Enterprise AI agents are gaining attention as companies move beyond simple chatbots toward software that can execute business processes. At the VB Transform 2026 conference, SAP senior solution advisor Max McPhee explained why the technology still needs a solid foundation built on knowledge graphs and strong governance.
Context is the missing piece
McPhee told VentureBeat Research analyst Rob Stretchay that the key difference between a chatbot and a true agent lies in grounding the system in a company’s own data. “Where we’re starting to see more emergent behavior of it feeling like a coworker rather than an assistant, is where we’re able to provide context on the actual enterprise rather than being able to use more of the standard knowledge,” he said. That context separates most enterprise chat software from genuinely agentic systems.
To give agents the needed context, SAP relies on knowledge graphs combined with vector‑embedded data. “When you are onboarding a new agent, I think it’s important to acknowledge how you might onboard a new employee, but tune that for an agent,” the advisor said. Knowledge graphs present information in a format that is easy for an AI to retrieve, helping the system avoid stumbling over internal shorthand or acronyms common in SAP environments. The result, according to the executive, is a smoother interaction than a typical chatbot that might ask, “What does that acronym mean?”
Related: Kimi K3 Weights Now Fully Disclosed
Governance, identity and security
SAP’s long history with process control gives it an advantage when it comes to governing autonomous agents. “That’s where SAP really has a good home, around that governance and process control,” McPhee noted, adding that the company’s 50‑year experience is being modernized to handle the flexibility of AI‑driven actions. One practical outcome is the renewed role of machine learning for validating agent behavior. Customers are now layering anomaly detection and machine‑learning‑based validation as guardrails, echoing SAP’s earlier use of intelligent approval recommendations.
Identity and permissions are also central to the governance model. Under SAP’s approach, both the human user and the AI assistant—named Joule—must have rights to access a given system. Even if a user can access SAP S/4HANA, they cannot do so through Joule unless the assistant itself has been provisioned for that access, reducing the risk of agents bypassing established controls.
Integrating diverse environments
Much of the work focuses on reconciling SAP’s own knowledge with the varied, often non‑SAP, settings of its customers. He cited recent acquisitions such as LeanIX, which he likened to “Google Maps for your architecture,” and the process‑mining firm Signavio. These tools help map the broader IT environment so that agents can understand how different systems interconnect. SAP has also invested in Berlin‑based automation company n8n, embedding it into Joule Studio, a low‑code environment for building intent‑based agents.
Companies that rely heavily on legacy, on‑premises systems may encounter performance bottlenecks as they expand autonomous agent use. “You’re going to probably run into throughput issues, and you’re kind of trying to drive a Ferrari around a dirt track,” the advisor warned. “You’ve got to upgrade the track first if you want to drive a Ferrari.” The comparison highlights the need for infrastructure upgrades before agents can deliver their full potential.
Related: New camera does it all on the go
The push toward autonomous agents reflects broader industry trends where firms seek to reduce manual effort and accelerate decision‑making. By anchoring AI in knowledge graphs, enterprises can ensure that agents draw from accurate, company‑specific data rather than generic internet sources. Governance frameworks keep those agents from overstepping security boundaries, a concern that becomes more salient as AI capabilities expand.
In practice, the combination of contextual grounding and robust controls can turn an AI tool from a simple question‑answering bot into a functional participant in business workflows. This shift could free up staff for higher‑value tasks while maintaining compliance with internal policies.
Overall, SAP’s strategy shows that without the right data structures and governance mechanisms, autonomous agents risk becoming unreliable or insecure. By investing in knowledge graphs, machine‑learning validation, and permission‑based execution, the company aims to make enterprise AI agents both useful and trustworthy.
