In Google President’s words on SAP Relationship, Ron said, "We (Google) don't want to compete with SAP on industry technologies in that way. SAP are experts at the business process layer. We think we can add value with AI, ML, and data analytics."
SAP CEO Christian Klein recently quite assertively stated that “while it’s all nice and wonderful that the big hyper-scalers like Microsoft, Google Cloud are forging tighter partnerships with SAP, those hyper-scalers have to remember that the partner controlling the strategic customer relationship (customer data) is SAP”.
So, what is in it for SAP customers in these two statements?
SAP has clear road map for an Intelligent Enterprise. SAP with its own S/4 HANA based Predictive Analytics Library (S/4 HANA PAL), provides ability for customers to build their own (and sometimes provides off-the-shelf solutions) intelligent enterprise framework.
At the same time, obviously, Google would like to have access to the SAP Customers’ transactional data and leverage its Artificial Intelligence/Machine Learning expertise to offer insights into the customer data.
Now, let’s have a closer look both the solutions offerings:
SAP’s view:
SAP’s Intelligent Enterprise is about enabling customers to effectively user their data assets to achieve their desired outcomes faster. At the core of the IE is Machine Learning. The world’s most relevant enterprise data is part of SAP’s own system on which the customers run their business.
Customers can manage the package delivered models by retraining and debriefing them directly within their applications. Furthermore, customer and partners can adapt predictive use cases, or create new ones based on their own business needs in SAP Predictive Analytics in S/4 HANA. Some examples are discussed below with some technical details:
Machine Learning: Process Out-sorted Billing Documents in the Utiliteis Industry
SAP’s ML utilizes previous decisions from the agent to release out-sorted billing documents or to correct these documents (such as reversal). Since the customers’ system usually contains a huge number of historical documents that have already been processed, the machine learning solution can use the corresponding master data, transactional data and additional calculations based on it to make a prediction: The release of the out-sorted document is recommended and can be processed automatically (label: Release Document) or the agent must process and check the document manually (label: No Release).
To enable such ML platform to work, SAP ships pre-packaged untrained model (model UTI_BI_OUTSORTED) and CDS Views & ABAP Report REML_BILLING_OUTSRTD_RELPRED. SAP suggests that the Machine learning-relevant information for each run is stored at DB level.
In the above scenario, here are the advantages that SAP has over a third party entering into its system to read and build machine learning:
SAP owns the data in its own proprietary database
SAP provides CDS Views (a new data modeling infrastructure where data models are defined and consumed on the database rather than on the application server)
Predictive Analytics Library provides repository of classic and universal predictive analysis algorithms in data mining categories like Clustering, Classification, Regression, Association, Time Series, Preprocessing, Statistics, Social Network Analysis, Recommender System, and Miscellaneous.
For customers using SAP’s PAL to build ML capabilities, advantages are:
IS-U Master Data, Transactional Data & Machine Learning Processing Information, stays in Industry Solution – Utilities (IS-U) system.
Machine Learning takes advantages of CDS views
ML algorithms need not have to go out of their native infrastructure which provides great advantage over time, speed and data transfer latencies.
Hyperscaler’s (Google) view:
Google to deploy its ML on SAP’s Enterprise Data. To achieve this, Google should get copy of the enterprise transactional data as it may not interpret the SAP’s CDS views and leverage the data model.
This may result in latency while google is attempting to provided suggesting for the user, in real time scenarios.
Key take away:
Our opinion is that SAP S/4 HANA based Machine Learning implementation helps utilities to avoid data redundancies, speed and flexibility.