New directions with digital AI assistants
An SAP ERP system is already a real automation machine today. But why are so many possibilities still unused? In the area of finance, for example, almost all invoices from service providers go through at least two pairs of hands. Or logistics: A system automatically generates just 30 percent of all production orders in repetitive manufacturing from planned orders.
The reason is often that discrepancies between the system and the physical reality still exist. However, these discrepancies affect the precision of the suggestions determined by the system. For example, heterogeneous material master data leads to suggestions that dispatchers distrust. The consequence? They manually adjust the data according to their own knowledge and experience!
One solution lies in the use of AI assistants. These assistants use not only the data at hand, but also historical data. There are millions of such data in every ERP system - and these data represent patterns. The AI learns from these patterns and can reduce the existing discrepancies in this way. For example, there are AI models that determine whether a similar process was successful in the past when master or transaction data is created or changed.
Or they check which decision a person made in a comparable context in the past. Some of the findings are based on historical data, while other data is only available temporarily. Here, the data must be collected using different solution approaches.
And there are other areas of application: AI assistants can support users in their work by automatically completing critical data such as account assignment information. Or the assistants can actively warn a user of a foreseeable exceptional situation - such as the incorrect payment of a supplier invoice - even before a document is saved.
Many companies are familiar with AI from the field of predictive analytics. There are some AI solutions that are already working successfully here. But the potential applications - as shown by the examples of AI assistants - are far greater still. In order for the assistants to actively intervene in an ongoing process, however, important aspects must be taken into account: in particular, timing, data availability and reaction speed.
The latter is already enormous today: An AI assistant for document validation in financial accounting needs less than 200 milliseconds for the forecast. The question of interaction with the user must also be answered. Does the assistant warn the user with a warning tone, or does it even suggest values?
Our assistants, for example, initially use a Silent Mode, in which an assistant only acts in the background and logs suggestions and predictions. These define the Active Mode, in which the assistant interacts with the user.
Where is the journey going?
Last but not least, the high degree of individualization of ERP systems requires adjustments. For example, the AI models must learn the patterns from the company's individual data and they must be specifically adjusted. Another challenge lies in the so-called feature engineering: After all, not all information that can be extracted in tables is available at the moment of document processing, since it is only generated in the update process.
The first AI assistants are now available to tap new potential in automation and value creation in operational processes as well. And standardized S/4 Hana solutions are just accelerating the development of such assistance systems.
Companies should fully exploit the potential of the SAP automation engine with its help now in order to remain competitive and gain an edge in terms of digital transformation. The topic of artificial intelligence is gaining momentum, and companies would do well to take the first step now.
