Snowflake intros SnowWork to automate tasks with AI
The platform, which is now in research preview, is designed to provide users with agent-powered collaboration capabilities that can automate actions and workflows.
Snowflake on Wednesday introduced Project SnowWork, an AI-driven platform designed to make business employees more efficient and productive by autonomously executing previously manual tasks.
The new platform -- which is in research preview and only available to a limited group of customers -- integrates Snowflake's data platform with agentic AI capabilities so that users can express what they want and have SnowWork carry out the work.
SnowWork includes preconfigured capabilities that give it context about specific domains within organizations such as finance, sales, marketing and operations so the AI understands the appropriate workflows, terminology and key performance indicators. In addition, the platform features multi-step task completion capabilities and built-in security and access controls.
Combined, SnowWork's capabilities aim to enable users to plan and automate workflows based on their Snowflake data, generate analysis that includes recommended actions, and automate data, AI and business systems to reduce the time it takes to execute tasks and take actions.
Given that SnowWork enables users to go beyond the limits of dashboards and reports, which only address specific subjects, and do so in a secure manner, SnowWork will be an important addition for Snowflake users once it is enterprise-ready, according to Sanjeev Mohan, founder and principal of analyst firm SanjMo.
"It's significant because … it's interactive and it's also collaborative -- that's the beauty of it," he said "Users aren't limited by what their dashboard was built for. As a businessperson, they can ask all kinds of questions on data that is sitting inside their enterprise. And because it knows who I am, it will apply the roll-based access control so if I'm not supposed to see a different campaign, I'm not going to see it."
It is not likely that SnowWork will give Snowflake a competitive advantage over other data platform vendors, Mohan continued, noting that Databricks on March 11 launched Genie Code, an agent that can carry out tasks such as building pipelines and debugging failures.
However, capabilities such as SnowWork and Genie Code will give data management vendors an advantage over large language models (LLMs) that don't have secure access to the proprietary data that makes agents contextually aware of an enterprise's unique characteristics.
"I think [all the data platform vendors] will have something similar," Mohan. "People ask me why they need something from Snowflake or Databricks when an LLM can do similar things. If you compare SnowWork with LLMs like Google Gemini, the difference is stark because none of the model providers have access to enterprise data with all the security and governance."
AI everywhere
Based in Bozeman, Mont., but with a campus in Menlo Park, Calif., Snowflake is a data platform vendor that, like peers such as Databricks and Teradata as well as hyperscale cloud vendors including AWS and Microsoft, has added AI development capabilities over the past few years.
It's significant because … it's interactive and it's also collaborative -- that's the beauty of it. Users aren't limited by what their dashboard was built for.
Sanjeev MohanFounder and principal, SanjMo
Snowflake, however, did not embrace AI immediately after OpenAI's November 2022 launch of ChatGPT marked significant improvement in generative AI technology and sparked surging interest among enterprises in AI development. Instead, Snowflake started nearly a year later than Databricks and other vendors, finally making the creation of an AI development environment and infusion of AI capabilities throughout its platform a priority when Sridhar Ramaswamy was named CEO in February 2024.
Now, the vendor is not only making AI part of its platform with tools such as Snowflake Intelligence, whichenables data exploration using natural language rather than code and Snowflake OpenFlow, which automates data ingestion and integration, but making AI a ubiquitous companion.
Bala Kasiviswanathan, Snowflake's vice president of developer and AI experiences, noted that AI to date has been good at responding to questions, but not at executing actions based on insights generated from queries. SnowWork represents Snowflake's attempt to extend AI beyond helping analyze data to building workflows based on analysis.
"Businesses have spent millions on data, yet humans are still stuck manually moving files, clicking buttons, and chasing approvals to turn an AI's 'insight' into a result," he said. "Project SnowWork is the bridge between talking and doing. ... We're moving past the 'chatbot' phase and into an era where AI doesn't just tell you what's wrong, it helps you fix it."
