Solution Architecture

In this chapter, we define the solution architecture for embedding artificial intelligence into ERP software. The previously discussed ERP product qualities, ERP reference processes, ERP reference architecture, and ERP application patterns for artificial intelligence build the foundation for the solution architecture that we propose in this chapter. There are simple use cases like trending and predictions, which can be served with classic algorithms as regression and clustering. Usually, those algorithms also not require a lot of data, memory, or computing power. Therefore, we propose to implement this type of use cases within the ERP platform and call them embedded artificial intelligence. However, there are more complex user cases like image recognition or processing of natural language that need neuronal networks and deep learning. These algorithms demand high volume of data and system resources. To avoid that the critical business processes in ERP systems are negatively affected from excessive hardware consumption, we suggest scaling out this category of use cases to corresponding AI technology platform and called side-by-side artificial intelligence. While we focus in this chapter on the overall architecture, we will discuss additional aspects (e.g., data protection, extensibility) in the next chapters.