Dataiku Agent Management Maps Enterprise AI Agent Sprawl

Enterprise AI agents are multiplying faster than anyone can count them. That is the bet behind Dataiku's newest product: a standalone layer that finds every agent running inside a company, measures what each one is worth, and flags the agents carrying the most risk. The company unveiled Agent Management on September 24, 2026, at its Succeed conference in New York City, with general availability scheduled for October.

Agent sprawl is the problem the launch names. According to IBM's "AI in Motion" research cited by Dataiku, fewer than one in five organizations maintain a complete, current inventory of their AI systems. Most agent platforms can only observe the agents built on top of them, which leaves the rest of a company's agent estate without clear ownership, purpose, or risk profile.

One Inventory Across Six Competing Platforms

Agent Management connects to AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, and Snowflake Cortex, alongside Dataiku's own platform. Custom environments can report in through OpenTelemetry support, so home-grown automations show up in the same view as vendor-built assistants.

Once connected, the system scans every connected environment into a single inventory. Dataiku says each record captures the agent's owner, its purpose, the systems it touches, and its last run. The tool also maps the structure beneath each agent, including the models and tools it depends on, so a change in a foundation model or API surfaces everywhere it matters.

This is the layer most companies are missing today. Every large company already keeps an account of the software it runs: who owns it, what it costs, and when it renews. Almost none can say the same about their AI agents. Dataiku positions agent management as the missing ledger for the agent economy, a control layer that sits above the platforms where agents are built and operated.

From Discovery to Drift Alerts

The product is designed as more than a directory. Dataiku says Agent Management monitors each agent's health, usage, cost, quality, and behavior over time, and raises alerts when an agent moves away from its normal pattern. A coverage roundup from the AI agent community noted that the tool tiers agents by risk, giving operators a triage order instead of a flat list.

For agents that touch customers, sensitive data, or live transactions, the product keeps standing records of certification status, named risks, and scheduled tests. The point is that an audit trail already exists when a manager or regulator asks for one, rather than being reconstructed after the fact.

Among the risks the company names are excessive privileges, shared credentials, and the absence of a human override. Independent reporting on the launch confirmed the connector set spans the major cloud and data platforms, and noted that pricing is per-instance annually with monitoring metered per agent; Dataiku has not published a specific dollar figure.

"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," said Florian Douetteau, Dataiku's co-founder and CEO, in the launch announcement. "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."

The timing is telling. The launch arrives the same week that NVIDIA rolled out an open agent safety platform and analysts are publicly asking who actually governs production agents — a wave of infrastructure news that suggests the industry is moving from asking how to build agents to asking how to run them safely at scale. For readers following our AI News coverage of the agent build-out, from frontier lab funding to Anthropic's leaked IPO figures, the Dataiku story is the enterprise IT counterpart: less glamorous than model launches, more important than most.

The practical takeaway for teams running agents is straightforward. Map your current footprint first — owners, data access, service levels — then validate discovery and drift detection in a short pilot before scaling, whether with Dataiku's tool or an equivalent. In the agent economy, the companies that can't count their agents won't be able to govern them either.

Sources

• BigDATAwire — Dataiku press announcement (New York, Sept. 24, 2026)
• Pondero — "Dataiku launches Agent Management to inventory AI agents across six enterprise platforms"
• Aree Blog — "Dataiku Launches AI Agent Inventory for Enterprise Tracking"
• AI Agent Store daily news — September 25, 2026 roundup
• Lapaas Voice — "Dataiku Agent Management Maps AI Sprawl"