As artificial intelligence (AI) continues to evolve for the world of business, including through the recent emergence of generative AI, building out a robust ethical framework that provides governance, principles, and policies for AI-based solutions and their strategic development will be increasingly important for businesses to consider.
In a recent ASUG webcast, Puntis Palazzolo, AI Ethics Lead at SAP SuccessFactors; and Robert Richardson, Strategy Advisor, Human Experience Management at SAP, detailed the fundamentals of ethical AI strategy at SAP, intending to open an inclusive dialogue on AI ethics while offering comparative models for enterprises to adopt in strategy and practice.
AI usage is on the rise in the ASUG community, especially as SAP continues making this emerging technology a priority for its member base. The 2024 ASUG Pulse of the SAP Customer research shows ASUG Members’ interest surging in AI use, benefits, and impact, balanced by concerns about current generative AI solutions, implementation, and skill sets. This growth in interest and adoption only heightens the need to develop and adopt AI frameworks to facilitate ethical AI adoption in the SAP ecosystem.
AI is ‘More Salient’
“AI is not new” to SAP, said Richardson. “I’d just like to say that it’s [now] more salient. With the advent of generative AI, it’s not a surprise that it’s on the public conscience right now.”
Richardson discussed the SAP Business AI Product Development Principles, a “framework for developing our products,” during the webcast. The framework principles and definitions, which provide a strategic foundation for SAP, dictate that AI implementations should be:
- Relevant – Using the best partners, SAP technologies, and an outcome-driven focus to maximize value to customers and employees.
- Reliable – Accurate, generative AI outputs must be trustworthy and accessible, ensuring AI can be used by all employees.
- Responsible – Ethical approaches that mitigate customers’ legal and ethical concerns and encourage safe, secure, and transparent use of employee data.
While the two presenters noted they generally focus on HR customers leveraging SAP SuccessFactors, Richardson said, “One of the conversations that is often getting lost [in AI] is the safety and security of these solutions. In HR, these are high-risk use cases. We’re making decisions about real people’s lives that affect their families and individuals. Our ethical framework needs to make sure [our solutions] are responsible.”
Palazzolo pointed out that privacy is among the five primary challenges and “ethical dimensions” that SAP grapples with each day in AI product development, alongside automated decisions, fairness, transparency, and human dignity.
‘Put Humans at Center’
“We need to ensure that AI respects human rights and puts humans at the center and allows for human input and evaluation instead of only relying on machines,” Palazzolo said.
Transitioning to more detail on SAP AI practices and processes, the experts reminded the audience that there is no single governing authority or AI roadmap.
“One thing I spend a lot of time talking about with clients is how do you manage governance around this when you don’t know what the roadmap looks like?” Richardson remarked.
At SAP, AI ethics governance operates through three interconnected bodies:
- An AI Ethics Advisory Panel of internal and external experts providing input on the company’s AI guiding principles and how to operationalize the principles.
- An AI Ethics Steering Committee developing and enforcing the principles through active company-wide processes—such as the Ethics Policy Framework—which also assesses high-risk AI use cases.
- The Trustworthy AI Workstream creating the means necessary for AI development processes, implementation, and compliance.
Policy Framework
In addition to the principles, the governance entities provide a proactive and responsive presence applying the SAP AI Ethics Policy Framework, which acts as a filter for every SAP AI-related use case and development. The three-pillar Policy Framework mandates:
- Human Agency and Oversight for safeguarding human autonomy, particularly for automated decision-making.
- Addressing Bias and Discrimination to identify patterns of marginalizing, inequality and discrimination must not be encoded into AI.
- Transparency and Explainability checks that prioritize both the transparency of the development process and an AI’s decisions.
Both Richardson and Palazzolo emphasized it’s important that SAP customers know of and understand the AI principles, policy, and governance practices the company continues to expand as AI advances and challenges business in new ways.
“Certainly, the AI is not going to be the one that is culpable for any errors that happen,” Richardson said. “It’s going to be you or your organization.”
“That’s exactly why SAP is trying to put all these pieces for AI ethics in place because our customers hold us responsible at the end of the day,” added Palazzolo.
Watch “Towards Ethical AI: Navigating Challenges and Establishing Worthiness” on demand here.