Dataiku launches Cobuild to speed governed AI projects

Dataiku has launched Dataiku Cobuild, a product designed to help enterprise teams create governed AI projects from business objectives.

Cobuild is generally available and targets organisations trying to move AI work from experimentation into operational use without relying on code-generation tools alone.

Dataiku is positioning the launch around a persistent problem in large companies: many have built data platforms and adopted AI tools, but still struggle to turn prototypes into systems that can be reviewed, approved and deployed under internal controls.

Cobuild starts with a business problem and turns it into a Dataiku project. It uses AI models to identify relevant data, design workflows and generate components such as data pipelines, machine learning models, agents and applications.

The output is presented as a visual flow so different stakeholders can inspect, edit and approve a project before it goes into production. The product works within Dataiku's existing governance and permission frameworks, allowing wider use across business teams while keeping oversight with risk and IT functions.

Governance focus

The launch comes as companies adopt AI-assisted software development tools more widely but face questions over how those outputs can be checked and managed in regulated or tightly controlled environments. Dataiku argues that business and governance teams often lack the skills or tools to review code-heavy outputs, while standalone agent-building tools can produce work outside established enterprise systems.

The issue is particularly acute in industries with stricter operational and compliance demands. Pfizer was among the users Dataiku cited as highlighting the need for stronger controls around AI-generated work.

"AI-assisted building compresses the distance between an idea and a production-ready workflow. But in an enterprise and especially in pharma, the output has to be more than impressive. It has to be explainable, auditable, and safe to put into production. That's the gap Dataiku Cobuild closes," said Neil Patel, Senior Director, Analytics Experience, Pfizer.

Customers can use Cobuild through Dataiku's own AI services or connect external models through its LLM Mesh layer. Supported options include Snowflake Cortex AI, Databricks AI Gateway, AWS Bedrock, Google Gemini, Microsoft Foundry, OpenAI and Anthropic.

That model choice matters for large organisations that want to retain control over data residency, oversight and vendor selection as they expand AI use. By keeping model access inside an existing governed framework, Dataiku is seeking to appeal to companies that want flexibility without creating separate development tracks outside core systems.

Enterprise gap

The wider commercial opportunity lies in the gap between AI experimentation and deployment. Many companies have invested in data foundations and AI strategies, but turning a business request into a production-ready system remains slow, especially when review and sign-off involve several teams.

Dataiku's argument is that plain-language inputs could reduce that friction if the resulting project remains visible and reviewable to business users, governance staff and IT teams. Rather than centring the process on specialist developers, Cobuild is aimed at broader enterprise teams that need to take part in building and approving AI projects.

Clément Stenac, one of Dataiku's co-founders, said that approach was central to the product's design.

"AI-assisted development only matters if the output can survive contact with the enterprise," said Clément Stenac, Co-Founder and CTO of Dataiku.

"That means it has to be understandable to the people closest to the business, governable by the teams responsible for risk, and production-ready for the IT teams that run it. Cobuild was built to that standard: AI brings the speed, while enterprise teams bring the business ingenuity, and IT keeps the control," Stenac said.

Dataiku sells software for building, deploying and governing AI and analytics work across enterprise systems. Its platform sits above data platforms, cloud infrastructure, business applications and AI services, with a focus on centralised governance across multi-vendor environments.

The Cobuild launch extends that pitch by targeting a common complaint in enterprise AI programmes: new tools can generate output quickly, but not always in a form that meets internal standards for explainability, auditability and production use.