Collibra brings runtime governance to enterprise AI agents

Van de Maele said the stakes have risen as autonomous agents move from answering questions to taking action across enterprise systems. He said the difference from two or three years ago is that people used to be in the middle and their judgment could be counted on. “That has gone away, and so governance has moved from documentation, design time [and] policy setting to runtime,” he said.

Live Map is designed to build reusable context for agent fleets. Van de Maele said graphs are becoming an important context layer for enterprise agents, but he does not view them as a universal retrieval method. Depending on the task, he said, agents might use vector stores, structured files, semantic search or a graph grounded in business ontology. “There’s not going to be one mechanism that is able to answer every potential question,” he said. “The question is, ‘How do we route the right context to the right retrieval mechanism at the right time?’ Graph is key.”

Live Map prepares reusable context from curated unstructured documents so an agent does not have to reconstruct relationships with every query, according to SiliconANGLE. Collibra’s broader Knowledge Graph is based on Neo4j and supports traversal of relationships among enterprise assets. Van de Maele said organizations have many unstructured documents from SharePoint and do not want a model to parse through the data for every question. Building and engineering that context through Live Map, he said, means hitting it once offline so the answer is ready every time, which he called a massive token efficiency.

Collibra Maestro applies agents to governance tasks traditionally handled through manual stewardship. Van de Maele said the system is designed to help governance operations keep pace with expanding artificial intelligence workloads. “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?” he said. Maestro, he said, gives customers the ability to build governance steward agents that automate up to 80% of governance work, allowing them to govern quickly and efficiently enough to keep up with new AI workloads.

Agent Contracts define permitted behavior based on factors such as data sensitivity, operating context and risk, according to Van de Maele. Guardian Agents enforce those contracts before an action occurs. He said examples of agents escaping are well known, and Collibra’s approach is to encode and compile Agent Contracts and deploy them into existing AI gateways or AI orchestration tooling. Before an agent can take an action, the system monitors whether it aligns with the Agent Contract; if it does not, the action is blocked and the agent never has the opportunity to take it, he said.

The discussion took place during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio, at GraphSummit. SiliconANGLE reported that theCUBE is a paid media partner for GraphSummit, and that neither Neo4j, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.

Editor's Summary

Collibra is extending governance to runtime for autonomous AI agents through Live Map, Maestro, Guardian Agents and Agent Contracts. The tools are designed to supply reusable enterprise context, automate governance work and block actions that violate defined contracts. The move reflects pressure to control agent behavior as AI systems shift from answering questions to taking action.