Experts question whether Salesforce demo breaches SAP API policy

SAP experts are questioning whether Salesforce's recent public demo of its AI technology would violate the German vendor's API policy, which was introduced in April. During Salesforce's Dreamforce conference in San Francisco, Katie O'Neil, Salesforce's product marketing director, demonstrated an AI agent performing supplier onboarding with SAP using the company's AI agent, Marshall.

The demo showcased how the AI agent could learn the required fields and perform actions within an SAP sandbox, with the AI reasoning through business rules behind the process before executing them.

SAP consultant Mario de Felipe argued that the demo would not be allowed under SAP's API Policy, as it involved impersonation techniques and bypassing SAP-endorsed architectures. According to the policy, any autonomous AI must go through SAP-endorsed architectures to access SAP applications. De Felipe noted that Salesforce had connected to the SAP API to update a supplier and captured the SAP application logic, which is not allowed.

However, Salesforce maintains that the demo did not violate any API use restrictions, as it interacted with SAP through its user interface using service accounts, without directly using SAP APIs. The company pointed out that the API-use restrictions cited in the LinkedIn post do not apply to the activity demonstrated.

Critics argue that the demo does break the rules, as SAP expects documented or endorsed access paths when an LLM or autonomous agent starts planning and executing calls against its applications. SAP's API Policy was introduced to prevent such scenarios, as some critics argue that it would prohibit the integration of AI systems outside its endorsed architectures.