Dataiku announced on September 24, 2026, at its annual Dataiku Succeed conference, the launch of Agent Management: a standalone product that discovers, measures, and audits the performance of every AI agent a company has in production, regardless of which platform built it. It will reach general availability in October 2026.
According to figures from IBM's "AI in Motion" research, cited in the announcement, fewer than one in five organizations keeps a complete, up-to-date inventory of its AI systems. The pace at which teams are creating and adopting agents — in sales, support, finance, engineering — has far outrun IT and security teams' ability to know how many agents exist, who built them, what access they have, and whether they're actually generating value.
The product discovers agents deployed across the most common enterprise platforms — AWS, Databricks, Google Cloud, Microsoft, Salesforce, Snowflake, among others — regardless of whether they were built with those vendors' native tools or with third-party frameworks. From that inventory, it measures each agent's technical and business performance, and explicitly flags which ones represent the greatest risk, whether because of the access they hold, a lack of monitoring, or observed behavior.
Dataiku's bet is explicitly "cross-platform": it isn't competing to be the platform where agents get built, but to be the governance layer that sits on top of every platform a company already has. It's a signal of where the enterprise AI tooling market is heading: from "how do I build more agents faster" to "how do I know how many agents I have and whether they're out of control."
If your organization already has agents in production on more than one platform — increasingly common even at mid-sized companies — the question Dataiku is betting you'll ask yourself is simple: can you name, right now, every agent your company has running, what access it has, and who approved it? If the answer is no, it doesn't have to be this specific product, but it is a sign that agent governance needs to be treated as a core IT function, with the same rigor as employee identity management, rather than a detail each team figures out on its own.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies in Guatemala and Latin America. Contact him for a consultation.
Contact Carlos Montiel