On March 24 and 25, TAC Insights and SAP hosted the SAP EHS and Product Compliance Conference in Amsterdam. In a great atmosphere, the SAP EHS and Product Compliance community met in person and benefitted from a full agenda and various networking opportunities. For more information, see this blog post.

One hot topic at the conference was agentic AI and its use in EHS and Product Compliance processes. Agentic AI refers to autonomous systems that act as agents. These make decisions and execute multi-step tasks to achieve specific goals without constant human guidance.

In his conference keynote, Gunther Rothermel, Chief Product Officer and Co-General Manager for SAP Sustainability, explained SAP's product strategy for sustainability. He also described how embedded AI can drive efficiency and reduce operational risks across the product lifecycle. SAP is developing AI agents for EHS and Product Compliance. The EHS Workplace Safety Agent transforms safety observations into continuously mitigated risks and up-to-date safety instructions. It uncovers overlooked hazards, enhances risk assessments, and recommends corrective and preventive actions. The agent automatically generates or updates safety instructions for a safer, more compliant workplace. In Product Compliance, the planned GHS Classification and Labeling Agent supports and accelerates the process of classifying and labeling products based on the Globally Harmonized System (GHS). The agent is based on the latest regulations, reduces manual workload, and minimizes errors. Through seamless integration into the product compliance solution, the agent determines and collects the required input data. It then calculates and proposes the GHS classification and label elements for use in subsequent product compliance processes. The AI agent checks for data gaps and validates data against regulations to ensure accuracy. Users can review the results using explanations of how the agent reached its conclusions.

In a panel discussion, Gunther Rothermel and two SAP customers discussed the real-world impact of AI in EHS and Product Compliance processes. All panelists emphasized the importance of clean and reliable data as a foundation for AI. AI was also part of several customer and partner presentations. These presentations included, for example, the role of AI in safety data sheet authoring. In the Discovery Hub room, attendees could get hands-on experience with the SAP solutions for EHS and Product Compliance. In the EHS room, participants could, for example, try out the AI-assisted compliance management and AI-powered safety instruction generation for risk assessments.

Toward the end of the conference, participants discussed how agentic AI could support or change their work in a lively roundtable. Some SAP customers said that they already use agentic AI to propose potential substance substitutions, explain classifications through a Hazard Communication Chatbot, and support recipe development. One customer expressed concerns that the use of AI is restricted in European countries due to labor laws. The customer also noted that they don't want to share their formulations in the cloud, for example, for AI-based classifications.

When asked about ideas and wishes for use cases supported by agentic AI, customers discussed the following:

• AI-supported analytics could provide implications for the whole product portfolio after a regulatory change. Going one step further, AI could propose potential future regulatory changes and their potential impact on the product portfolio.
• AI could simulate compliance before a product is released.
• Some sustainability requirements are vague, complicated, and data is often missing. AI could propose values for missing data. In general, AI could help identify data gaps and clean up data.
• AI could help analyze customer requests.
• Safety data sheets contain a lot of information and are provided in various formats. AI could extract the key information.
• SAP’s new Product Compliance solution uses a different data model than the classic solution. For the migration from the classic to the new solution, data needs to be mapped. AI could support this mapping.

What do you think about these use cases? Do you have additional ideas? Provide them as comments on this blog post.