Over the past two years, AI has helped employees generate content, answer questions and accelerate analysis. But most AI systems still wait for instructions. The next leap is proactive collaboration: agents that understand your business, learn how you work, act safely across enterprise systems and help drive better decisions and faster execution.

With Snowflake CoWork (formerly Snowflake Intelligence), companies can shift from using AI in reaction to requests to using it for active, collaborative assistance.

CoWork is a personal work agent for every knowledge worker. It can reason deeply, automate routine tasks and accelerate the path from ideas to decisions to action. By combining deep understanding, automation and enterprise context, CoWork helps organizations turn AI into measurable business outcomes.

_{Figure 1: Snowflake CoWork: A personal work agent for every knowledge worker.}

Understands your business deeply

An agent is only as good as its understanding of your business. Historically, getting AI systems to interpret enterprise data correctly has required weeks of setup before delivering meaningful value.

Cortex Sense (set for private preview soon) will introduce a foundational context layer that automatically learns how a business defines its data and operations. It will understand essential elements like key business metrics, relationships between data sources, and standard analytical processes without requiring extensive manual configuration.

Using signals from query history, metadata, dashboards in tools like Power BI and Tableau, and enterprise data outside Snowflake, Cortex Sense will understand concepts like revenue definitions, fiscal calendars and snapshot tables. This will enable CoWork to deliver more accurate, grounded answers from day 1.

This means knowledge workers will stop waiting on data teams to "set up AI" and start getting value from day one. Based on internal testing on complex enterprise queries, CoCo and CoWork achieved an 83% accuracy rate when using Cortex Sense compared to 47% for CoCo and CoWork alone, and 23% for Frontier Coding Agents with Snowflake MCP. This enables significantly more dependable business reasoning and decision support.

To accelerate adoption even further, Cortex Sense will include a series of ready-to-use plugins (private preview soon) for domains such as finance and sales. These plugins will combine skills, business logic and MCP connectors to help enterprises deploy production-ready agents faster. They will automatically ground agents in operational knowledge and enterprise context, enabling more accurate, domain-specific actions and insights. Teams can go from zero to a context-aware agent in minutes, not months.

This capability will work alongside Deep Research (set to be generally available soon), allowing users to thoroughly examine complicated business questions across both structured and unstructured enterprise data. Powered by the agent swarm orchestration system built by Snowflake's AI Research Team, Deep Research delivers state-of-the-art performance across structured and unstructured enterprise data, outperforming single-agent systems by over a third on Snowflake's Hybrid Deep Research Benchmark.

_{Figure 2: Investigate complex questions across enterprise data using Deep Research.}

Analytical Search (set for public preview) will incorporate multi-step analysis of unstructured data by conducting aggregation, comparison and trend analysis across vast collections of documents. Questions that once took weeks of manual effort will now be answered instantly with computed results.

For questions such as why gross margin is down in the EU or why renewals are slipping in APAC, CoWork will decompose the problem, run parallel research agents across structured and unstructured data, cross-validate findings and return a fully cited report in minutes. A finance leader can now ask, "Why is the EU gross margin down this quarter?" CoWork will analyze revenue, cost trends and currency effects from governed data, incorporate relevant business context and deliver an answer complete with citations. Teams can now complete in minutes work that would typically take a senior analyst days.

One personal agent that knows you

Every employee interacts with AI differently. A CMO needs campaign performance and budget pacing. A supply chain director needs inventory alerts and supplier lead times. With these innovations, every user gets a single agent that adapts to how they work instead of forcing them into the same workflow.

Multi-agent orchestration (coming in public preview soon) will allow CoWork to automatically direct every question to the best data, skills and tools, eliminating the need for users to manually select which workflow or system to engage. With personalized MCP connections and persistent memory across sessions, CoWork is designed to adapt to how each employee works and deliver a more seamless, context-aware experience over time.

Memory (public preview soon) for remembering user preferences and facts will give CoWork continuous personalization based on user interactions. CoWork will remember preferences, role-specific patterns and recurring tasks, so teams can move faster instead of starting over every session.

User Skills (coming in public preview soon) will let any user record and transform a multi-step routine into a reusable skill for the entire organization. These skills can automate routine tasks such as preparing a Monday recap, running a multi-step variance analysis, and drafting customer follow-ups, simply by describing the workflow in natural language.

_{Figure 3: Turn workflows into reusable skills easily in Snowflake CoWork.}

Skills can also invoke the Code Execution Tool (public preview), allowing users to execute business logic and generate outputs such as PDFs and PowerPoint presentations directly from CoWork, making it easy to share polished deliverables across teams. Knowledge workers will get polished, shareable deliverables, not just text answers, and builders never have to configure the underlying compute.

_{Figure 4: Generate reports and presentations automatically.}

Over time, organizations will build a library of trusted automations and reusable workflows. For example, a sales ops manager could describe a workflow such as: "Pull pipeline changes by region, flag deals that slipped stage, and draft a summary email for leadership." Teams can share this skill with others in the organization, driving productivity for all. Users can easily discover these skills, creating shared operational knowledge rather than one-person workarounds locked away in someone's notes.

