Dataiku launches Cobuild on Snowflake for AI workflows

Dataiku has launched Cobuild on Snowflake for joint customers running AI workflows on Snowflake.

The product links Snowflake Cortex AI with Dataiku's orchestration software, allowing users to turn natural-language prompts into visual AI workflows, agents and applications inside Snowflake environments. It is designed to give customers more control over how those workflows are reviewed, validated and approved before deployment.

Both companies are targeting large organisations that want broader use of AI tools without handing software generation entirely to coding assistants. They already have more than 350 joint customers globally.

The launch comes as businesses try to balance the speed promised by generative AI tools with tighter demands around compliance, auditability and cost management. In many corporate settings, leaders have been wary of systems that produce code or processes that are difficult for non-specialists, risk teams or IT departments to inspect.

Cobuild on Snowflake turns a business objective into a visual workflow rather than a block of generated code. The workflow can cover data preparation, machine learning tasks, AI agents and business applications, with teams able to review and modify each step before release.

Governance focus

Dataiku is positioning the product as an alternative to consumer-style AI coding tools in regulated or high-stakes settings. It argues that organisations need a way for business users, data teams and governance staff to work on the same project in a controlled environment.

"Consumer AI tools can make code appear instantly, but enterprises cannot afford to unleash opaque, unvalidated workflows into environments where accuracy, compliance, safety, and cost control matter," said Florian Douetteau, Co-Founder and Chief Executive Officer of Dataiku.

"With Cobuild on Snowflake, organisations can bring the speed of AI-assisted development into a governed process with visual workflows that teams can inspect, validate, and improve before production," he said.

Snowflake said the product builds on Cortex AI, its layer for accessing large language models on data held within the Snowflake platform. That allows customers to work with AI models close to their business data rather than moving workloads into separate tools.

For Snowflake, the launch adds another software partner to its effort to make its data cloud a place where customers not only store and analyse data but also build AI systems. For Dataiku, it deepens a long-running technical integration with Snowflake by adding a prompt-based workflow creation tool for both technical and non-technical staff.

"Snowflake Cortex AI brings leading models directly to governed enterprise data, so organisations can build and run AI where their business context already lives," said Baris Gultekin, Vice President of AI at Snowflake.

"With Dataiku's Cobuild, customers can accelerate how business intent is translated into visual, governed workflows that can be inspected, optimised, and operationalised at scale. That combination matters as enterprises move AI into production and need accurate systems that they can trust," he said.

Existing tie-up

The product builds on an existing partnership between the two companies. Dataiku said its software already works with Snowflake for in-warehouse data preparation, machine learning training through Snowpark Container Services, batch and real-time inference, Iceberg catalog support and several Cortex services, including Cortex Agents, Cortex Analyst, Cortex Search, Cortex REST API and Cortex AI Functions.

That breadth matters because many customers want AI development tools to fit into existing data estates rather than create a separate environment. By keeping generation and workflow creation within a customer's Snowflake setup through a secure REST API integration, the companies are seeking to address concerns about data movement and oversight.

The initial rollout is for joint customers already building with Dataiku on Snowflake, with wider access planned later for Snowflake customers that want to add Dataiku to their stack.

Users can ask the system in plain language to prepare data for analysis, improve a model or create an AI agent for a business process, and Cobuild will generate a full Dataiku workflow for teams to inspect and refine before it reaches production.