly feeds training data for each use case into the system. The AI runs regularly and reclassifies articles. In a Fiori app, the results – enriched with various visual evaluations – can be viewed with regard to data distribution. Plausible results can then be adopted as validated master data. Validated articles are automatically added to the training data.

Each training data feed improves the AI. And the people involved are also learning, about quality requirements for article master data in the new SAP world as well as about the limits and possibilities of working with artificial intelligence and machine learning.

This shows that although AI may not yet be able to work completely autonomously and without follow-up control, it can very well provide meaningful and useful support in elementary areas of day-to-day business with a manageable investment of time and effort.

Special thanks to AGRAVIS's master data team for the great cooperation in this groundbreaking project! We are already looking forward to the next joint steps on the way to the digitisation of AGRAVIS.

From our competence centre Digital Transformation
Categories: SAP MDG - Material Master | Robotic Process Automation (RPA) | Artificial Intelligence (AI)