Runtime governance is becoming critical as autonomous agents move from answering questions to taking action across enterprise systems. Organizations need to govern both the business context agents receive and the boundaries they operate within.
Collibra BV is addressing those requirements through Live Map, Maestro, Guardian Agents and Agent Contracts. Together, these runtime governance capabilities are intended to reduce manual verification, rework and risk. Those costs constitute the “hallucination tax,” according to Felix Van de Maele (pictured), chief executive officer and co-founder of Collibra.
“I think the stakes just have gotten so much higher,” he said. “The big difference today versus two or three years ago is that we used to have people in the middle, and we could count on the judgment of people. That has gone away, and so governance has moved from documentation, design time [and] policy setting to runtime.”
Van de Maele spoke with theCUBE Research’s John Furrier at GraphSummit, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how a shared knowledge layer could help agents use enterprise context, explain their answers and reuse knowledge across tasks. (* Disclosure below.)
Live Map builds reusable context for agent fleets
Graphs are becoming an important context layer for enterprise agents, but Van de Maele doesn’t view them as a universal retrieval method. Depending on the task, 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 doesn’t have to reconstruct relationships with every query. Collibra’s broader Knowledge Graph is based on Neo4j and supports traversal of relationships among enterprise assets.
“You have a lot of unstructured documents from your SharePoint. What you don’t want is for the model to [parse through the data] for every single question,” Van de Maele said. “Building, engineering that context and making that available through Live Map, you hit it once offline, and so you have the answer ready every time. That’s a massive token efficiency.”
Runtime governance turns policies into enforcement
Collibra Maestro applies agents to governance tasks traditionally handled through manual stewardship. 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?” Van de Maele said. “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.”
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.
“We’ve all read the examples where the agent escapes,” Van de Maele said. “The way we’re solving that is by encoding and compiling these Agent Contracts and deploying those into your existing AI gateways or AI orchestration tooling. Before the agent is able to take the action, we monitor: Does it align to the Agent Contract? If not, we block it, and the agent never has the opportunity to take that action.”
Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the GraphSummit:
(* Disclosure: TheCUBE is a paid media partner for GraphSummit. Neither Neo4j, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)
