Experts question whether Salesforce demo breaches SAP API policy

Experts are raising concerns over whether Salesforce's recent AI demo, showcased at the Dreamforce conference, could potentially violate SAP's API policy. During the event, Salesforce product marketing director Katie O'Neil demonstrated an AI agent, Marshall, performing supplier onboarding with SAP, using data from the global engineering manufacturer Siemens. The demo purportedly showcased AI reasoning and deterministic execution, allowing users in Salesforce's collaboration platform, Slack, to onboard customers in SAP.

However, SAP experts argue that the demo could be in violation of the company's API policy, particularly when it comes to impersonation techniques and autonomy. According to an independent SAP consultant, Mario de Felipe, the API policy specifically prohibits the use of API calls by "semi-autonomous or generative AI systems that plan, select, or execute sequences of API calls" outside of SAP's endorsed architectures.

Furthermore, the demo involved connecting to and capturing SAP application logic, which goes against the policy.

Another SAP consultant, Marian Zeis, agreed with de Felipe's interpretation, stating that while robotic process automation (RPA) is generally acceptable, autonomous agents planning and executing API calls raises concerns. He also noted that SAP's API policy might seem overly restrictive, but it is necessary for governance purposes.

John Appleby, a long-time SAP expert, suggested that the demo was likely designed to be "edgy" but ultimately not compliant with SAP's licensing agreements. He also questioned the practicality of the policy, as it may not align with how customers typically work.