Enterprises have spent years building AI strategies. Most of them are still waiting for the output. Dataiku is betting Cobuild changes that.
Enterprise AI has a gap that nobody talks about enough. Companies have the data, the strategy, and increasingly the ambition — but the distance between an idea and a production-ready AI workflow remains stubbornly wide. Code-generation tools move fast but produce outputs that governance teams cannot review. Standalone agent builders create prototypes that live outside enterprise infrastructure. The backlog grows. The business waits.
Dataiku‘s answer to that problem is Cobuild — an AI building agent that the company is making generally available on June 18, 2026. The pitch is direct: take a business objective, expressed in plain language, and turn it into a complete, governed, production-ready AI project — without writing a line of code and without stepping outside the enterprise controls organizations depend on.
Cobuild starts with a business problem and works outward from there. Using frontier AI models, it identifies the relevant data, designs the appropriate workflows, and generates the underlying components — data pipelines, machine learning models, agents, and applications. The output is rendered as a visual flow that every stakeholder, technical or otherwise, can inspect, edit, and approve before anything reaches production. Governance is not added at the end. It is built in from the start.
The enterprise context matters here. Pfizer is among the early adopters, and the framing from its team captures what Cobuild is trying to solve. “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 of Analytics Experience at Pfizer.
Cobuild operates entirely within Dataiku’s existing governance and permissioning frameworks. It can be powered through Dataiku AI Services or connected to a company’s own models via Dataiku’s LLM Mesh, with support for Snowflake Cortex AI, Databricks AI Gateway, AWS Bedrock, Google Gemini, Microsoft Foundry, OpenAI, Anthropic, and others. Model choice, data residency, and oversight requirements stay under enterprise control throughout.
Also Read: Karthik Ranganathan on Why AI’s Future Starts With Data Infrastructure
“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.”
The broader argument Dataiku is making is that the build-govern gap — the reason so many AI pilots never make it to production — is not a talent problem or a budget problem. It is an infrastructure problem. And Cobuild is designed to be the infrastructure that closes it.
Dataiku Cobuild is generally available to customers from June 18, 2026.
