Knowledge Graphs Emerge as Critical Infrastructure for Enterprise AI Agents
At GraphSummit 2026, industry leaders revealed how knowledge graphs are becoming essential for grounding AI agents in shared enterprise context, enabling governance at scale, and improving the safety and accuracy of autonomous systems.

Effective AI agents require more than access to raw enterprise data; they need the organizational context in which that data operates. Knowledge graphs are positioning themselves as a unified layer that can consolidate the business context currently distributed across multiple AI agents.
Today, developers manually embed business context into AI agents through prompts. This approach forces teams to recreate the same contextual work each time they deploy a new agent, rather than leveraging a shared knowledge foundation. According to Jesús Barrasa, field chief technology officer of gen AI at Neo4j Inc., a centralized knowledge layer solves this problem. "This knowledge layer, this idea of capturing your enterprise knowledge and using it as the context engine for your agents, not only gives you the consistency … it gives you the explainability," he said during an exclusive broadcast on theCUBE, SiliconANGLE Media's livestreaming studio. "That's the source of my data. These are the elements that I use to produce the answer."
During GraphSummit, theCUBE Research's John Furrier conducted interviews with technology leaders from Collibra BV, Blitzy Inc., Raymond James Financial Inc., Qiagen N.V. and other organizations. Their discussions centered on how knowledge graphs can anchor AI agents in verifiable enterprise knowledge, strengthen explainability, enable safer large-scale coding practices, accelerate pharmaceutical research and reshape how businesses operate.
https://www.youtube.com/embed/CJlkmav-r5k?feature=oembed
Runtime governance keeps autonomous agents aligned with business rules
Enterprise AI deployments face a critical governance challenge: how to ensure that increasingly autonomous agents behave appropriately. Decisions that previously required human judgment must now be made by agents in seconds, necessitating a shift toward governance mechanisms that operate at machine speed, according to Felix Van de Maele, chief executive officer and co-founder of Collibra BV.
Collibra has built several tools—Live Map, Maestro, Guardian Agents and Agent Contracts—on top of Neo4j's graph database infrastructure. These components work together as a control framework designed to ensure AI agents access appropriate information and comply with organizational policies.
"The challenge is: how do you govern at the speed of AI, which is different than the way we had to govern five years ago?" Van de Maele told theCUBE. "Maestro gives our customers the ability to build these governance steward agents that automate up to 80% of the governance work. You're able to govern quickly and efficiently enough to keep up with the demands of new AI workloads."
https://www.youtube.com/embed/FdtCk9IbVxQ?feature=oembed
Agent Contracts represent another layer of this governance approach, establishing and enforcing boundaries around what agents can do. By tracking agent activities and preventing actions that breach predetermined rules, they serve as a safeguard against unintended or unauthorized behavior.
Knowledge graphs enable AI coding agents to operate safely at scale
Minor modifications to software can trigger unexpected failures elsewhere in a system. For Blitzy Inc., which orchestrates thousands of AI coding agents simultaneously to help organizations execute large-scale infrastructure changes, a knowledge graph provides the mapping needed to understand dependencies and assess the risks inherent in any modification.
By giving agents a structured understanding of system relationships, knowledge graphs allow them to make changes confidently without being buried under millions of lines of code, according to Neeraj Deshmukh, director of engineering at Blitzy. The company supplements this with human-reviewed action plans and automated validation procedures to maintain quality throughout the coding process.
https://www.youtube.com/embed/j8P2nLXLldI?feature=oembed
"On top of that, we have agents who are watching other agents to make sure that they abide by the spec, the plan that we had created and approved by the human in the loop," Deshmukh told theCUBE. "That ensures that there is no drift or hallucination."
Financial services firms use knowledge graphs to validate AI accuracy and build confidence in agent-based operations
At Raymond James Financial Inc., knowledge graphs help direct AI agents toward data from authoritative sources that the organization has already validated. According to Naveen Mahankali, vice president of technology at Raymond James, this approach provides a foundation for verification.
"We have golden sources; we have systems that we can rely on today, and we're able to use agents to build and compare what we are building against the golden sources," he told theCUBE. "With the knowledge graph, even when we pull the data from different sources, it sits in the Neo4j, and we run evals compared to the golden sources and make sure they're 100% accurate before launching."
https://www.youtube.com/embed/8OgyswtEyho?feature=oembed
https://www.youtube.com/embed/videoseries?list=PLHBeVIUN0Be0
Raymond James is also investigating how AI agents can enhance the productivity of its financial advisors. The firm has already deployed Raimond, its proprietary AI assistant, which Jeff Griffith, vice president of AI engineering and data science at Raymond James, described as having achieved significant adoption. "We have an AI assistant … it's gotten huge adoption, and it's just going to get smarter," Griffith said during the interview alongside Mahankali. "That's part of the core evolution of what we're doing in terms of engineering. We're changing how people — how everyone — works with a desktop that gives people the best AI tools to move from doing all the menial tasks to managing agents that are doing those tasks. In terms of skills and what we are looking for, it is shifting just from people who can do tasks to people who have a drive to solve a meaningful problem."