GenAI & Agents Product Updates

Discover Dataiku Cobuild: the queue stops here

Jun 11, 2026

2 min read / Julia Tran

From Shadow AI to Shadow Agents
The Hidden Cost of Autonomy
Why Traditional Governance Breaks Down
Visibility Is the Missing Primitive
Human-in-the-Loop as a Design Choice
The Real Shift: From Limiting AI to Seeing It

You've been in this meeting before.

Someone from the business side walks in with an idea: an analyst, a sales ops lead, a finance manager who knows the data better than anyone. A model they want to run. A workflow they've been thinking about for months. They've done the homework. They have the numbers, the business case, and a slide deck all buttoned up.

The meeting goes well. Everyone agrees it's worth building. And then it goes into a queue.

Maybe it comes back in Q3. Maybe it gets deprioritized when something more urgent lands. Maybe the analyst, tired of waiting, builds something in a consumer AI tool that nobody in IT knows exists: fast, useful, completely ungoverned, and now quietly running inside a production process.

This is the state of enterprise AI right now. Demand has never been higher. Model capabilities have never been greater. But data and AI teams can't keep up, so business teams are trading control for speed. They're filling the gap with workarounds that are unsanctioned and disconnected from enterprise systems, and IT can't govern effectively either.

It's a mess. Dataiku Cobuild is built on the belief that it doesn't have to be. Starting June 18, we're officially launching Cobuild, so you won't have to choose between fast and governed anymore.

The fast path and the governed path are the same

Consumer AI tools like ChatGPT have reshaped what people expect from software. The idea-to-output loop that used to take weeks now takes minutes. That expectation isn't going away, and it shouldn't. The problem is that the speed came without the structure enterprises actually need: opaque outputs, no observability, pipelines nobody can inspect before they ship.

Cobuild gives teams the speed without the tradeoff.

Just describe what you want to build in plain language. Cobuild assembles it across data preparation, model training, agentic workflows, and production-ready applications as a complete, visual Dataiku Flow. Every step is inspectable. Every transformation is editable. The analyst who had the idea in that meeting can see exactly what got built, change what doesn't look right, and sign off before anything moves to production.

Not a prototype handed off to an engineer to audit. Not a workaround living outside the stack. A governed, production-ready AI project built by the person closest to the problem, inside the environment IT already trusts.

For the people who have to approve

Scale without visibility isn't a solution. The people responsible for what's in production need to see everything being built.

Access to Cobuild is controlled at the admin level. All Cobuild interactions produce a standard Dataiku flow, inheriting the same review and approval controls already in place.

Don't give everyone the keys and hope for the best. Make the fast path and the governed path the same path, so the analyst doesn't have to choose between moving quickly and staying inside the rails.

What actually changes

With Cobuild, the queue gets shorter. The ideas that have been waiting, always good enough to pursue but never urgent enough to prioritize, finally have somewhere to go.

The data science team doesn't disappear from this picture. Work that requires technical depth still goes to them, just with a much shorter line. And the workarounds stop being necessary. Not because IT banned them, but because there's finally something better that works for everyone.