Agentic AI in Manufacturing ERP: The Gap Between Claims and Live Capabilities

Hiroshi Ozaki connects SAP’s operating signals to executive decisions

After 35 years of watching enterprise software evolve—from the green screens of Fujitsu’s early systems to SAP R/3’s client-server revolution, and later the in-memory promise of HANA—I’ve learned that the distance between a vendor’s marketing narrative and a factory manager’s real shift report can be measured in years. Today, that gap is nowhere sharper than in the wave of agentic AI claims pouring into the manufacturing ERP space. The headline is that AI agents will autonomously adjust production schedules, reroute supply chains, and execute quality inspections. But for executives and transformation leaders trying to separate investment signal from noise, the operational reality is far more modest than the stage demonstrations suggest.

This is not a criticism of ambition. It is a recognition that, as with previous technology waves, the market’s vocabulary outruns the systems that must deliver. The recent analysis by ERP.Today, “Agentic AI claims vs. delivered functionality in manufacturing ERP systems,” provides a structured evidence base that I believe deserves C-suite attention—not just from practitioners, but from anyone sizing up SAP’s commercial trajectory.

The Business Signal

Vendors, SAP included, are increasingly relying on AI narrative premium to support cloud transition momentum and pricing power. SAP’s cloud backlog growth, now frequently cited in quarterly earnings, depends in part on customers believing that AI-embedded processes will soon be essential—and that waiting means falling behind. The company has woven “Business AI” into its messaging around Joule and the S/4HANA platform, with agentic capabilities featured prominently at events like Sapphire.

Yet measured against the yardstick of live, generally available functionality in manufacturing, the evidence gathered by ERP.Today shows that most agentic AI claims are still in an adjective phase—used to describe future product intent, not to independently execute complex, unsupervised tasks on a shop floor. The report’s systematic scorecard distinguishes between capabilities that are demonstrably operational and those that exist only in roadmaps, pilots, or tightly controlled demo environments. The result is a clear pattern: the industry’s talk of autonomous production agents is running notably ahead of the code that manufacturing IT teams can actually deploy today.

For investors and analysts, the signal is that SAP’s AI monetization in the manufacturing sector is still in its infancy. While cloud revenue growth benefits from the halo effect, the margin expansion from premium AI add-ons will depend on shipping robust, auditable functionality—not just large language model integrations into planning screens. Until agentic features clear the bar of factory-floor reliability, the revenue impact remains aspirational.

What It Means for SAP Customers

For the manufacturing executive, the ERP.Today analysis is a timely strategic planning tool. I recommend using the five-point evaluation framework it suggests to strip away vendor adjectives and focus on evidence of operational readiness. My own experience in automotive and electronics implementations leads me to propose a parallel, manufacturing-specific readiness lens:

1. Data Integrity – Can the AI agent rely on real-time master data, inventory, and machine status without manual correction? In most

References

• AI Reality Check: Separating Agentic Claims from Manufacturing Substance
• AI Reality Check: Separating Agentic Claims from Manufacturing Substance
• SAP AI Core Documentation

References

• SAP AI Core Documentation
• AI Reality Check: Separating Agentic Claims from Manufacturing Substance
• AI Reality Check: Separating Agentic Claims from Manufacturing Substance