Enterprise AI Data Agents Snowflake Summit 26

Snowflake Summit Day 2: Is It Magic — or Trusted Data Agents?

Techmediaglobal

| 6 min read

7,100+

SNOWFLAKE COCO USERS

4 Acts

KEYNOTE FRAMEWORK

5 Years

UNDER ARMOUR + SNOWFLAKE

100s

PRODUCT LAUNCHES / QUARTER

Day 2 of Snowflake Summit 26 in San Francisco offered a sweeping vision of what enterprise AI can become — not the rogue, black-box intelligence of science fiction, but autonomous agents grounded in trusted data, governed access, and human-centred design. Led by Christian Kleinerman, EVP of Product, and Benoit Dageville, Co-Founder and President of Product, the keynote made one thing clear: when AI feels like magic, it is because the data foundation beneath it is rock solid.

From HAL 9000 to Helpful Agents

Christian Kleinerman opened the Day 2 keynote with a striking cultural reference — Arthur C. Clarke's classic assertion that any sufficiently advanced technology is indistinguishable from magic. It was a deliberate framing: where HAL 9000 in 2001: A Space Odyssey represented the dangers of unchecked autonomous intelligence, Snowflake's vision for the agentic enterprise is one where AI agents operate not in a black box, but anchored to verified, governed enterprise data.

"I've been saying this for years because when we see a great product, it feels magical — and we all live in a time where this proves true almost every single week," Kleinerman told the audience. The stakes are real: Snowflake ships hundreds of product launches every quarter, and the pace continues to accelerate.

Meet CoCo and CoWork: Snowflake's AI Duo Gets New Names

Two products have emerged as standout performers over the past year: Snowflake Cortex Code and Snowflake Intelligence. At Day 2, Kleinerman announced both are being rebranded to better reflect their position within Snowflake's broader Cortex AI portfolio.

Cortex Code — which launched in February 2026 as an AI-powered coding assistant that translates a single prompt into production-ready data pipelines and applications — has already been affectionately nicknamed "CoCo" by its growing user base of over 7,100 engineers and developers. It is Snowflake's fastest-ever growing product, and the name is now official: it will be known as Snowflake CoCo.

Snowflake Intelligence — the conversational assistant that allows employees to query company data in natural language, build AI agents for specific teams, and maintain data protection without additional security configuration — will now be known as Snowflake CoWork. Together, the two products align under Snowflake's Cortex AI umbrella, including Cortex Agents, which lets users build agents that retrieve, reason over, and act on governed enterprise data.

"CoCo is truly changing how we think about Snowflake — the surface area and how you're more productive and agile to get things done."
— Christian Kleinerman, EVP of Product, Snowflake

Anthropic and Snowflake: Accelerating the Path to Production

One of the summit's most notable partnerships on display was the deepening collaboration between Snowflake and Anthropic. Enterprise adoption of Anthropic Claude within Snowflake Cortex AI has been growing rapidly, prompting CEO Sridhar Ramaswamy to highlight on stage how the two firms are working together to help customers move faster from AI experimentation into full production deployment.

The partnership, expanded in December 2025, integrates Claude models directly into Cortex AI across all major cloud platforms and establishes a joint go-to-market strategy for deploying AI agents across customer data. The division of labour is clear: Anthropic provides the Claude model, while Snowflake integrates it directly within customers' existing secure data and governance environments — ensuring sensitive data never leaves those environments.

Christian's Four-Act Framework for the Modern Data Enterprise

Kleinerman structured his Day 2 keynote as a four-act theatrical production — a deliberate narrative device to walk the audience through Snowflake's full vision for enterprise AI transformation. Act One centred on simplification: removing traditional friction from data management so that complex architectures become accessible and actionable. Under Armour's five-year journey with Snowflake — moving from transactional partnership to a fully democratised data ecosystem — was the anchor example.

Act Two addressed security and agent identity — introducing the concept that enterprise systems should be able to recognise when activity is being performed by an agent versus a human, and apply appropriate masking policies accordingly. Act Three focused on openness: ensuring data flows freely across tech stacks without vendor lock-in, with Snowflake claiming full bidirectional support across catalogue engines. Act Four represented the destination: autonomous, intelligent workflows powered by CoCo and CoWork — turning governed data into enterprise-wide AI action.

Netflix was cited as a leader in multi-party collaboration — building shared secure environments where multiple organisations can analyse and activate data together without exposing raw PII or proprietary information, a capability Snowflake sees as foundational to the next phase of enterprise data work.

The Agentic Enterprise: One Platform, Zero Silos

Benoit Dageville, Co-Founder and President of Product at Snowflake, reinforced a critical architectural principle during his session: the agentic enterprise cannot be built on fragmented stacks. Enterprises need two things to survive the agentic era — a world-class data foundation, and a unified architecture where AI and data exist on a single integrated platform.

"Building an isolated AI stack will lead to sales silos, forgetting governance, and driving up cost and complexity," Dageville warned. Snowflake's own foundation principles — unifying structured and unstructured data, decoupling compute from storage for scalability, and delivering a fully managed zero-maintenance platform — are precisely designed to prevent this outcome.

The real-world proof point came from Under Armour. In partnership with Elementum and Snowflake, the company has launched an internal AI agent that allows employees to instantly query data and automate workflows. Patrick Duroseau, Chief Data AI Officer at Under Armour, captured the shift: teammates can now talk, chat, and engage with the agent to get answers quickly on their own — powered by the trusted data foundation that Snowflake and Elementum provide.

"That partnership with Elementum and Snowflake gives us the trusted data that our teammates are looking for."
— Patrick Duroseau, Chief Data AI Officer, Under Armour

Key Takeaways

→Snowflake rebranded Cortex Code as Snowflake CoCo and Snowflake Intelligence as Snowflake CoWork, consolidating its two fastest-growing AI products under a unified Cortex AI identity.
→CoCo is Snowflake's fastest-ever growing product, with more than 7,100 users since its February 2026 launch — converting single prompts into production-ready data pipelines and apps.
→Snowflake and Anthropic deepened their partnership to integrate Claude models into Cortex AI across all major cloud platforms, enabling secure deployment of AI agents directly within existing customer data environments.
→Christian Kleinerman's four-act keynote framework — Simplification, Security & Agent Identity, Openness, and Autonomous Workflows — outlined a complete roadmap for how enterprises transition to the agentic era.
→Benoit Dageville reinforced that a fragmented AI stack creates silos, governance failures, and rising complexity — positioning a unified data and AI platform as the only viable architecture for the agentic enterprise.
→Under Armour's live deployment of an internal AI agent — built with Elementum on Snowflake's data platform — demonstrated how trusted data transforms from a reporting asset into an autonomous, enterprise-wide productivity tool.