Dataiku has launched Agent Management, a standalone product designed to give enterprises a consolidated view of AI agents operating across multiple platforms, addressing a growing governance gap as agent deployment accelerates.
The product discovers AI agents regardless of where they were built, measures their business and technical performance, and identifies those presenting the greatest risk. Dataiku says fewer than one in five organizations currently maintain a complete, up-to-date inventory of their AI systems, highlighting the difficulty enterprises face in establishing ownership, purpose, risk and business value across increasingly fragmented agent estates.
Agent Management connects with AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, Snowflake Cortex and Dataiku, while OpenTelemetry support extends visibility to custom environments.
“Ask a bank how many servers it runs, and you get an answer to the decimal. Ask how many AI agents it’s running, and you get a shrug or a guess. Nobody set out to build it this way. Teams built agents faster than anyone could count them. Agent Management tells you what’s actually out there, and what it’s actually worth.”
— Florian Douetteau, Co-founder and CEO, Dataiku
Beyond simply cataloguing agents, the platform identifies their underlying models and tools, providing visibility into how individual agents operate. For higher-risk agents handling customers, sensitive data or live transactions, it maintains certification status, identified risks and scheduled testing records, creating an evidence trail for management, audit and regulatory requirements.
Dataiku is positioning the product above individual agent platforms rather than within a single vendor ecosystem. This allows organizations to identify unmonitored agents, locate concentrations of risk and assess which agents are generating sufficient business value to justify their cost.
The move reflects a broader challenge in enterprise AI: deployment is progressing faster than governance. As agents increasingly move from experimentation into business processes and transactions, organizations need visibility across the entire agent estate rather than isolated monitoring within individual technology stacks.
Agent Management is scheduled for general availability in October 2026, with annual instance-based pricing and monitoring metered per agent.
