Snowflake joins the race to own the agentic AI control plane

File photo: CIOs gather at the CIO Association of Canada’s Peer Forum in Vancouver. — Photo by Jennifer Friesen, Digital Journal

Enterprise AI is about to stop being human-facing. According to John Encizo, field CTO and director of enterprise AI for North America at Lenovo, AI agents will begin interacting directly with one another over the next 12 to 18 months, changing how companies handle everything from APIs to security.

“When your customer is another AI, lots of things happen. Your API strategy becomes your agent strategy, and authentication becomes much more complicated,” said Encizo.

Encizo was speaking at the CIO Association of Canada’s Peer Forum in Vancouver last week, describing the next phase of enterprise AI to a room of Canadian technology leaders. 

Every major data and cloud vendor is now racing to ship the governance and infrastructure that companies will need before that moment arrives. Last week, Snowflake made its move.

Snowflake’s response

The latest updates to Snowflake Intelligence and Cortex Code aim to create a “control plane for the agentic enterprise.” 

In infrastructure terms, a control plane is the layer that decides what’s allowed to happen, who has access, and where things run. 

Snowflake Intelligence is being pitched as a personal work agent for business users, with new Model Context Protocol (MCP) connectors into Salesforce, Slack, and other enterprise tools. Cortex Code, the developer side, now integrates with Databricks, Postgres, and AWS Glue, with a Claude Code plugin in preview.

Snowflake says more than 9,100 customers use its AI products weekly. Cortex Code, which launched in February, is now active in more than half of the company’s customer base, according to the release.

The idea is that whoever owns the governed data layer also owns the seat where enterprise AI is granted permission to act. A category is forming around that idea. 

The premise is that AI can’t be trusted to act until the data underneath it can be.

Back in January, I interviewed Anahita Tafvizi, Snowflake’s chief data and analytics officer, about what she called documentation debt.

“If you don’t have it documented to train a human, then how do you train an AI agent,” says Tafvizi.

That argument is now the basis for Snowflake Intelligence and Cortex Code.

At machine speed, undocumented data doesn’t produce results the business can trust. Snowflake’s answer is to make the governed data layer the control plane, the place where AI gets permission to act.

Snowflake’s bet is that the same system that manages your data should also control how AI is allowed to use it.

Why CIOs might be skeptical

A governed, consolidated platform sounds like exactly what most enterprises need. 

In practice, it’s the kind of decision CIOs have learned to make carefully. The vendor that becomes your data platform today is the vendor you depend on to move anywhere in the future. 

The sovereignty and lock-in conversation is already well underway, as Digital Journal covered last month through the work of former federal AI technologist JP Lalonde, who said that the organizations building exit paths now will have options that those who wait won’t.

James McGregor, CTO of the City of Kelowna, described how he’s thinking about that at CIOCAN’s Peer Forum.

“We have a partnership with Microsoft, and we use Copilot in an enterprise way. As a parallel effort, we’re now exploring Google, and we’re exploring Anthropic to try and figure out other tools that we can start to bring in for different use cases, but also look at optionality so we’re not getting locked in,” McGregor said.

McGregor’s instinct to keep other doors open shapes how the control plane pitch might be received. 

Snowflake, Databricks, Microsoft, and the other vendors selling governance as the reason to centralize are also the vendors a CIO has to be most deliberate about adopting. 

The appeal of a single platform and the cost of betting on one are inextricably linked.

Where your data lives will shape how your AI works

Tafvizi’s argument in January was that the data foundation is the gating item for AI. Canadian CIOs at CIOCAN’s Peer Forum discussed the same constraint within their own organizations. 

Snowflake’s announcement is the productized version of the argument, and the positioning is already showing up across the category.

The decision facing technology leaders is how deliberately to choose the platform through which their data and agents will route. 

Until the foundations underneath are in place, the cost of choosing wrong won’t surface.

Final shots

• The protocols that let AI agents from different vendors work together, like MCP, are becoming the connective tissue of the category. Watch which vendors adopt them and which build closed systems instead.
• Companies will need to know which AI did what and be able to trace it after the fact, and most don’t have the tools to do that yet.
• The data readiness question sits underneath all of it. The platform layer depends on the data. Classic garbage in, garbage out.