9.3    SAP S/4HANA and SAP Leonardo

SAP Leonardo is a digital innovation system from SAP that delivers software tools and micro-services to enable organizations to support the development of digital transformation applications. The digital transformation scope includes and is not limited to cognitive tools such as machine learning and artificial intelligence (AI), the integration and use of Internet of Things (IoT), the integration of blockchain, and analytics to support integration, understanding, and decision-making in flexible and agile big data environments.

9.3.1    Machine Learning

Machine learning is an old concept with formal origins as we know them today from the early 1950s. The statistical and probabilistic principles that support the development of machine learning were limited by the technology development, but the principles and concepts were already defined.

Machine learning, then, is the development of statistical regression models on how computers can learn based on data. A current main research area is for computer programs to automatically learn to recognize patterns and support intelligent decision-making based on data.

The digital in-memory revolution expedited the possibilities to develop integrated solutions for business applications. Some of the most common machine learning methods to train computers are the supervising learning, unsupervised learning, semi-supervised learning, and active learning methods.

9.3.2    Supervised Learning

In supervised learning (also called classification methods), the supervision in the learning comes from data that is labeled or classified in the training data set. For example in the financial accounts machine learning application, the machine learning data set contains all the records of the different banks, and based on each record, the model classifies which record belongs to which bank, how the payments are recognized, and how the records link to the SAP S/4HANA relevant account receivable.

The initial data set is trained, that is, algorithms interact with data, to achieve a statistical level of confidence. The more data is provided in the training data model, the more accurate the confidence becomes.

9.3.3    Unsupervised Learning

When the data set isn’t labeled, then data training becomes a process of clustering to discover classes within the data. The concept of clustering data or grouping data refers to the process group data in similar sets. For example, the SAP Leonardo image classification model collects data from images of a figure, in this case, the image of a “crystal bear.” It then groups each set of data of the image and recognizes its classification based on the data available and the trained model of similar “crystal figures” and properly classifies it. This application integrated into an SAP S/4HANA business scenario could help a crystal figures manufacturer find replacement figures based on a picture of a broken item. After the machine learning runs the process and finds the right figure, it will prepare and set the relevant sales order document to supply the product to the customer. All of this is done automatically—just waiting for the relevant sales supervisor to release the sales order.

9.3.4    SAP Leonardo Integration

The integration of the SAP S/4HANA platform and the SAP Leonardo platform is targeted at automating the current business processes in SAP S/4HANA and turning them into intelligent business processes that automatically learn based on data generated using machine learning.

Consider the following example: The financial application built to clear accounts receivables reads data directly from external banks that receive a deposit from clients, and the SAP Leonardo applications automatically assign the payment to the relevant account receivable. This application solves the issues of the variety of forms and data records of each bank that are unique. The application reads the data from the bank reports, and based on machine learning, trained models determine where the payment was generated, which client is making the payment, and the relevant account receivable and document to be cleared. The statistical level of confidence is around 95% of the machine learning process.

In preliminary internal testing, our teams were able to calculate a reduction of more than 80% of the time required to manually identify those accounts in SAP S/4HANA before the integration of this SAP Leonardo application.

SAP is developing applications across all business functions, and it has a very aggressive road map to integrate your business functions to the next digital level of optimization, automatic learning, and integration of actionable and tactical business operations.

Next, we’ll discuss the architecture of SAP Leonardo, which will help you understand the concept behind this powerful digital innovation system.

9.3.5    SAP Leonardo Architecture

The SAP Leonardo architecture has three main components in its core: business services, functional services, and training and lifecycle management.

The Business Services Component

The integration with functional applications and the SAP S/4HANA system is the main function of this component. This is where the customer, finance, HR and other business functions are integrated with the AI components of SAP Leonardo.

The Functional Services Component

The repository of complex AI or machine learning components resides in the Functional Service component. These components are classified in natural language processing, image and video processing, tabular and time series, audio and speech recognition, and processing subcomponents.

This subcomponent is the foundation of prebuilt SAP Leonardo functions and can be used as the components of your own development of custom applications.

Training and Lifecycle Management Component

Designing, building, and testing AI and machine learning applications require the use of large amounts of data that will help the “training” of the machine learning or AI algorithms.

The concept of training isn’t the traditional concept of educating a person or a group of people on a particular subject. The term is used to define the activities required to collect, prepare, store, and execute AI or ML algorithms that will learn based on the data provided to predict a specific outcome in a level of statistical confidence or the Machine Learning Confidence Score accepted.

Figure 9.20 represents the SAP Leonardo Machine Learning Architecture Foundation. (The figure was prepared by the SAP Leonardo Team.)

Machine Learning Architecture Foundation

Figure 9.20    Machine Learning Architecture Foundation

The Training and Lifecycle Management component is normally used by the team members in charge to define particular machine learning and AI applications for your organization.

SAP Leonardo offers a great variety of prebuilt applications and components that are easy to implement and that don’t require team members specialized in data science to redevelop them.

Deloitte Cognitive Data Fabric Powered by SAP Leonardo

The Deloitte Cognitive Data Fabric (DCDF) is a technical and functional architecture that integrates big data, machine learning, analytics, and SAP S/4HANA. The platform minimizes complexity by automating processes and integrating workflows to simplify business processes. The DCDF is powered by SAP Leonardo.

As you’ve learned, SAP S/4HANA is a digital application that opens new horizons and possibilities for elevating your analytical capabilities to the highest levels of analytical maturity. It also enables you to integrate your sophisticated transactional systems with big data, machine learning, and AI tools; organization supply chain business processes; financial controls; sales and marketing operations; and manufacturing and transportation processes. These new possibilities provide you with a technical tool that will expand your competitive advantages, adapt to the fast digital requirements of your clients, and adjust to the new market realities. The DCDF integrates all these technologies and also facilitates the discovery of business use cases that bring value to your organization. The DCDF architecture approach allows the use of big data generated from IoT data, social media, other integrated systems, and Software as a Service (SaaS) applications, among others.

The big data volumes are managed through the use of SAP HANA smart data streaming via SAP Leonardo cognitive tools and integrating them in actionable applications in SAP S/4HANA and their relevant analytics tools. The use and integration of SAP technologies and platforms with big data repositories solve the requirements of data, versatility, high volumes, and the required speed to process big data in the new digital era.

9.3.6    Integrating a Big Data Platform and SAP S/4HANA

Big data is defined by three Vs:

The huge amount of high-speed data generated by a variety of sources, if analyzed and put in action, can drive improved business performance. The generated data can be classified into two categories:

To take advantage of both of these types of data, SAP recommends leveraging the Lambda architecture (see Figure 9.21). The Lambda architecture provides a framework for ingesting streaming data (e.g., sensor data) as well as data from multiple enterprise sources, which can be used to derive insights across the organization.

DCDF: Lambda Architecture

Figure 9.21    DCDF: Lambda Architecture