A Strategy for Innovative Production at the Bilstein Group

Preparing for Artificial Intelligence

The Bilstein Group faced a series of complex challenges that made reliable production planning difficult. Its software failed to provide accurate forecasts because it could not reflect real operating conditions such as machine failures, maintenance, or short-term staff shortages. In addition to unreliable production forecasts, rising customer expectations for transparent product availability statements drove the desire for a modern, data-driven solution.

AI was intended to remedy this situation, in particular by using historical data for realistic forecasts and optimized processes. However, the company lacked both methodological and technological AI expertise internally. A preliminary project demonstrated that without external support, neither the IT department nor specialist departments could carry out a viable initiative.

In addition, ambitious requirements for the new system demanded realistic planning, accurate delivery dates, simulation scenarios, inventory calculations, and mapping the flow of materials across locations. The need was not for a selective solution, but a strategically conceived, data-supported, overall strategy.

A Success formula

SoftwareOne intentionally chose not to rush into a technological solution. Instead, it opted for a well-founded, step-by-step approach. The consulting approach was divided into three phases: Envision, Discover, and Solution – the foundation of the ‘Intelligence Fabric’ strategy.

• Envision

  In this phase, SoftwareOne and the Bilstein Group’s specialist departments developed a vision for AI-supported production planning. The collaboration aimed to identify specific use cases, define realistic goals, and align all stakeholders. This phase established a clear picture of what AI can and cannot achieve in Bilstein’s specific context.

• Discover

  An analysis of existing data sources (including SAP) was conducted to establish technical and organizational framework conditions. At the same time, the Discover phase served to build knowledge as the Bilstein team was methodically introduced to topics such as machine learning, forecasting models, and data quality. The outcome was a common understanding of the prerequisites for successful AI application.

• Solution (planned)

  The next step is a proof of value (PoV) using a specific production line. We demonstrate how completion dates can be predicted much more accurately using historical data. Real factors such as machine utilization, shift models, and downtime are all taken into consideration. The PoV forms the basis for the decision to roll out the system to other production lines.