After years of hearing about the potential of AI and ML capabilities, these solutions are now at a point where they can have actual, tangible, and demonstrated effects on a business and its processes. According to the ASUG Pulse of the SAP Customer 2021 survey, 60% of respondents believe these technologies will have significant effects on their organizations in the next two years.
SAP S/4HANA offers customers a variety of ML and AI tools and capabilities. ASUG recently sat down with Siar Sarferaz, chief software architect at SAP, and Raghu Banda, senior director of AI production strategy at SAP Labs. The two recently published an SAP Press book: Implementing Machine Learning with SAP S/4HANA. Read on to see Siar and Raghu discuss their book, the AI and ML capabilities available to SAP S/4HANA customers, and how customers can roll these functionalities out in their business.
ASUG: Tell us about your backgrounds and how they have served you as you were writing this book.
Siar: I am a chief software architect at SAP. In this role I drive digital transformation by focusing on AI and predictive analytics. I began my career as a method researcher at Siemens AG, before moving to SAP, where I have now worked for more than 20 years, holding various positions. I am the lead architect for ML implementation in SAP S/4HANA and am in charge of all concepts for infusing intelligence into business processes. I studied computer science and philosophy and hold a Ph.D. in computer science.
Raghu: I am a product manager at SAP working on creating a strategy and definition to infuse AI technologies into the SAP S/4HANA business processes by leveraging the different tools and technologies available internally and externally. I began my career working for a startup firm (founded by a group of Indian Institute of Science back in the mid-90s in Bangalore) with clients such as NASA, BHEL, etc. Later I worked with Wipro, one of the big five consulting firms based out of India, and then moved to the U.S. in 1997. I joined SAP in 2001 and worked in various roles, from research and development to solutions engineering and product management. I was fortunate enough to be involved with predictive analytics and ML ever since SAP started the journey into the AI world in 2011 and 2012. I hold a bachelor’s degree in computer sciences and engineering. While getting that degree in India, I authored a brief write-up on AI during my final semester in 1994. Currently, I am pursuing a global leadership and business management certification course with the international business school INSEAD and will be graduating this year.
ASUG: You open your book by discussing predictive intelligence. What is your working definition of predictive intelligence? How is it being embedded into the SAP Intelligent Enterprise?
Raghu: I coined this term “predictive intelligence” for SAP intelligent offerings, though it is not an official term. Intelligence has been embedded into SAP offerings in the form of rule-based analysis before SAP S/4HANA was released. But there has always been this need to understand and anticipate the customers’ needs ahead of time and suggest recommendations accordingly. Even intelligence of a business process can be analyzed, predicted, learned, and fine-tuned over a period of time with the historical implications and the other variable factors–which is what explains “predictive intelligence” in a nutshell. Embedding this predictive intelligence happens in different ways: embedding them into the SAP S/4HANA business processes and leveraging the SAP Analytics Cloud to enhance the business user experience. The book talks about all these in detail.
ASUG: Can you give us a high-level view of how predictive analytics and ML have been leveraged in SAP solutions before the introduction of SAP S/4HANA?
Raghu: Like I said earlier, even before the inception of SAP S/4HANA, SAP—and other ERP companies besides SAP—always did some kind of analysis and embedded rule-based intelligence into the business process of an organization. SAP, being a leader in the ERP world, has pioneered the journey of enhancing rule-based intelligence with predictive analytics and ML—even before the inception of SAP S/4HANA. Using the home-grown predictive analytics technologies and the algorithms from the SAP HANA library, we have built packaged apps that sit on top of the core ERP across different domains. The customers could leverage the benefits of these SAP HANA predictive algorithms that served as extensions to the core ERP processes, even before the inception of SAP S/4HANA. To also help the customers realize this more easily, we have created a cloud image using our SAP HANA appliance that could be hosted on Amazon Web Services (AWS). This was helpful to understand the customer pain points in those business processes. This definitely gave us an understanding of how the intelligent technologies could provide exponential support if built into SAP S/4HANA.
ASUG: What are some of the key tools and services available to SAP customers they can leverage to get access to the full capabilities of predictive analytics and ML?
Raghu: SAP has been a pioneer not only in developing business software, but also in leveraging the intelligent technologies to infuse intelligence into the business processes. There are different ways for operational users, business users, data scientists, business analysts, application developers, or SAP consultants to access the already available predictive and ML functions or build new functionality. Broadly there are three different ways of leveraging predictive analytics and ML with SAP S/4HANA: embedding them into the SAP S/4HANA processes, consuming the ML services available from the SAP Cloud Platform, or enhancing the predictive capabilities with SAP Analytics Cloud. The book details this completely in chapters three, four, five, and six.
While embedding predictive analytics and ML into SAP S/4HANA uses the Intelligent Scenario Lifecycle Management (ISLM) framework by leveraging the SAP Fiori apps with SAP HANA Automated Predictive Library (APL) and SAP HANA Predictive Analysis Library (PAL) algorithms, the approach of consuming ML services that leverages the AI Foundation and the SAP Data Intelligence tools. Additional extensions of predictive analytics services can be leveraged using the SAP Analytics Cloud features such as the smart insights, search to insights, and smart predict, etc.
ASUG: What is the architecture of ML and predictive analytics in SAP S/4HANA? How are these technologies embedded into the ERP solution?
