Collibra CEO: AI agent scale-up hinges on inventory, contracts and context

Felix Van de Maele, founder and CEO of data governance company Collibra, said at VentureBeat's AI Week in New York that enterprises running AI agents at scale need inventory, contracts and maintained context to keep risk and cost under control.

Van de Maele warned that the risk, cost and oversight burden grows exponentially with each additional agent, citing data that more than 75% of enterprises hit a production wall when scaling from a single assistant to 10-15 autonomous agents. A Harris Poll commissioned by Collibra found that nearly 30% of companies caught agents accessing or surfacing data they were not permitted to see.

He recommended building an accurate inventory of agents, establishing a deployment lifecycle that defines what an agent must clear before production, and maintaining a context layer — an ontology describing the business — that is independent of any model or platform. He also described agent contracts as machine-readable governance rules enforced deterministically by guardians at runtime.

Van de Maele argued that model improvements alone will not fix enterprise agent failures, since business queries often have exactly one correct answer, and stressed that differentiation comes from capturing proprietary knowledge and making it available to agents efficiently.