Artificial (Enterprise) Intelligence: the future of SAP in a new era of technology

In the past, SAP has stated that Artificial Intelligence (AI) and Machine Learning (ML) would not be a core focus for their solutions, even though it did already weave several applications of these technologies in its Business Technology Platform (BTP) years ago. Today, legacy solutions such as SAP Predictive Analytics and the Leonardo-suite of applications no longer exist, but their functionalities and new ones have been integrated into both SAP Business Solutions (e.g. SAP S/4HANA and other components of Cloud ERP, as part of what SAP calls ‘Business AI’) and data and analytics solutions such as SAP Data Intelligence and SAP Analytics Cloud (supported by the APL and PAL within the HANA database). These ‘embedded’ functionalities are often tailored to the more classic variants of Machine Learning use cases, such as regression, forecasting and classification.

Aside from this, we also have the AI portfolio on the BTP, which consists of SAP AI Launchpad and SAP AI Core. The Launchpad allows customers to manage AI models across the organization from a central access point, as well as connect multiple AI runtimes with each other. SAP’s AI Core is the main platform through which developers can create tailored services more aimed at deep learning, either through a graphical UI, Python or SQLScript respectively. Support for Bring Your Own Model (BYOM), through for example TensorFlow, has also seen an increased number of use cases in the past few years. Machine Learning functionalities in SAP Data Intelligence such as the support for both PyTorch and TensorFlow 2.0 in the deep learning library or the use of the JupyterLab environment with Python have been applied with much enthusiasm by the (SAP) data science community as well.

Seeking to extend its affiliation with the growing influence of AI, SAP cemented its commitment to LLMs at Sapphire 2023 by publishing a reference architecture that allows developers to bind BTP applications to external LLMs (on hyperscale environments such as Azure) through SAP’s proprietary Cloud Application Programming (CAP) model. SAP is also working to incorporate different LLMs as well as LangChain (a framework used to simplify app development through LLMs) into the BTP. The Business AI components we mentioned earlier will also be expanded upon with new generative AI scenarios. SAP further demonstrated its broad approach towards AI as an open ecosystem by strategically investing in AI companies Anthropic, Cohere and Aleph Alpha in July. On the data and analytics front, we see roadmaps being shaken up to accommodate AI-related features and development paths. From Natural Language Processing (NLP) capabilities in SAC, realized through SAP’s 2022 acquisition of Askdata, to its partnerships with Databricks and DataRobot support of SAP Datasphere; AI and machine learning-functionalities are increasingly taking center stage in SAP’s solution briefs.

Now that we have established that SAP itself is seemingly shifting gears when it comes to applying these technologies, what do we actually see happening at customers and in the market at large?

SAP customers are generally larger enterprises that host vast and intricate IT landscapes, seldom based exclusively on SAP solutions. What is often the case is that SAP technology forms a backbone for one or more primary business processes, making it an important link in the functioning of the organization. As SAP consultants, we have seen that for years, SAP’s ecosystem has been relatively closed, focused on integration and cooperation with peer SAP solutions (the infamous ‘vendor lock-in’). However, with the exponential growth of (types of) data and the mainstream onset of AI and ML technologies, both industry competitors (Google, Amazon and Snowflake, to name a few) and open source technologies have forced a shift in this once well-established paradigm. Flexible cloud-based data warehouses, open APIs leveraged through Python or R and the rise of LLMs have all changed the way through which customers evaluate the choices they make in their data landscape. Combine these technological influences with a new generation of digital-natives, and we open up a whole new world of data-rich possibilities.

Another factor to take into consideration here is the evolution of the ‘data scientist’-profession. Data science as a field has rapidly developed over the last decade, with the inclusion and expansion of AI and ML, and thus the profession has as well. Both analytical and technical competences have risen as requirements for the average data scientist position, resulting in three separate but interrelated roles required to effectively (practically) deploy a Machine Learning model. As can be seen in the image below, each step in the life cycle of a Machine Learning project requires specific skills; a combination of these skills is the key to utilizing Machine Learning to improve (business) processes.

So, while programming knowledge in for example Python is important, statistical (algorithm) know-how is crucial, and the ability to wrangle data and train a model is paramount, modern data science cannot be put into practice without these combined capabilities. After all, having a state-of-the-art, high-precision and structurally reliable Machine Learning model is great, but it will not yield any business benefits unless it is deployed and maintained properly. That is why organizations across the world will not be looking for just the classic data scientist, but for data-and-machine learning engineers that synergize with that scientist as well. Understanding a challenge, gathering the right data to tackle it, training a ML model and finally applying and maintaining it, is an iterative process that will be present in more and more corners of an organization’s IT architecture. This is Expertum’s take on today’s developments and something that we are already experiencing (and are a part of) at some of our customers. While our consultants already have years of experience with many of the data engineer’s processes, we are also working to develop our expertise in the context of the data scientist and machine learning engineer roles to help our customers get the most out of their SAP solutions in terms of AI and ML.

In a world that will be increasingly driven by Machine Learning technologies and where the expertise underpinning these concepts is developing at a lightning pace, SAP has to evolve to stay relevant and support its customers’ (future) business goals. From more mainstream-oriented steps such as the inclusion of AI-assistants in SAP Grow and Rise, to the broader adoption of generative AI from a development perspective and the integration with third-party technology (e.g. IBM Watson); SAP does not seem to be hesitating.

At Expertum, we recognize and even encourage this movement, because if one is not progressing, one is regressing. The role of SAP technology in organizations will not change overnight, but the degree to which it iterates in the context of contemporary developments will determine the role of Walldorf-developed solutions in tomorrow’s world. We are also cautious about the impact these changes will have on established business processes, solution developments and people’s roles. Having the knowledge and expertise to realize successful applications of Machine Learning technology is one thing; keeping these applications sustainable in the larger context of the corporate IT landscape is another. The advance of AI and ML will have organizational, technological and social implications, none of which we can neglect as consultants. If the past few months have shown us anything, it is that by expanding existing Machine Learning applications within its portfolio of solutions and investing in new, AI-driven initiatives, SAP is shaping up to meet the challenges of tomorrow’s intelligent enterprise. This movement is something no one should miss, and at Expertum, we are eager to help you discover how you can leverage AI and ML within your SAP solutions to take that next step.