Siar: ML and predictive analytics are natively integrated into SAP S/4HANA and can be easily used across the entire organization. However, incorporating ML and predictive analytics capabilities into ERP solutions is a challenging task. This is due to the high complexity of ERP systems. SAP S/4HANA on-premise stack, for example, consists of more than 250 million lines of code and 140,000 tables. It supports 25 industry verticals, localization for 64 countries, and thousands of business processes. This means whenever you touch such an ERP system to add new functionalities, it is a very complex task. That is what we tried to solve with the proposed architecture, which we explain in the book. The architecture is also the foundation for how we are implementing ML in SAP S/4HANA. The proposed architecture solves basically two substantial challenges. First is to systematically build ML and predictive analytics into business processes so that intelligence is provided to the right person in the right place at the right time. Second is to make ML and predictive analytics enterprise ready so that qualities like compliance, security, life cycle, scalability, robustness, extensibility, configurability, operations, supportability, globalization, or auditing are ensured.
ASUG: Let’s move to the technical implementation of ML and predictive analytics. What are the different implementation approaches customers can take? What are their benefits and drawbacks?
Siar: Prior to defining the solution architecture, we intensively analyzed the ML and predictive analytics use cases in order to understand the technology requirements. Thus, we identified different application patterns, which required also disparate implementation approaches. Simple use cases like ranking, categorization, and prediction can be solved with classic algorithms like classification, clustering, regression, or time series analysis. Usually, those algorithms do not allocate much memory and CPU time. Thus, they can be implemented within the SAP S/4HANA stack, where the application data for model training and the ML consuming business processes also are located. This embedded ML architecture has very low TCO. This is because everything is on the same platform, so we don’t need to move data.
However, there are also more complex use cases like image recognition, sentiment analysis, and natural language processing, which require deep learning algorithms based on neural networks. For model training, usually these kinds of algorithms demand a huge volume of data and graphics processing unit (GPU) time. Therefore, these kinds of scenarios are scaled out to a ML platform to keep the load in the transactional SAP S/4HANA system low. This side-by-side ML approach provides scalable infrastructure with state-of-the-art algorithms but causing higher TCO.
ASUG: What about business implementations? How does a company’s industry affect the way it would roll out ML and predictive analytics in SAP S/4HANA?
Siar: ML and predictive analytics enable disruptive innovation in many domains and have a major impact on the enterprise software market. Today, most organizations have enterprise software that uses rule-based processing to automate tasks. This opens the door for innovation. It’s very powerful and helpful, but this approach can’t learn and improve with experience as humans do. ML helps to close this gap. Self-learning algorithms take enterprise software to a new level. These algorithms learn from the data and from the changes, and adapt themselves accordingly.
Market analysts agree that ML is a core driver for economic development. McKinsey and Company estimates that the global ML applications market will reach a value of $127 billion by 2025. Benefits will be gleaned in almost all industries and sectors. For example, validating that a product has been produced exactly to its specifications and configuration is an important step in final quality assurance and readiness checks for product shipments. Image-recognition algorithms support this via visual product quality checks. Thus, the accuracy and automation level of quality processes in production can be increased, which leads to fewer returns, higher customer satisfaction, and better profitability.
ASUG: What do the next three to five years hold for the evolution of ML and predictive analytics as they relate to SAP S/4HANA?
Siar: As already mentioned, ML and predictive analytics positively impact almost all industries and sectors. The benefits go beyond cost savings. They also include identifying untapped opportunities, exposing hidden risks, optimizing operations, and creating more personalized customer services and enhanced user experiences. Thus, SAP S/4HANA will provide more and more intelligent business processes. However, the objective will be on assisting the users by automating repetitive tasks and freeing up humans to focus on high-value tasks. Early adopters of ML and predictive analytics can disrupt entire industries and formulate completely new business models. However, the success of ML depends on its broad social acceptance. SAP defined ethical principles to steer the development and deployment of ML software to help the world run better and improve people’s lives.
Raghu: That was very well explained by Siar. I can also add a couple of points here. ML and predictive analytics are no longer the best-kept secrets at SAP, and we have a lot of enhancements coming into the business processes with SAP S/4HANA. You will see how SAP is transforming the enterprise world with the intelligent enterprise. As you know, 77% of world’s transactions touch an SAP system. The journey just started and there is a lot of potential for SAP not only to be a leader in the intelligent enterprises, but also how you can intuitively handle all these in the cloud. We will be having frequent openSAP courses on the topic, continued conversations in our blog series, and a newly started podcast series as well. The difference with SAP is that you will see intelligent end-to-end processes that leverage AI to the full extent and not just a use case here and there. SAP understands the customers pain points to help them get to a brave new world of Intelligent Enterprises.
ASUG: What do you want readers to take away from this book? What do you want to empower them to be able to do?
Siar: ML and predictive models have been created by data scientists for decades. However, often these models resided in special tools, were consumed by experts only, and therefore added hardly any value for enterprise applications. The book closes this gap and proposes how to systematically built-in ML and predictive analytics into business processes and to ensure the required enterprise qualities. There are three main objectives to this book. First is to explain the ML and predictive analytics tools, services, and architecture of SAP S/4HANA. Second is to teach how to implement your own ML and predictive analytics applications based on the SAP S/4HANA programming model. Finally, we want to describe the ML business applications that have been already delivered with SAP S/4HANA.
Make sure to register for ASUG Best Practices: SAP S/4HANA. Spread across four weeks in March, this virtual conference series will take attendees through every step of the SAP S/4HANA journey, from developing a strategy to effectively adopting the ERP solution.