Runs proactively

The most valuable agent work happens before anyone asks.

Consider a VP of sales on a Monday morning. Before a single question is asked, the agent will have already surfaced three at-risk renewals, flagged a consumption drop at a top account and drafted a check-in email for the account executive, based on overnight analysis across customer and operational data. No dashboard was checked. No analyst was pinged.

Automations and time-based subscriptions (coming in public preview soon) make this possible. CoWork will run in the background, checking for specific conditions, identifying anomalies and providing timely updates via email, Slack or mobile alerts.

One simple configuration: "Every Monday, compare each account's consumption to the prior week. If any drops more than 20%, brief me with the root cause and recommended action." The agent will handle the rest.

This is the difference between a dashboard you check and an agent that checks for you.

Some business problems cannot be solved in a single interaction. They require longer-running analysis, coordination across systems and iterative reasoning. Async Agent API (generally available soon) will allow agents to work on complex tasks that may take a few minutes or hours to complete, while continuing to run asynchronously in the background. This enables knowledge workers to delegate complex investigations and workflows to CoWork with minimal guidance.

Together, these capabilities will transform AI from a question-and-answer interface into a proactive work partner that helps teams move faster and stay ahead of what matters.

Builds shareable team knowledge

One-off answers do not scale. CoWork outputs will now be reusable across organizations.

Artifacts (set for general availability soon) will let users save and share robust outputs, including analyses, charts and entire conversation histories, preserving all original context. Every artifact will stay live: It will refresh with the latest data and respect role-based access (RBAC) per viewer. A finding from one analyst will become a starting point for the next, not a screenshot buried in a slide deck.

_{Figure 5: Share analyses, not screenshots.}

Artifacts on CoWork now include interactive dashboards (public preview soon). Authors can publish multi-tile North Star dashboards directly in CoWork, scoped by RBAC. Users can ask follow-up questions directly against these dashboards, filter by region or period, and pull findings into the next presentation.

Every leader will see the same source of truth, all governed by the same policies that protect the underlying data.

_{Figure 6: Use interactive dashboards for collaborative decision-making.}

Conversation Sharing (generally available soon) will make it easy to share a thread with a colleague, complete with context and citations, so collaboration can happen around real answers, not paraphrased ones.

For teams that live in Slack, the new Snowflake CoWork Slack app (private preview soon) will bring the conversational agent experience with deep links to CoWork for more. Users can ask CoWork questions in any conversation, get streaming responses with rendered charts inline and follow up in thread without ever leaving Slack. Multi-user OAuth ensures every response respects the asker's permissions, so the same governance that protects data inside CoWork extends seamlessly into the flow of team communication.

Acts across the tools your teams already use

Insights only matter when they turn into action. Through MCP connectors, CoWork can draft emails in Gmail, update Jira tickets, post to Slack and log activity in Salesforce, all from the same conversation and within Snowflake's governance boundaries, all within the flow of work. Teams can move from insight to action without switching tools or workflows.

_{Figure 7: Enable one-click MCP connections to enterprise tools.}

The iOS mobile app (set for general availability soon) makes the complete CoWork experience accessible to anyone on the team, wherever they are working. Face ID unlocking and complete conversation history mean a knowledge worker can approve an agent-drafted customer email from the conference floor or review a morning briefing from the airport.

Every response reflects the data the user is authorized to see. Every action executes within admin-defined policies. Every interaction is fully auditable.

_{Figure 8: iOS mobile app: Your personal work agent, anywhere.}

Built for the people who scale agents

Behind every great user experience is a powerful builder experience. Agent Studio (generally available soon) is the redesigned experience for building, managing and governing agents.

An agent readiness wizard will make it faster to go from prototype to production, guiding creators through agent configuration, evaluation and deployment. A single pane of glass will give builders visibility into usage across agents in the organization, latency, eval scores and tool utilization across every agent in the org, so builders can easily improve and monitor agent quality.

And to accelerate time-to-value even further, builders turning their dashboards into conversational apps can now instantly create a governed semantic view directly from any existing Power BI dashboard (public preview).

Admins can now use RBAC, along with a complete audit trail and admin-defined policies to ensure autonomous actions stay within the organization's boundaries. Admins will decide what an agent does on its own, what needs sign-off and what it cannot use, with every action logged with its policy reason.

Every user in your business now has a personalized work agent

Snowflake CoWork started with a promise: Help teams act on what happened, not just understand it.

Now, that promise extends beyond delivering answers to help teams execute work, automate repeatable processes and turn individual insights into reusable organizational capabilities.

Snowflake CoWork is the personal work agent where you work smarter. It's trusted, proactive and continuously working for you.

Get started

• Try it: Snowflake CoWork quickstart
• Watch: Summit 2026 keynote and product demos
• Learn more: Snowflake Cortex documentation

All references to CoWork above refer to Snowflake CoWork.

Forward-looking statements

This article contains forward-looking statements, including about our future product offerings, and are not commitments to deliver any product offerings. Actual results and offerings may differ and are subject to known and unknown risk and uncertainties. See our latest 10-Q for more information.