The actual business impact of such machine learning applications depends on a variety of different factors (technological, corporate, social), making its determination on an application level quite specific and individual. A more generic business impact becomes visible, if you look at a business process holistically: The orchestrated application of machine learning along a business process with all its potential synergies drives a visible business impact and results in acceptance and adoption.
One example is the process of corporate purchasing. This process is from particular interest, since it directly touches a company´s employees, probably the most critical target group when it comes to the question of company-wide acceptance and adoption. For this target group, self-service ordering must be as straightforward as possible, reflecting the user experience the employees are used to from their private consumer context.
But you´ll only have a holistic process improvement, if also the remaining process steps are improved. Because no matter where a process inefficiency occurs, it will always negatively impact the whole process. For the next process step, which is the approval of the ordered item, it is the manager, who is confronted with the daily effort of reviewing approval requests. Thus, the manager determines the speed and accuracy of this process step. Performing it with maximum efficiency is not only of the manager´s own interest but also positively impacts the whole process as well as the employee experience.
Also, the final process step, the order execution, is crucial in providing a holistic process improvement. Closely monitoring the order and being able to react in case of potential bottlenecks or inefficiencies is of essential importance. The three machine learning capabilities delivered with SAP S/4HANA 1909 (as part of the SAP Procurement Intelligence License) in the area of Procurement represent a great example for this approach to improving business processes holistically with the help of machine learning.
Proposal of catalog items based on images
With this machine learning capability, the employee may create a purchase requisition (1) based on a photo that he takes with his mobile device (2). A machine learning algorithm for image recognition matches it to images stored in a purchasing catalog. As a result, he gets a list of matching catalog items (3), from which he may select the required item to order it (4). The employee neither needs to know details about the item to order (brand, ID, etc.) nor he needs to know how to search. On top of this it is possible to place the order from wherever the requirement comes up or the photo is taken.
Intelligent approval of purchase requisition items
Approving purchase requisitions with increased accuracy and efficiency is enabled by this machine learning. Based on historical decisions, machine learning calculates the confidence level of an item approval (1). Evaluating the similarity of a set of parameters from prior decisions with the same parameters from the pending decision, it indicates the certainty, with which an approval decision is made correctly. Leveraging this decision support, the manager is able to make accelerated decisions while keeping decision accuracy and even increasing decision consistency (3). With the top influencing parameters being shown below (2), there is additionally a way to quickly comprehended where the calculated confidence level comes from (explainable AI) and to understand the “why” of the decision to be made.
Prediction of delivery date for purchase order items
Closely monitoring a purchase order is facilitated by this machine learning application. Embedded in an SAP Fiori analytical list page, which provides various options to analyze upcoming purchase orders graphically (2) and with flexible granularity (1), the machine learning leverages historical data to predict the arrival date of purchase orders (3). This means far more than enhanced flexibility for an operational purchaser. It equips him with the ability of taking proactive action, since it allows him to identify a delay, before it happens and thus opens a whole new set of options to react. Informing the employee that the order will be late is only one of these options. The one that keeps the employee experience positive. Other reactions like double-checking with the supplier or re-ordering from an alternative source are possible as well.
The example shows, that each application of machine learning surely has a positive impact on the work or process step, which is executed by a specific role:
- The employee has a more efficient, flexible and state-of-the art way of ordering
- The manager makes accurate and consistent decisions more efficiently
- The operational purchaser can monitor purchase orders proactively
But more importantly, it shows that all the steps in sum result in an optimized purchasing process offering more efficiency, higher accuracy and an enhanced employee experience. Watch the video to get a first impression.
The introduced solution is only one of the manifold innovations enabled by SAP S/4HANA. To find out more, stay tuned for our upcoming videos and blogs and feel free to have a look here:
SAP S/4HANA Feature Scope Description: http://spr.ly/6009Ds0UT
SAP Fiori Application Library: http://spr.ly/6006Ds0Uw
SAP S/4HANA Trials: http://spr.ly/6005Ds0Uc