From Manual Compliance Efforts to an Efficient AI Backbone

The steel manufacturer GMH Group uses AI to process mill certificates and other documents. With cbs AID (Advanced Integration of Documents), data is extracted with 99.79% accuracy and transferred to SAP. This increases the level of automation and process reliability, while the replacement of outdated OCR solutions creates cost advantages. The realized efficiency and effectiveness gains support the GMH Group in its transformation toward green steel.

About the GMH Group

The GMH Group is a leading German steel producer and processor headquartered in Georgsmarienhütte, Lower Saxony. The group comprises more than 15 medium-sized production companies with approximately 6,000 employees. As a pioneer in the production of green steel, the GMH Group uses state-of-the-art electric arc furnaces in its manufacturing process, which generate 80 percent fewer CO2 emissions than conventional production methods. The process relies almost exclusively on recycled materials.

Establishment of an AI backbone

Extracting information from documents and entering it into an SAP system: This error-prone process causes delays and prevents employees from performing higher-value tasks. The GMH Group has set itself the goal of comprehensively reducing the effort involved in document processing and relieving the burden on employees. Starting with mill certificates, the steel producer has created a powerful AI backbone that is being expanded to include additional document types. With the G.A.D.G.E.T. project, GMH has turned a compliance bottleneck into a strategic advantage.

It was no coincidence that GMH started with mill certificates. In steel production, every shipment of raw materials is accompanied by such a document. These are test and acceptance certificates that are essential for SAP-based production planning. They contain quality and compliance-related information such as batch, order number, and chemical composition. As a pioneer in low-carbon steel production, the GMH Group uses nearly 100 percent recycled material. This presents challenges: The number of suppliers and their deliveries is very large, so a high volume of mill certificates must be recorded and transferred into the SAP system. Roman Michel, Senior Consultant SAP MM at GMH Systems, explains:

“Shop floor documents are extremely complex. Given their high volume and importance to our production processes, we needed a solution that could process these documents both accurately and with minimal effort.”

Previously, conventional OCR (Optical Character Recognition) solutions were used. These relied on templates for recognition and did not meet the requirements for flexibility and cost-efficiency. The structure of factory documents frequently changes, which necessitated equally frequent and costly software adjustments. Individual plants also extracted the information entirely manually, which was time-consuming and error-prone.

The GMH Group decided to systematically digitize these factory certificates using a standardized, modern AI solution. The choice fell on cbs AID – Advanced Integration of Documents. “There was a need for a modern and easy-to-maintain solution that ensures a standardized process for all suppliers with high accuracy. This is exactly where cbs AID proved to be the optimal choice,” says Roman Michel. In addition, cbs helped design the architecture and implementation.

Automation through AI

cbs AID uses Agentic AI and does not require templates or vendor-specific training. The SAP BTP-based solution first captures all available information from the document and then performs a context-based analysis to identify and extract the relevant data. Thanks to integration via SAP BTP into the backend ERP system, the data is verified, including a warning system for end users in case of discrepancies. After a final human review via a dashboard, the data is automatically transferred to the SAP system. As a custom AI, cbs AID is tailored to each customer’s specific requirements, significantly outperforming standard AI solutions in terms of accuracy and flexibility. With SAP BTP as a layer, the AI services are cleanly decoupled from the SAP core while remaining deeply integrated into the processes.

Following the proof of concept, the GMH Group initially rolled out cbs AID in its pilot plants and achieved a near-perfect extraction accuracy of 99.79 percent. “That’s an outstanding figure. The solution significantly outperforms conventional extraction methods in terms of quality and speed. We save costs, minimize errors in the process, and improve data quality and operational efficiency,” explains Roman Michel. The impact is measurable. Since cbs AID reliably recognizes changing formats and new suppliers without requiring retraining, maintenance costs are reduced by 95 percent. There is no effort involved in onboarding new suppliers or layouts. As a result, a positive return on investment of 230 percent is achieved within a three-year period. The solution is now being rolled out continuously to additional plants.

Focus on Additional Document Types

The cbs AID Framework is flexible, scalable, and easily expandable to include additional document types, from invoices to complex forms. Building on the experience gained, the GMH Group is currently rapidly expanding the use of cbs AID to delivery notes and order confirmations.

“Our goal is to be innovative not only in the field of green steel, but also in the use of AI. Accordingly, we are driving forward additional use cases,” emphasizes Roman Michel. To this end, an AI backbone has been created that strengthens the GMH Group’s green steel strategy and its position as an industry innovator through smarter, more reliable, and faster processes.