The emergence of artificial intelligence (AI) tools such as AI agents, will bring more automation and efficiencies across work environments, but it does not mean humans can be left out of the equation.
SAP, for instance, is heralding the transformation of business operations through “autonomous, collaborative AI agents”, where the next frontier of AI will unify people, data, and processes to augment outcomes.
Specifically, its AI copilot Joule is built with the ability to autonomously “reason, plan, and execute” collectively across multiple applications.
This means, for example, its Joule platform can bring together AI agents to settle a payment dispute, which involves monitoring and assessing accounts, analysing data, providing recommendations, and generating reports.
However, it does not mean humans are no longer needed to oversee such tasks, said SAP’s CTO and chief AI officer Philipp Herzig, said in an interview with FutureCIO.
The German software vendor adheres to a “strict” ethics policy that requires human oversight in every use case for AI, whether it involves traditional AI tools or emerging ones such as AI agents, said Herzig.
Humans remain in the loop and always have to be the last decision maker, he said.
Asked if this was necessary even as AI advanced, he replied affirmatively, noting that SAP initiated this AI policy long before EU AI Act was established and before industry discussions centred on the need for GenAI (generative AI) regulations.
“It’s an important design principle…[and one] we believe is key to create trust and necessary, for many years on, to make sure whatever goes into an SAP system is sound and safe from a compliance perspective,” he said.
It also is critical that the AI systems can carry out tasks according to a company’s business rules, which Herzig noted can prove to be a tough engineering challenge.
This requires the ability to leverage large datasets to drive the right outcomes with respect to logic and reasoning, for various business functions, such as HR and financial planning, he said.
Organisations have to ensure AI-driven workflows do not deviate from business rules that still need to be applied to these tasks, he noted.
This involves providing the AI models with the metadata, description of APIs (application programming interfaces), and semantics, amongst others, so the AI agents can carry out the various steps to fulfil a task. Fulfilling a customer order, for example, will require checking the inventory and looking at the production capacity, he said.
The AI algorithm needs to have the capability to fulfil such tasks in the right, logical way, that adheres to all required business rules. He pointed to SAP Knowledge Graph, which he said was designed to address this for enterprise customers.
The knowledge graph taps the software vendor’s data, including business process metadata and data models, to ground AI models. To carry out a payment dispute, for instance, Joule AI agents would leverage the knowledge graph to pull together related orders, invoices, and payments from SAP applications, to carry out the task.
Powered by contextual data across core business functions, such as CRM, HR, and ERP, SAP further believes its agentic AI products can “seamlessly integrate” business operations and transform “the future of work”.
Change management plays an important role as workplace evolves
It highlights the importance of change management as the way people work is evolving, said Herzig, who assumed the newly created role of chief AI officer in February last year and in January 2025 added CTO to his resume.
His predecessor and former CTO Juergen Mueller stepped down last September following an incident at a company event, during which Mueller was deemed to have behaved “inappropriately”.
Asked what typically would be discussed during his meetings with other SAP’s leadership team such as the CFO and HR head, Herzig said: “A discussion we always have is the impact on human. Change management with AI is a very important topic because it’s very clear that the way people work is fundamentally changing.”
Noting that SAP’s 40,000 developers worldwide use various GenAI and other AI tools, he underscored the need to ensure such tools are adopted to improve efficiencies and to increase AI literacy across the organisation.
Any concerns that AI will replace jobs also need to be addressed, he said, adding that companies should work to support employees along their AI journey.
In a separate announcement this week, SAP Labs Singapore said it was working with the National University of Singapore (NUS) to hire and train nine researchers in its AI team over the next five years.
Parked under the Industrial Postgraduate Programme, this initiative will see new researchers recruited from NUS and hired to work on AI research areas that “complement” the work of SAP Labs’ local team.
The Singapore lab was established in 2022 along with the pledge to hire 200 AI engineers. It currently has a headcount of 420, with hires from local universities and institutes of higher learning, SAP said. It added that the local engineering team also supports the company’s agentic AI roadmap and capabilities.