Like Snowflake, many data management and analytics vendors have introduced agentic AI capabilities since agents replaced generative AI as the cutting edge of enterprise AI in late 2024 and early 2025.
For example, ThoughtSpot is planning to automate its entire platform with agents for building dashboards, creating semantic models and developing embedded analytics applications. Informatica similarly now provides agents for data quality, data exploration and data discovery, among other tasks.
With the introduction of SnowWork, Snowflake is going beyond building task-specific agents. Instead -- once out of the research and preview stages -- SnowWork will provide agentic AI capabilities that essentially collaborate with users to execute workflows.
"SnowWork is poised to be a strong addition for Snowflake users," said Stephen Catanzano, an analyst at Omdia, a division of Informa TechTarget. "It bridges the gap between data insights and actionable outcomes, enabling business professionals to execute complex workflows. ... Its ability to turn governed data into immediate, measurable results makes it a significant step forward in enterprise AI."
Regarding the individual capabilities that comprise SnowWork, multi-step task completion is perhaps the most valuable because it removes the time previously needed to go from insight generation to building workflows based on those insights, Catanzano continued.
"The multi-step task completion stands out as the most valuable because it allows users to move seamlessly from data queries to actionable deliverables without manual intervention, drastically reducing time and effort while ensuring accuracy and consistency," he said.
Mohan likewise named multi-step task completion the most valuable of SnowWork's capabilities. He noted that LLMs have had the reasoning to perform multi-step workloads since late last year, but lacked the proper context to do so in an enterprise scenario.
Snowflake's combination of semantic modeling capabilities which define data, and situational awareness on top of a semantic layer, now enable agents to perform complex workflows.
"The missing link has been the context," Mohan said. Once we have semantics and context, then we can fully leverage the capabilities of LLMs. Snowflake now has the glue between the end user and the capabilities of the LLM."
Meanwhile, though Mohan predicted that most data platform vendors will soon provide similar capabilities, Catanzano noted that SnowWork has the potential to be a differentiator for Snowflake given SnowWork's integration with Snowflake's data platform and inclusion of security and governance.
"While other vendors are exploring similar AI-driven capabilities, SnowWork’s integration with Snowflake's governed data platform and its focus on secure, enterprise-grade execution sets it apart, potentially positioning Snowflake as a leader in this emerging space," he said.
Looking ahead
As Snowflake plans product development, working with customers to refine SnowWork as it moves from research preview to eventual general availability is one priority, according to Kasiviswanathan. Another is improving tools such as Snowflake Intelligence to generate AI-powered insights and Cortex Code to build and deploy AI workloads, he continued.
"Our focus is on helping customers move from AI experimentation to real business impact," Kasiviswanathan said. "That means making it dramatically easier for both business users and developers to put AI into production."
One thing Snowflake could do to refine SnowWork is extend its capabilities to external data in addition to the data stored in Snowflake, according to Mohan.
He noted that while Snowflake customers might keep much of their data in the vendor's secure environment, they also might, for example, keep data in Apache Iceberg tables or in data stores with the Model Context Protocol as their interface. However, if Snowflake extends SnowWork to external data, security and governance will be crucial.
"Snowflake has done a great job building context on top of all the data sitting inside Snowflake," Mohan said. "What Snowflake can do next is acknowledge that they need to build context across data that is both internal and external and then put SnowWork on top. The problem is that security and governance become problems when you're going external. [Still], it's an extension for Snowflake to consider."
Catanzano similarly suggested that Snowflake make SnowWork interoperable with external systems while also noting that the vendor could expand the platform's persona-specific capabilities.
"To continue serving its users and attracting new ones, Snowflake could expand SnowWork’s persona-specific skills to cover even more roles and industries or enhance its interoperability with third-party tools, ensuring it becomes an indispensable part of every enterprise's workflow ecosystem," he said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.
