Snowflake announced a set of updates to Snowflake Intelligence and Cortex Code on April 22, positioning its data platform as the centralized “control plane for the agentic enterprise.” The updates add Model Context Protocol connectors to six enterprise applications, natural-language workflow automation via Skills, an iOS mobile app, multi-step reasoning through a deep research feature, and cross-platform data support that extends Cortex Code to AWS Glue, Databricks, and Postgres. Over 9,100 customers are using Snowflake’s AI products weekly, according to the company’s announcement.
The timing is deliberate. Google just rebranded Vertex AI to the Gemini Enterprise Agent Platform at Cloud Next 2026. Microsoft shipped Agent Governance Toolkit v3.2.2 with end-to-end encrypted agent messaging the same week. Snowflake, which does not own a foundation model or a hyperscale cloud, is making a different argument: the company that already governs your data should also govern the agents that act on it.
What Snowflake Actually Shipped
Snowflake Intelligence, first launched as a data exploration assistant, now functions as a context-aware agent for business users. The core additions, detailed in Snowflake’s blog post on Cortex Agents:
MCP connectors (generally available soon) provide native integration with Gmail, Google Calendar, Google Docs, Jira, Salesforce, and Slack. The governance model that applies to Snowflake database objects, including roles, grants, and audit logging, extends to these external MCP servers. A sales operations team can connect Salesforce and Jira instances and interact through Snowflake Intelligence to track pipeline, upload customer reports, and create tickets automatically.
Skills (generally available soon) allow users to describe multi-step workflows in natural language. Snowflake Intelligence then executes them independently: preparing presentations, running forecasts, generating reports, sending follow-ups. Skills are modular and shareable across teams. A forecasting skill built by an analytics team can be reused by operations, sales, and marketing agents without duplicating code.
Deep research (public preview soon) uses an agentic architecture to reason across structured data, unstructured data, and external context in multi-step reports. Each report documents the agent’s reasoning chain with citations.
Code Execution Tool (public preview soon) gives agents a sandboxed Python environment with session-level isolation. Additional packages are available through Snowflake’s Artifact Repository, and external network access can be granted via network rules.
iOS mobile app (public preview soon) extends the entire interface to smartphones.
“Snowflake gives customers one place to bring their data together, connect the systems they rely on, and turn AI into something that actually helps teams get work done,” Baris Gultekin, VP of AI at Snowflake, said in a statement reported by SiliconANGLE.
Cortex Code Becomes a Cross-Platform Builder Layer
Cortex Code, launched in November 2025 as an AI coding agent for data workflows, has been adopted by more than 50% of Snowflake customers, according to Techzine. The expansion includes:
Support for external data systems including AWS Glue, Databricks, and Postgres, allowing developers to build applications without migrating data into Snowflake. Integration with other AI systems through MCP and Agent Communication Protocol (ACP), enabling interoperability with existing agent frameworks. A VS Code extension (private preview) and a Claude Code plugin let developers access Cortex Code from their preferred environments. A new Agent SDK supporting Python and TypeScript embeds Cortex Code capabilities into custom applications. Cortex Code Sandboxes in Snowsight provide browser-based cloud environments for running code end-to-end without local setup.
The cross-platform support is the most strategically significant element. As SiliconANGLE reported, the new capabilities extend Cortex Code across external data systems, “allowing developers to build applications without migrating data.” This is a direct concession that enterprise data lives everywhere, and Snowflake’s path to becoming a control plane runs through interoperability, not data gravity.
The Competitive Landscape: Three Models for Agent Governance
The enterprise agentic platform race has crystallized into three distinct approaches over the past two weeks.
Google’s full-stack ownership model. At Cloud Next 2026, Google rebranded Vertex AI to the Gemini Enterprise Agent Platform, consolidating agent development, runtime, governance, security, and orchestration into a unified stack. Google shipped ADK v1.0, Agent Memory Bank, Agent Identity, A2A protocol, and TPU 8t/8i custom silicon purpose-built for agent workloads. As Forbes reported, Google announced bidirectional federation with Databricks Unity Catalog, Snowflake Polaris, and AWS Glue, conceding that enterprise data will not move to a single cloud but positioning Google Cloud as “the query and reasoning layer over data that lives elsewhere.”
Microsoft’s governance-first model. Microsoft released Agent Governance Toolkit v3.2.2 with end-to-end encrypted agent messaging via Signal protocol, wire protocol specifications, and registry/relay services. Microsoft’s approach focuses on the security and compliance layer for agents built on Azure.
Snowflake’s data-native model. Snowflake argues that the platform that already governs enterprise data, including row access policies, role-based grants, and audit logging, should extend that governance to agents. The pitch is that enterprises don’t need to build new governance infrastructure for agents; they can reuse the policies they’ve already defined for their data.
