Snowflake CoWork: AI Agent for Knowledge Workers

Snowflake is betting big on proactive artificial intelligence with its newly branded Snowflake CoWork, formerly known as Snowflake Intelligence. The company aims to transform AI from a tool that merely answers questions into a collaborative partner for knowledge workers.

The core idea is to equip every employee with a personal work agent that understands their business, learns their work habits, and acts safely across enterprise systems. This shift from reactive to proactive assistance is designed to drive better decisions and accelerate execution.

Understanding the Business Context

A key challenge for AI agents has been grasping enterprise-specific data and operations. Snowflake CoWork addresses this with Cortex Sense, a foundational context layer set for private preview. It automatically learns business definitions, data relationships, and analytical processes.

Cortex Sense uses signals from query history, metadata, and external data sources to understand metrics like revenue definitions and fiscal calendars. Internal tests show an 83% accuracy rate for CoWork using Cortex Sense, a significant jump from 47% without it.

Ready-to-use plugins for domains like finance and sales are also planned, enabling faster deployment of context-aware agents.

Deep Dives and Data Analysis

The platform also enhances its analytical capabilities with Deep Research, moving towards general availability. This feature leverages an agent swarm orchestration system to analyze both structured and unstructured enterprise data.

Analytical Search, entering public preview, will perform multi-step analysis on unstructured data, answering complex questions in minutes that previously took weeks of manual effort. For instance, investigating a drop in gross margin could be decomposed, researched across data sources, and presented with citations.

This capability allows finance leaders to ask questions like, "Why is the EU gross margin down this quarter?" and receive a comprehensive, cited answer.

A Personalized Agent for Every User

Snowflake CoWork is designed to adapt to individual users. Instead of forcing employees into a single workflow, each user gets a personalized agent.

Multi-agent orchestration, coming to public preview, will automatically direct queries to the most appropriate data, skills, and tools. This eliminates the need for users to manually select systems.

Features like Memory (public preview) and User Skills (coming soon) will enable continuous personalization. Memory will retain user preferences and recurring tasks, while User Skills will allow employees to convert multi-step routines into reusable organizational automations described in natural language.

These skills can automate tasks like preparing weekly summaries or drafting customer follow-ups. They can also invoke the Code Execution Tool to generate outputs like PDFs and presentations directly within CoWork.

Proactive Assistance and Collaboration

The agent is designed to work proactively. Automations and time-based subscriptions will allow CoWork to run in the background, checking for anomalies and providing timely updates via email or Slack.

For example, a VP of sales could receive a summary of at-risk renewals and flagged consumption drops before their week even begins, based on overnight analysis.

The Async Agent API, generally available soon, will support longer-running analyses and iterative reasoning for complex tasks.

Building Shareable Team Knowledge

Snowflake is emphasizing reusability and collaboration. Artifacts, set for general availability, will allow users to save and share outputs like analyses and charts with preserved context. These artifacts will remain live, refreshing with new data and respecting access controls.

Interactive dashboards within CoWork will allow users to ask follow-up questions directly against the data and integrate findings into presentations. This ensures a single source of truth governed by the same policies as the underlying data.

Conversation Sharing will enable easy sharing of entire threads with colleagues, facilitating collaboration around verified answers.

A new Snowflake CoWork Slack app (private preview) will bring this conversational experience directly into team communication channels.

Integration with Existing Tools

CoWork aims to act across the tools teams already use. Through MCP connectors, it can draft emails in Gmail, update Jira tickets, post to Slack, and log activity in Salesforce, all within Snowflake's governance framework.

An iOS mobile app (general availability soon) will provide access to the full CoWork experience on the go, allowing users to approve emails or review briefings from anywhere.

Empowering Builders

For developers and administrators, Agent Studio (generally available soon) offers a redesigned experience for building, managing, and governing agents. It includes an agent readiness wizard and a centralized dashboard for monitoring usage, latency, and performance.

Admins will have granular control over agent actions, requiring sign-offs for sensitive operations and ensuring all autonomous actions remain within organizational boundaries.

Snowflake CoWork represents a move toward trusted, proactive AI that works continuously for users, helping them execute work and turn individual insights into organizational capabilities.

Snowflake announced the new capabilities on June 2, 2026.