Snowflake adds governed AI workflows
- Dataiku launched Cobuild on Snowflake on May 14, and Valid Systems rolled out a Snowflake integration on May 18 for production AI workflows. - Dataiku said hundreds of joint customers can use Cobuild, while Valid said its platform processes more than 70 million transactions monthly. - Snowflake’s Cortex AI and agent features remain available in Snowflake documentation, where customers can review native workflow and governance capabilities.
Dataiku and Valid Systems released separate Snowflake integrations in the past week that push more AI work inside Snowflake’s own data environment. Dataiku said on May 14 that Cobuild on Snowflake lets joint customers turn natural-language prompts into governed AI agents and workflows on Snowflake. Valid Systems said on May 14 that its Snowflake AI Data Cloud integration lets financial institutions run fraud and risk models directly in Snowflake for real-time decisioning. Together, the launches show software vendors packaging AI workflow creation and execution inside the data platform rather than exporting data to separate tools. ### Why are two different vendors building deeper into Snowflake now? Dataiku said its new product pairs Snowflake Cortex AI with Dataiku’s orchestration layer so customers can move from a plain-language request to a visual workflow that stays natively on Snowflake. The company said the product is aimed at Global 2000 customers that want governance, transparency and cost controls as they move AI projects into production. Valid Systems said its integration is designed for deposit fraud mitigation and account-risk decisions, with machine learning models running directly in the Snowflake cloud environment. The company said the setup is meant to give banks and other financial institutions instant decisioning without the conventional middleware and infrastructure work that often slows deployment. (dataiku.com) ### What does Dataiku’s launch actually add for Snowflake customers? On May 14, Dataiku described Cobuild on Snowflake as a way to translate “natural-language business intent” into governed AI agents and workflows. The company said the integration uses Snowflake Cortex AI’s access to large language models while keeping Dataiku’s controls around orchestration and enterprise deployment. Snowflake’s own documentation says Cortex AI functions are exposed as SQL and Python features inside Snowflake, and that Snowflake also offers Cortex Agents for multi-step task planning and execution. (businesswire.com) Snowflake says those features are delivered within its service perimeter and support document pipelines, retrieval-augmented generation and other AI workflows. ### Why does the Valid Systems deal matter beyond one fraud product? (dataiku.com) Valid Systems said the Snowflake integration is intended for financial institutions “of all sizes,” and outside coverage said the offering is aimed in part at smaller banks that already operate on Snowflake. The company said Snowflake provides the scale needed to process high transaction volumes in real time, while letting institutions access Valid’s models inside an existing cloud environment. (docs.snowflake.com) TechIntelPro, citing the company announcement, said Valid processes more than 70 million transactions a month and guarantees more than $6 billion in immediately available funds. That gives a measure of the production workloads the company is using to market the Snowflake integration. ### Where is Snowflake itself in this stack? (businesswire.com) Snowflake documentation says Cortex AI functions can be called directly in SQL, while Cortex Code includes built-in skills for agent creation, machine learning, data engineering and governance workflows. Snowflake has been positioning those features as native ways to build AI applications without moving data out of the platform. (techintelpro.com) Dataiku and Valid are adding their own layers on top of that base. Dataiku is supplying workflow generation and orchestration for enterprise users, while Valid is supplying a packaged fraud-decisioning application for banks and credit unions, based on the company’s description. That division of labor suggests Snowflake is becoming the execution layer for partners that want to sell governed AI systems without asking customers to rebuild data pipelines elsewhere. (docs.snowflake.com) ### What happens next for customers evaluating these products? Dataiku said Cobuild on Snowflake is available to joint customers now, and the company said hundreds of shared customers could use the integration immediately. Valid Systems said any organization already using Snowflake can access its offering through the new integration. Snowflake’s documentation pages for Cortex AI functions, Cortex Agents and Cortex Code remain the reference points for customers comparing what Snowflake provides natively with what partners are adding on top. (dataiku.com)