Analyst Assessment: Catching Up or Pulling Ahead?
Industry analysts are divided on whether Snowflake’s approach represents a genuine competitive advantage or table-stakes feature parity.
David Menninger, analyst at ISG Software Research, told TechTarget that the capabilities are “required investments but not necessarily unique features.” He noted that Databricks offers Genie Code, an agent for generating code, and a VS Code extension. “It’s a race among data platform vendors to build the largest AI ecosystems with the most capabilities to attract and retain users,” he said.
Stephen Catanzano, analyst at Omdia (a division of Informa TechTarget), offered a more favorable assessment to the same outlet: “While competitors like Databricks, AWS and Google Cloud offer similar tools, Snowflake’s focus on combining enterprise-grade governance with seamless AI integration across data, tools and workflows sets it apart.”
Michael Leone, VP and principal analyst at Moor Insights & Strategy, told InfoWorld that the dual-audience approach is “worth slowing down on.” Most vendors pick one target, users or builders, and come back to the other later. “Snowflake is putting both on the same governed data foundation. That’s a harder engineering problem, but I’d argue it’s a cleaner answer to the question enterprises are actually asking, which is how to open AI up to more people without losing control of the data underneath.”
Igor Ikonnikov, advisory fellow at Info-Tech Research Group, was more measured. He told InfoWorld that Snowflake “has caught up with the competition, but not yet surpassing it,” noting that “common semantics are still buried inside database models and code.”
Sanchit Vir Gogia, chief analyst at Greyhound Research, identified the central tension in Snowflake’s strategy in the same InfoWorld report: “Control without ownership of the systems where work is executed introduces dependency that is difficult to fully resolve.” He characterized Snowflake as “one of the most credible contenders in a race that will be defined not by who builds the smartest AI, but by who can make that AI work reliably inside the enterprise.”
The 95% Problem
A data quality challenge looms behind every platform announcement. A July 2025 MIT report found that 95% of organizations have not yet gotten any return on their investments in AI, according to TechTarget. A survey of 540 data practitioners by The Modern Data Company, cited by SiliconANGLE, found that 89% identified finding the right data as a top-three time drain, 61% struggle with poor naming conventions, and 60% lack tools to discover what data exists.
Will Allen, head of Snowflake Intelligence and agents, addressed this directly in TechTarget’s coverage: “Time and time again, we hear from customers that the bottleneck for enterprise AI isn’t the models, but rather tapping into the data and context needed to make those models useful.”
This frames Snowflake’s pitch precisely. In a market where every vendor claims agent capabilities, Snowflake’s argument is that the data layer, not the model layer, is the binding constraint. If agents can’t access governed, high-quality data in real time, their autonomy is theoretical. Snowflake, which already sits on the data, is betting that extending governance from data objects to agent actions is a shorter bridge than building data governance from scratch atop a cloud or model provider.
The “Coming Soon” Risk
A notable portion of the announcement is forward-looking. Skills, MCP connectors, the mobile app, deep research, and Code Execution Tool are all listed as “generally available soon” or “coming soon in public preview.” Michael Leone at Moor Insights flagged this directly, calling the roadmap “ambitious” and noting “the number of items announced that are ‘coming soon’ or are in public preview,” according to InfoWorld.
For Snowflake, the execution timeline matters more than the announcement. Google’s Gemini Enterprise Agent Platform shipped ADK v1.0 and Agent Memory Bank as production-ready features at Cloud Next, with 150 organizations already running on the platform. If Snowflake’s capabilities remain in preview through Q3 2026, the window for establishing the control plane position narrows considerably.
The Architecture Bet
Snowflake’s strategy inverts the conventional platform play. Google and Microsoft own the compute, the models, and the cloud infrastructure, then pull data toward their platforms. Snowflake owns the data governance layer, then extends outward to agents, applications, and external systems.
Sanjeev Mohan, principal at SanjMo, told InfoWorld that the cross-platform support is significant: “What Snowflake is saying is that even if your data lives in a competitor’s system, Snowflake AI coding agent can be used. And vice versa, the VS Code extension and Claude Code plugin can be used on Snowflake data. In other words, it reduces vendor lock-in fears.”
The question is whether enterprises will choose their data platform as their agent governance layer, or choose their cloud provider, their model vendor, or a purpose-built agent platform. With $997 million in Q1 FY26 product revenue (up 26% year-over-year) and 9,100 weekly AI product users, Snowflake has the installed base to make the argument. Whether it can ship fast enough to make the case before the hyperscalers cement their own control planes is the $4.3 billion question.