Presented by SAP

At VB Transform 2026, Max McPhee, senior solution advisor at SAP, discussed with Rob Stretchay, lead analyst at VentureBeat Research, what enterprises need in order to move beyond chatbots and toward autonomous AI agents capable of carrying out real business processes. McPhee said the key distinction is whether those agents are grounded in a company’s own operational context, rather than relying only on general knowledge.

https://www.youtube.com/watch?v=SRf9t-wSZSo

"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," McPhee said.

That context, he argued, is what continues to separate most enterprise chat tools from systems that can operate in a genuinely agentic way. The distinction matters because enterprise agents are expected not only to answer questions, but also to interact with business applications, retrieve company-specific data and support workflows where accuracy, access control and auditability are central requirements.

Building enterprise context with knowledge graphs

McPhee said enterprises should think about onboarding AI agents in a way that resembles onboarding employees, while accounting for the fact that software retrieves and uses information differently than people do.

"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," McPhee said. "The way that is really powerful is using knowledge graphs and having vector-embedded data, because that's a really easy format for an agent to be able to find and retrieve information."

Knowledge graphs are commonly used to represent relationships among entities, processes and data sources, while vector embeddings help AI systems retrieve semantically related information. In an enterprise setting, McPhee said combining those approaches can help agents connect general language understanding with the specific structure of a company’s operations.

That grounding can also help agents understand internal terminology and company-specific shorthand, an issue McPhee described as especially important in SAP environments.

"Being able to provide that tribal knowledge in the format that's easy for it to consume helps to provide a really nice result with your agents versus a chatbot that might say, 'Well, what does that acronym mean?'" he said.

Governance, identity and security for autonomous agents

McPhee said governance is an area where SAP’s history as an enterprise process company is relevant, especially as controls evolve for systems that can act with more flexibility than previous generations of automation.

"That's where SAP really has a good home, around that governance and process control," McPhee said. "We're a 50-year-old process company, modernizing that governance to be able to handle the flexibility that comes with agents running."

He also pointed to a renewed role for machine learning in validating agent behavior.

"It's becoming a bit of a revival of machine learning," McPhee added, referring to customers that run agents inside a process and then add anomaly detection and machine-learning-based validation as a guardrail. He said this is the same type of approach SAP has long used for intelligent approval recommendations.

Identity and permissions extend that governance into execution. In this model, both the human user and SAP’s Joule, the generative AI assistant embedded across SAP’s cloud applications and Business Technology Platform, must have the rights required to access a particular system. Even if a user is authorized to access S/4, they cannot use Joule to do so unless Joule has also been provisioned for that access. McPhee said that structure is intended to prevent an agent from being used to bypass access controls.

Standard SAP and customized enterprise landscapes

A major part of McPhee’s work involves aligning SAP’s own knowledge with decades of customer customization and with systems outside SAP. He said many customers tell SAP, "You’re only 10% of my landscape," a reality that has influenced the company’s recent strategy.

Recent acquisitions, including LeanIX, which McPhee compared to "Google Maps for your architecture," and process-mining company Signavio are meant to help map that non-SAP majority so SAP’s agents can understand how enterprise systems are connected. SAP has also invested in Berlin-based automation company n8n and is embedding it natively into Joule Studio, the company’s intent-based, low-code environment for building agents.

That emphasis reflects a broader implementation challenge for large companies: AI agents often need to work across standardized software, customized business logic and third-party systems that were not originally designed around autonomous AI workflows. Mapping those dependencies is a prerequisite for giving agents reliable context about where data comes from and which systems are involved in a process.

McPhee said companies may also need to modernize older on-premises systems as they broaden their use of autonomous agents, or they could face technical constraints.

"You're going to probably run into throughput issues, and you're kind of trying to drive a Ferrari around a dirt track," he said. "You've got to upgrade the track first if you want to drive a Ferrari."

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they are clearly marked as such.