While SAP is a front runner among global companies addressing the ethics around artificial intelligence (AI), there is still plenty to learn.
In September 2018, SAP released guidelines for the ethical use of AI, which declare, for instance, that AI built by SAP shall be human centered, guarantee transparency, and keep data secure.
While SAP was the first company in Europe to put such initiatives in place, many other large enterprises, including Microsoft and IBM, have taken similar steps toward ethical AI. With no globally recognized norms or policies in place yet, SAP has emerged as a leading voice in the global conversation.
Markus Noga is head of SAP Cloud Platform Business Services, which includes AI, as well as a member of the steering committee,
Noga explains the mandate he sees for SAP: “Not only should our own AI be subject to ethical requirements, but we also want to help our customers to build and use AI ethically. Moreover, we aim to contribute our expertise to the discussions taking place at a global level.”
According to a November 2018 study by consulting firm McKinsey, nearly half of enterprises have already adopted AI while another 30 percent are currently running pilot AI projects for future deployment. A further 44 percent plan to look to providers like SAP for a majority of their AI needs.
In the public discourse, AI can be a controversial topic. When machines make decisions that affect the lives of human beings, it is a sure bet that some of the people affected will disagree. That leads to more complex questions with which SAP will have to grapple: How can those that develop and sell autonomous systems ensure that both the creator and the customer can feel moral confidence in an AI’s decisions?
Ethical AI at SAP
At SAP, the Intelligent Enterprise strategy means that there are different ways the company builds and sells AI.
On one hand, there is AI that is directly embedded in SAP products, such as SAP Cash Application software, which helps automate invoice reconciliation in the SAP S/4HANA Finance solution. For such use cases, SAP is working to implement procedures to better understand the intent of a use case and to help ensure that the concept complies with ethical guidelines.
On the other hand, when it comes to AI platforms, like the SAP Data Intelligence cloud service, an SAP customer can use the technology to build any AI use case. According to Noga, that means one must critically question some ideas and find a way to lead innovation toward ethical use cases. This is why SAP is exploring how to provide guidelines or even terms and conditions to customers in order to guide the ethically appropriate use of the AI platform.
“Out of consideration for the customers and colleagues involved, it is difficult to talk about concrete use cases,” Noga shares. “But I can say that cases have been sent to us on a wide range of issues — from facial recognition to government use of AI to cases of negative environmental impact.”
Which is why guidelines are so important: “We need to reconcile customer needs with the real possibility that SAP might be asked to do something against its policies.”
Internal and External Committees Work Hand in Hand
Noga represents SAP in the EU Expert Group on Artificial Intelligence, which is chaired by Pekka ala-Pietia, a member of the Supervisory Board of SAP SE. The European Commission is increasing its annual investment in AI by 70 percent under the Horizon 2020 research and innovation program. It will reach €1.5 billion for the period 2018 until 2020.
Earlier this year, the group presented its politics and investment recommendations for AI as well as guidelines for trustworthy AI. SAP also hosted an EU researcher in Walldorf, Germany, who investigated how large corporations can create reliable AI.
“We have also developed an openSAP course on AI ethics together with the members of the EU Expert Group,” Noga says. A handbook has been released based of both SAP’s guiding principles and the EU trustworthy AI assessment.
SAP has also implemented an external AI ethics advisory panel that meets quarterly.
“We discuss concepts on how different cultures see AI and how we can best align the ethical values of the different regions with our own ethical standards. We have also debated some fascinating legal concepts. The AI tools we provide are often used by our clients in a way that we may not even be able to imagine.”
Empowering Employees to Create and Sell AI Ethically
Looking back at SAP’s first year with ethical guidelines for AI, Noga summarizes: “The work on AI ethics is not finished, one year in. Indeed, there is still much to be done to operationalize the guiding principles in order to ensure that all SAP employees are empowered to create and sell AI in an ethical way.”
Mary Carol Madigan, AI product team lead for AI ethics activities at SAP, contributed to this story.