Dataiku Launches Agent Management to Track AI Agents Across Vendor Platforms

The product targets a gap in how large companies account for what they operate. Most corporations keep records of software ownership and license renewal dates, but almost none can produce the same record for their agents. Fewer than one in five organizations maintain a complete and current inventory of their AI systems, according to IBM Corp. research cited by Dataiku. Florian Douetteau, co-founder and chief executive of Dataiku, said a bank can state how many servers it runs "to the decimal," but when asked about its AI agents the answer is "a shrug or a guess."

Once connected to the platforms where a company's agents run, the software pulls them into a single inventory. Connectors cover Salesforce Inc.'s Agentforce and the agent services operated by Amazon Web Services Inc., Microsoft Corp. and Google LLC. Agents built on Databricks Inc. or Snowflake Inc. appear as well, along with Dataiku's own, and custom environments can report in through the OpenTelemetry standard.

Each entry maps the tools and models an agent depends on so supervisors can see how it works. The riskiest agents carry a standing record of certification status and named risks, and tests on them rerun on a schedule, leaving an evidence trail in place when an auditor or regulator requests one.

Dataiku is positioning the product against monitoring tools built into a single vendor's stack, which the company argues tend to favor that vendor's own agents. Sitting above every vendor's stack, Agent Management is designed to answer portfolio-wide questions such as where risk is concentrated or which agents cost more than they return. Users can pose those questions in plain language.

Plain-language questioning also underpins the Cobuild expansion. Dataiku made Cobuild generally available in June as an agent that converts a business request into a governed Dataiku project without code, and the update carries that approach across the platform. A feature called Cobuild Insights lets any employee ask a question of governed company data and receive what Dataiku describes as analyst-quality answers, then pursue the same line of inquiry into deeper data work inside Dataiku.

A second addition, Dataiku Headless, brings Cobuild to developers inside the coding agents they already use, including Anthropic PBC's Claude Code, OpenAI Group PBC's Codex and Space Exploration Technologies Corp.'s Cursor. Developers can build or change pipelines, models and agents from within those tools. Dataiku said an old Microsoft Excel workbook or Alteryx Inc. workflow can be converted into a governed project, and the tool can also be pointed at a live environment to investigate problems. Cobuild is embedded in Agent Management as well, allowing a team to have it recommend business-value metrics for an agent or diagnose and fix a risk alert. Its AI Catalog is expanding to cover all of a company's vetted assets, semantic models included, so a described business need can be matched with approved assets that fit an answer or a visual pipeline with full lineage.

Clément Stenac, Dataiku's co-founder and chief technology officer, said enterprises tend to run technical and business teams on separate platforms "without shared data or shared guardrails." A few pilots could tolerate that arrangement, he said, but at scale it becomes "a liability companies have to prepare for."

Dataiku disclosed timing and pricing only for Agent Management, which will be generally available in October. Access carries an annual per-instance fee, with monitoring metered by the agent.

At its Succeed conference in New York, the company also published the "Global AI Confessions Report: CIO Edition, 2026," based on a survey of 685 chief information officers conducted by The Harris Poll for Dataiku from July 9 to 29 at businesses with more than $500 million in annual revenue across eight countries. Nine in 10 respondents said they were confident they had complete tracking of their agents, yet 81% said they lacked complete oversight of agents built outside approved systems or formal channels. Eighty-four percent said employees build agents and applications faster than the information technology organization can govern them. Most of the CIOs also believe their careers will be shaped by their success with AI, and 76% said their role would be at risk if their company failed to show measurable business gains from AI by the end of 2027. Their AI budgets face earlier scrutiny, with 72% expecting cuts or a freeze tied to this year's performance.