Dataiku Launches Agent Management to Oversee Enterprise AI Deployments
Dataiku's new Agent Management tool aims to provide visibility and control over the rapidly growing landscape of AI agents within organizations.
Dataiku has introduced Agent Management, a new standalone product designed to address the growing challenge of overseeing artificial intelligence agents within enterprise environments. This tool aims to discover, monitor, and assess the performance and risks associated with all AI agents, regardless of their origin platform. The move comes as organizations grapple with the rapid, often unmanaged, adoption of AI technologies.
Traditionally, enterprises maintain detailed inventories of their software assets, tracking ownership, costs, and renewal dates. However, this level of oversight is largely absent for AI agents. Research indicates that fewer than one in five organizations possess a comprehensive inventory of their AI systems. This gap exists because most agent platforms can only provide visibility into agents built on their specific systems, leaving a significant portion of an organization's AI estate undocumented and unmanaged.
"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," stated Florian Douetteau, CEO of Dataiku. He explained that teams have been developing agents at a pace that outstrips an organization's ability to count or manage them. Agent Management seeks to provide clarity on what AI agents are actually in use and their associated value.
The Agent Management solution integrates with a wide array of existing enterprise platforms where agents are deployed. This includes major cloud providers and AI development environments such as AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, Snowflake Cortex, and Dataiku itself. It also supports custom environments through OpenTelemetry. By connecting to these platforms, it compiles a unified inventory of all agents, automatically identifying their underlying structure, including the tools and models they utilize, thereby offering supervisors complete transparency into their operational mechanisms.
For agents handling sensitive data, customer interactions, or live transactions, Agent Management maintains a continuous record of their certification status, identified risks, and scheduled re-tests. This ensures that a clear audit trail is readily available for managers, auditors, and regulators, simplifying compliance and risk management processes.
Unlike conventional monitoring solutions that are often vendor-specific and biased towards their own agents, Dataiku's Agent Management is designed to be platform-agnostic. It operates at a higher level, enabling it to answer critical questions across the entire AI agent portfolio, such as identifying unmonitored agents, pinpointing areas of concentrated risk, and assessing which agents are delivering tangible value. Users can query this information in plain language, receiving comprehensive answers that cover their entire ecosystem.
The introduction of Agent Management by Dataiku reflects a broader industry trend towards better governance and security for AI deployments. As AI agents become more integrated into business operations, the need for robust management and monitoring tools is paramount to mitigate risks and ensure responsible AI usage. This product aims to equip organizations with the necessary visibility to navigate the complexities of modern AI landscapes effectively.