Goodbye Data, Hello AI: My Biggest Takeaway from Snowflake Summit 2026

AI

Data Science

Snowflake

Technology

Programming

Goodbye Data, Hello AI: My Biggest Takeaway from Snowflake Summit 2026

Apache SeaTunnel

15 min read

·

Jun 11, 2026

--

By William Guo, CEO of WhaleOps & Snowflake Ambassador

I would like to thank Snowflake for inviting me to attend Snowflake Summit as a Snowflake Ambassador.

This Summit had a much greater impact on me than I had expected.

As many of you know, I have spent my career in the data industry. I started at Teradata, then moved to IBM. Later, I was responsible for big data initiatives at enterprises such as Lenovo, CICC, and Wanda Group. After that, I became a Member of the Apache Software Foundation, and today I am the CEO of WhaleOps Open Source. Because of this background, I have always paid close attention to developments across the data industry.

Before coming to the Summit, I originally thought Snowflake would launch several enterprise AI products or add AI-related capabilities on top of its existing data warehouse and data platform offerings.

For many years, people’s understanding of Snowflake has been quite clear: it is a cloud data warehouse company and a representative of the Data Cloud era. Its core strengths revolve around data storage, compute, performance, security, governance, sharing, and elastic scalability.

However, after spending two days at the Summit, my impression changed completely.

The biggest takeaway I had from this year’s Snowflake Summit was not that it had released some new data platform features. Rather, it is aggressively reconstructing its product positioning.

In my view, Snowflake is no longer satisfied with being defined as a Data Warehouse company. Nor does it simply want to become an AI Data Cloud. Instead, it aims to transform itself into an enterprise AI + Data platform, and perhaps even the foundation of the Agentic Enterprise, putting itself on a path that increasingly overlaps with companies like Anthropic.

If I had to summarize my personal impression of this Snowflake Summit in one sentence, it would be:

Goodbye Data, Hello AI.

Of course, “Goodbye Data” does not mean data is becoming less important.

On the contrary, data has become even more important.

What has changed is the way data platforms are expressed and understood.

In the past, when we talked about data platforms, we talked about how data should be stored, processed, shared, governed, and optimized for cost efficiency.

Today, Snowflake is talking about how AI can understand enterprise data, how Agents can use enterprise data, how business users can gain insights directly through natural language, and how enterprises can enable AI to execute tasks within secure and governed boundaries.

Snowflake Product VP Christian Kleinerman made a statement during the Platform Keynote that perfectly captures this shift:

Press enter or click to view image in full size

Your AI-native enterprise starts here.

If this sentence had appeared at a typical AI conference, it might have sounded like a standard marketing slogan.

But in the context of Snowflake Summit, it carries a very different meaning.

Because Snowflake is not an AI-native company. Historically, it has been a data infrastructure company. When a company with that background begins reorganizing its entire product portfolio around AI, it signals that AI is no longer an add-on feature — it is becoming a force that reshapes the enterprise itself.

1 Snowflake’s Transformation: From Data Warehouse to AI Platform

In the past, when I thought about Snowflake, the first things that came to mind were data warehousing, cloud-native architecture, elastic computing, the separation of storage and compute, data sharing, and unified governance.

What Snowflake solved were several long-standing problems in traditional data platforms: fragmented data, limited scalability, complex performance tuning, inconsistent governance, and high collaboration costs.

The central narrative of this year’s Summit was clearly different.

Snowflake still talks about All Data, All Workloads, and All Users. It still talks about structured, semi-structured, and unstructured data. It still talks about Iceberg, OpenFlow, Streaming, Zero Copy, and Horizon Catalog.

However, these capabilities are no longer being positioned simply as components of a better data platform. Instead, they are being framed as the foundation for a new goal: enabling enterprise AI and Agents to operate on a unified data platform.

Christian Kleinerman also made another highly important statement during the Platform Keynote:

“We need a unified architecture, both AI and data.”

This statement can almost be regarded as the strategic core of this year’s Snowflake Summit. It is not saying, “We support AI too.” Rather, it is saying that enterprises should not build a separate AI platform outside of their data platform.

Why?

Because if the AI platform and the data platform are separated, many of the same problems we experienced during the data era will reappear: new silos, new permission systems, new governance gaps, new cost black holes, and new security risks.

We spent more than a decade eliminating data silos. If we build AI on an entirely separate stack today, we are essentially creating AI silos all over again.

So Snowflake’s answer is clear: AI and Data must be unified. Data, compute, semantics, governance, security, applications, and Agents should all form a closed loop within a single platform.

Viewed from this perspective, Snowflake’s Summit slogan, Make AI Real for Business, is fundamentally about turning Data into the context, fuel, and execution foundation for AI.

In the past, data platforms were built for people. People wrote SQL, viewed dashboards, configured jobs, and performed analyses.

In the future, data platforms will increasingly be built for Agents. Agents will understand business questions, invoke data capabilities, generate analytical workflows, propose actions, and even participate directly in business processes.

This is what truly struck me at this Summit.

