Cobuild’s new capabilities help teams get trusted answers, build and reuse AI assets, and oversee agents from one platform
Dataiku, the Platform for AI Success, announced the expansion of Dataiku Cobuild, its natural-language agentic interface that gives business teams and AI builders a direct path from question to governed AI work.
Initially launched in June, Cobuild’s expanded capabilities were first shown to attendees at Succeed, the company’s annual flagship conference, showcasing how to turn business ideas into a governed Dataiku project without writing code. Cobuild’s expansion extends that same natural-language, governed experience across the platform: for business users who need trusted answers from their data, for developers working with the coding tools they already use, and for IT teams that keep their growing agent portfolios accountable.
Also Read: LTM Launches BlueVerse SovereignSphere Models to Redefining Enterprise Context and Governance
With this expansion comes new capabilities to make that possible:
- Cobuild Insights allows both technical and non-technical users to ask a question in plain language and get analyst-quality answers grounded in trusted, governed data, then carry that same thread into deeper data work in Dataiku when needed.
- Dataiku Headless is a new way to use Cobuild to harness the full power of the Dataiku platform via coding tools such as Claude Code, OpenAI Codex, and Cursor. Developers can build and modify data pipelines, models, dashboards, and agents without leaving their coding agent; migrate legacy work, such as an Excel workbook or an Alteryx workflow, into an inspectable, governed Dataiku project; and investigate issues in a live Dataiku environment. Now, teams can turn work that used to live in one person’s terminal into something the rest of the business can review and run.
- Agent Management now includes Cobuild, giving teams a way to extract insights, make decisions, and act across their entire agentic portfolio, from recommending business-value metrics for an agent to diagnosing and remediating a risk alert.
Along with these new features, Cobuild is expanding its universal AI Catalog, which now includes all of a company’s vetted assets: datasets, models, agents, and semantic models. This lets teams describe a business need, have Cobuild find the vetted assets that fit, and then assemble them into an answer or a visual, inspectable pipeline with full lineage, enabling teams to leverage and build on their company’s existing AI work rather than rebuild it.
“Enterprises run their technical teams on one platform and their business teams on another, without shared data or shared guardrails. That gap was manageable with a few pilots. At scale, it’s a liability companies have to prepare for. Cobuild closes it: one platform, one set of rules,” said Clément Stenac, co-founder and CTO of Dataiku.
Source: Businesswire