Dataiku, the Platform for AI Success, has announced the launch of Agent Management, a standalone product that finds every AI agent an enterprise is running. It works regardless of which platform built the agent. It also measures business and technical performance and flags the agents that pose the greatest risk.

The company unveiled the product at Succeed, its annual flagship conference. The event showcases how companies worldwide are achieving AI success. The new product will be generally available in October.

The launch addresses a gap between how fast agents are created and how well organizations can observe and manage them. Most large companies already keep an account of their software. They know who owns it, what it costs and when it renews. Few can say the same about AI agents.

In fact, fewer than one in five organizations maintain a complete, current inventory of their AI systems, according to IBM’s “AI in Motion” research. The reason is that most agent platforms can only see the agents built on them. As a result, much of a company’s agent estate lacks clear ownership, purpose, risk or business outcome.

“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, co-founder and CEO of Dataiku. “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 product connects to platforms that enterprise teams already use to run agents. These include AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry. They also include Salesforce Agentforce, Snowflake Cortex and Dataiku itself. OpenTelemetry support covers custom environments. Agent Management scans all of these into a single inventory.

It also identifies each agent’s structure automatically. This includes the tools and models the agent relies on. Supervisors can therefore see how an agent actually works, not just that it exists.

For the highest-risk agents, the product keeps a standing record. These are the agents that handle customers, sensitive data or live transactions. The record covers certification status, named risks and tests that rerun on a schedule. Consequently, the evidence trail already exists when a manager, auditor or regulator asks for it.

Traditional agent monitoring capabilities sit within a single vendor’s stack. They are designed to favor that vendor’s own agents. The new product takes a different approach and is deliberately agnostic. It sits above the stack.

This position lets it answer questions that no single platform can. Which agents are unmonitored? Where is risk concentrated? Which agents earn their cost? Teams can ask these questions in plain language. They get an answer that covers the whole portfolio.

With general availability set for October, Agent Management gives enterprises a way to count, assess and oversee the AI agents they already run.

Related post

View all