Snowflake is not simply adding an AI assistant on top of an existing Data Warehouse. It is using data to rebuild a new AI-native foundation for the Agentic Enterprise, which is also where OpenAI and Anthropic will ultimately compete.

That is why I believe Snowflake’s transformation is far more aggressive than I originally imagined.

2 CoCo, CoWork, and Desktop: Snowflake Is “Paying Tribute” to Anthropic — And Revealing a Bigger Ambition

If the first layer of change is strategic positioning, then the second layer is the product portfolio itself.

At this year’s Snowflake Summit, what impressed me most was not a traditional database feature or a performance improvement metric. Instead, it was the launch of an entire collection of AI Agent-centric products and components:

CoCo

CoWork

Desktop

Skill Catalog

VS Code Extension

Excel Add-in

MCP

ACP

Cloud Agents

Agent Teams

Automated Agents

When viewed together, they send a very clear signal:

Snowflake is reorganizing its product strategy in the same way an AI-native company would.

In fact, I would even say it is “paying tribute” to Anthropic.

Why do I say that?

Because AI-native companies such as Anthropic are no longer just building chatbots. They are building complete AI work systems, including Claude, Claude Code, Desktop, MCP, Artifacts, Skills, Computer Use, enterprise context, and security boundaries.

What they truly want to own is not merely a conversational interface, but the primary interface through which humans collaborate with software in the future.

The CoCo, CoWork, Desktop, Skill Catalog, and MCP/ACP announcements from Snowflake have remarkably strong parallels.

CoCo feels like Claude Code for the enterprise.

CoWork resembles an AI workspace for business users.

CoCo Desktop extends Snowflake’s AI capabilities beyond the web console and into users’ everyday work environments.

Skill Catalog packages Snowflake platform capabilities into discoverable, composable, and reusable skills that Agents can invoke.

So when I heard these announcements at the event, my first reaction was not:

“Snowflake has released a few more AI features.”

Instead, it was:

Snowflake wants to repackage the data platform as a complete Enterprise AI Agent Operating System and enter the same strategic battleground occupied by OpenAI Enterprise and Anthropic Enterprise.

Snowflake officially announced that Cortex Code would no longer be called Cortex Code. It has been renamed to Snowflake CoCo:

Press enter or click to view image in full size

“From here on, no more Cortex Code. It is officially Snowflake CoCo.”

This statement is worth paying attention to.

The name Cortex Code still carried the feeling of being a coding assistant.

CoCo, on the other hand, feels much more like a standalone AI product.

Behind this rebranding is a larger ambition.

Snowflake does not want CoCo to be merely an assistant that helps users write SQL, generate code, or explain syntax. It wants CoCo to become the AI operating interface for the entire Snowflake platform.

Christian also mentioned during the keynote that over the past several months, CoCo has evolved beyond CLI and SnowSight experiences and expanded into MCP, ACP, SDKs, Agent Teams, Cloud Agents, automation capabilities, and Skill Catalog.

Among these, Skill Catalog is especially important. It enables users to share, discover, and reuse Skills. In essence, it is modularizing Snowflake platform capabilities and turning them into reusable tools for Agents.

This is extremely important.

Snowflake also explicitly announced upcoming Excel add-ins, VS Code extensions, and Marketplace partner integrations for CoCo.

During discussions at the event, many of us felt the Excel integration was particularly powerful because Excel remains the most familiar data workspace for business users.

VS Code, meanwhile, remains the most familiar workspace for developers.

Rather than forcing everyone into SnowSight, Snowflake is bringing CoCo directly into the environments where people already work.

Press enter or click to view image in full size

This is also one of the most important principles behind AI-native products:

Do not force users to move into your interface. Bring your Agent into the user’s workflow.

Therefore, the significance of CoCo is not that Snowflake now has its own Copilot.

The significance is that Snowflake is moving away from a traditional platform UI and toward an Agent Everywhere strategy.

Beyond CoCo, Snowflake also placed a major spotlight on CoWork at this Summit.

To be honest, when I first heard about CoWork, I was a bit puzzled. If Anthropic were launching CoWork, I could easily understand it, because Agents naturally require enterprise-grade collaboration. But from the perspective of a traditional data platform, CoWork did not seem like the kind of product Snowflake would be expected to release. CoCo helping data engineers write SQL, fix pipelines, and build applications makes perfect sense. OpenFlow, Streaming, Iceberg, and Horizon Catalog are also clear enhancements to the data platform. But what does CoWork have to do with a data warehouse?

After listening to the presentation, I gradually understood it. CoWork reveals Snowflake’s ambitions even more clearly. It is designed for business users, with the vision of enabling CEOs, sales teams, operations teams, marketers, and other business professionals to interact directly with enterprise data and gain insights as if they had their own personal Jarvis. Samsung shared a use case illustrating this idea: CoCo serves as the AI operating interface for data engineers and developers, while CoWork serves as the AI workspace for business users. Snowflake is not just trying to serve data teams; it wants to become part of the daily workflow of every business user across the enterprise.

At that point, I finally understood CoWork’s role. CoCo is r

[truncated for AI cost control]