AI and SAP: A connection with a future?

Technologies that take on or even surpass human characteristics clearly exert a great appeal - and not only in cinema. Science has also been devoted to the topic for decades.

The beginning of AI as an academic discipline is considered to be the workshop "Dartmouth Summer Research Project on Artificial Intelligence", which took place in the summer of 1956 at Dartmouth College in Hanover (New Hampshire, USA).

In fact, progress after the Dartmouth Conference was initially quite manageable. In the recent past, however, various developments have helped to make the visions of the future held by female scientists and Hollywood directors more and more real:

The performance of computers has increased immensely and rapidly; companies are launching more and more digital and networked products on the market; in addition, enormous amounts of digital data are also being created in society and the economy.

What AI can and cannot do

So far, however, there is no valid definition of AI or what we mean by intelligence. What we can say is that computers are now capable of making decisions and acting autonomously to a certain extent.

To do this, they recognize and analyze patterns in data, language or images. And in many cases, they are much better at this - more comprehensive, faster and more error-free - than we humans are.

A whole range of individual technology-based methods are used - for example, machine learning, which is often mentioned in the same breath as AI.

All of this falls into the category of "weak AI": A software independently solves specific application problems. However, it does not achieve a deeper understanding or develop overarching cognitive abilities - which are the decisive criteria for "strong AI".

Some scientists assume that "strong AI" may even develop into a superintelligence with (simulated) consciousness in the future, quickly escaping human control. For the time being, however, this remains science fiction.

AI potential of 430 billion euros

Attention is therefore currently focused almost exclusively on "weak AI" and thus on approaches with a concrete benefit that can be easily controlled and are less sensitive from an ethical perspective.

According to forecasts by the consulting firm PwC, (weak) AI alone can generate value added of 430 billion euros in Germany by 2030. In another study, PwC examined the areas of application for artificially intelligent technologies.

A large proportion of the companies surveyed are currently using AI to automate existing business processes and for data analysis. AI is particularly helpful when it comes to individual, complex issues with an enormous amount of data.

The fact that AI has enormous potential for every company is fairly obvious. For SAP users, this raises the question of which technologies SAP offers them to upgrade the existing system landscape. And indeed, there are a few applications in the portfolio that sound like AI.

Leonardo: Approach to AI

It is particularly worth taking a look at the SAP Leonardo innovation platform, which was launched in 2017 to great acclaim, but which has become somewhat quiet. SAP has assembled various applications and microservices for a whole range of hype topics here:

For example, in addition to machine learning, IoT and blockchain were to be included as innovative technologies in the SAP universe. In the meantime, SAP seems to have abandoned this strategy.

Instead of being a comprehensive technology provider, Walldorf apparently wants to reemerge as a practice-oriented solution provider. This is supported by the fact that SAP Leonardo was initially subsumed under Intelligent Technologies and is now gradually being integrated into the entire product portfolio.

In concrete terms, this means that many of the technical and functional services that were created under the umbrella of SAP Leonardo will be retired in the next few weeks.

Automation through ML and AI

Instead, SAP focuses on the business processes and the requirements of the companies and integrates the innovations and individual solutions for this purpose. In this context, SAP offers customers essentially three paths into the world of artificial intelligence, which are aimed at different target groups and complement each other.

In the new ERP system S/4 Hana, SAP has embedded some AI functionalities that make individual applications of S/4 intelligent. The goal is to automate routine tasks in order to relieve employees so that they can concentrate fully on complex cases. Several approaches exist for this.

Predefined scenarios: SAP delivers predefined scenarios that only need to be trained using a company's historical data and can then be used to automate routine tasks. This works, for example, for the assignment of incoming payments to open invoices or the creation of purchase orders.

Robotic Process Automation: With Robotic Process Automation, companies, departments or users can create their own AI-based automations in which bots process recurring tasks based on recognized patterns.

Situation Handling: The new paradigm Situation Handling is concretized by the predefined scenarios and Robotic Process Automation: Users in the business department are only shown unusual events to which they can react in a targeted manner.

SAP Conversational AI: SAP Conversational AI is also used in the context of S/4. The platform, which can actually be used to implement chatbots as part of SAP Leonardo, enables interaction via speech within S/4.

SAP Analytics Cloud

As the strategic analytics solution of the future, SAP Analytics Cloud (SAC) includes AI-based predictive functionalities called Smart Predict by SAP in addition to traditional BI functionalities focused on the past and present.

The tools are designed for citizen data scientists - i.e. for ambitious business users. In view of the continuing shortage of data scientists, they are currently receiving more and more attention.

For example, the Smart Predict applications contain everything needed to equip business users for typical predictive procedures: for example, to process classification, regression and time series-based questions and thus forecast sales. The AI models required for this can be trained with the company's historical data.

SAP Data Intelligence

SAP Data Intelligence, the most elaborate AI solution from SAP, runs on the SAP Cloud Platform (SCP) and can best be understood as the successor to SAP Leonardo. However, the business focus is again clearly recognizable here:

SAP Data Intelligence combines the somewhat more conservative business world with the open source world in which Data Scientists feel at home, and is open for corresponding applications: for example, for Jupyter Notebooks from Project Jupyter, for Python, and for Python-based machine learning frameworks such as pandas, scikit-learn, or TensorFlow.

Focused AI strategy

The data models created in this way can be transferred to the SAP environment with SAP Data Intelligence, where they can be further processed, automatically tested, and finally delivered to a highly available and scalable productive landscape - including subsequent performance monitoring. And very important in the business context: AI scenarios can be audited so that they meet the requirements of an audit, for example.

SAP has recognized that AI is an important - if not the most important - topic for the future and is committed to expanding its existing portfolio to include corresponding technologies.

Always with a view to the business requirements of the companies. However, SAP continues to hold back when it comes to active AI research. Here, companies such as Facebook, Google and Microsoft, as well as open source initiatives, are the driving forces.

Conclusion and future

If companies want to implement AI scenarios, they should, in our opinion, first evaluate (of course, after the benefits of the respective use case have been fundamentally examined) what exactly they need and to what extent SAP provides the corresponding technologies.

Where SAP leaves gaps, these can possibly be closed with SAP-based applications from SAP partners. Or third-party solutions are used, which are becoming increasingly easy to integrate.