Snowflake Inc. today is digging deeper into the emerging agentic enterprise model with the introduction of a research preview of an autonomous artificial intelligence platform designed to help business users automate complex tasks using Snowflake-governed data.

The company said the platform, called Project SnowWork, is intended to move AI beyond queries and toward systems that can plan and execute multi-step workflows using enterprise data. While positioned as allowing users to “simply ask for what they need and have SnowWork securely complete the task,” the platform works best with existing processes, said Bala Kasiviswanathan, vice president of developer and AI experiences at Snowflake.

“Project SnowWork is most powerful when it’s working with governed data, defined workflows and an understanding of how the business operates,” he said. “Where it really adds value is not just automating tasks, but accelerating decision-making. It can synthesize data, surface insights and recommend next steps.”

SnowWork runs on governed enterprise data stored on Snowflake’s platform and integrates business context, such as metrics, definitions and access policies. Snowflake executives said grounding AI agents in enterprise data is essential if companies want AI to move beyond experimentation and become a trusted operational tool.

“The real issue enterprises are facing right now isn’t just building agents,” Kasiviswanathan said. “It’s that most approaches aren’t actually solving the business problem fast enough, or in a way that’s grounded in real enterprise context.”

SnowWork orchestrates tasks such as querying datasets, analyzing results, generating reports and preparing presentations within a single interaction. The company said the system can handle workflows that span multiple enterprise systems while observing the same governance and security rules applied to the underlying data.

Kasiviswanathan actions taken by the AI agent remain transparent and auditable.

“Every action inherits role-based access controls, data policies and audit logging automatically,” he said. “That means it can only act on data the user is allowed to see, and every step is fully traceable.”

The company also emphasized mechanisms it has put in place to reduce the risk of unreliable outputs or hallucinations when the system performs complex tasks.

“Enterprises can inspect the steps, validate outputs and maintain control over how and when actions are executed,” Kasiviswanathan said.

Snowflake positions the new offering as the next stage in a long-running industry effort to democratize access to analytics. Previous waves of “self-service” data tools promised to eliminate the need for specialized data teams, but many organizations still rely heavily on analysts to create reports or interpret dashboards, it noted.

Kasiviswanathan said SnowWork aims to go further by completing tasks rather than simply returning answers. “We’re collapsing the entire chain from question to analysis to outcome into a single interaction,” he said. “Project SnowWork doesn’t just surface insights, it carries them through to a finished deliverable or recommended action.”

Another key point of differentiation is context, he said. “Project SnowWork understands how the business actually operates,” he said. “This isn’t just democratizing access to data, it’s democratizing the ability to act on it.”

Snowflake didn’t say when Project SnowWork will be generally available.

Image: SiliconANGLE/Microsoft Designer