At SAP Connect in Las Vegas today, CEO Christian Klein declared that “The Autonomous Enterprise has arrived.” While this framing serves as the company’s vision for the future, the actual substance of the event focused on moving Joule Assistants from conceptual pilots into specific, operational business workflows. SAP is positioning these agents to execute tasks directly within existing systems, relying on established data, authorizations, and audit trails to maintain control.
The shift toward operationalizing agents is visible in several named customer deployments. Novartis is piloting a Joule Sourcing Assistant, while Nestlé is deploying an Accounts Receivable Assistant across 140 countries to prioritize collections based on risk. PwC is utilizing a Billing Assistant to identify potential issues in accounting, and Morgan Foods is applying agents to supply chain management. SAP also reported that its own internal deployment of Joule Work across 110,000 employees has yielded a 20% productivity gain in finance, HR, and procurement. It is important to note that these figures and customer testimonials are provided by SAP and have not been independently verified.
The technical delta here is the attempt to standardize how these agents interact. SAP and Google Cloud have co-developed the Agent2Agent protocol, an open standard (v0.3.0) designed to support interoperability with platforms like Google Vertex AI, Microsoft Copilot Studio, and AWS Bedrock AgentCore. However, the bidirectional Agent Gateway required for full, seamless interoperability is not yet generally available. This remains a critical bottleneck for enterprises hoping to move beyond siloed agent deployments.
For the humans involved, the integration of these tools changes the nature of daily work. Steve Hankins, CFO of Morgan Foods, noted that the company’s “superpower will come from people on top of agents,” emphasizing that the existing workforce remains the primary driver of supply chain operations. This perspective highlights a recurring theme in the agent economy: the goal is often to create slack in high-volume, repetitive tasks—such as the 50% reduction in trade classification effort reported for the International Trade Assistant—rather than immediate, wholesale replacement of roles. The human cost and the shifting leverage of these workers remain the primary variables to watch as these tools scale.
Despite the momentum, the industry still faces a significant measurement gap. Our maturity thesis indicated that while 93% of leaders believe AI will eventually run core business functions, only 7% have successfully operationalized these systems. A recent audit of 150 deployments across 119 organizations found that only 22 disclosed specific cost or financial impacts. SAP’s announcement, which includes a standard forward-looking statement disclaimer, reflects this broader trend: companies are moving toward production, but the hard data on long-term ROI remains scarce. As SAP expands its ecosystem through partnerships with Moody’s for risk monitoring and ORO Labs for procurement orchestration, the challenge for practitioners will be moving from these initial, vendor-reported successes to consistent, measurable operational performance.