Meet Snowflake CoWork and CoCo: the AI colleagues that work directly on your data

A Sanofi field rep can now ask one question and receive a full physician pre-call plan in seconds, a task that previously took hours of manual research. That was the sharpest proof point from Snowflake Summit 26, held June 1–4 at San Francisco’s Moscone Center, where Snowflake pitched Snowflake CoWork and CoCo not as assistants, but as AI colleagues embedded directly in enterprise data.

Why Snowflake CoWork and CoCo are more than rebrands

CoWork, formerly Snowflake Intelligence, is aimed at knowledge workers. CoCo, formerly Cortex Code, targets developers, data engineers and AI builders. Both are designed to follow users into the tools where work already happens, including desktop, mobile, Slack, Snowflake environments and developer workflows, rather than forcing everyone back into one interface.

The bigger shift is autonomy. 

CoWork brings business-aware research, reusable artifacts and context-driven reasoning. CoCo adds cloud agents, automations and task execution, moving beyond code suggestions into governed data engineering work. Snowflake says more than 7,100 customers were already building with CoCo before the Summit push.

The reason this matters is context. 

Snowflake says CoCo achieved a 72.1% pass rate on ADE-Bench, a dbt-created benchmark for evaluating AI agents on real-world analytics and data engineering tasks. That put it ahead of both Anthropic’s Claude Code and OpenAI’s Codex, which Snowflake says each scored 65.1%. The point is not that CoCo is a smarter chatbot; it is that an agent built inside the data stack can reason with schemas, permissions, lineage and production workflows that generic coding agents often miss. In regulated sectors such as healthcare, finance and legal, that grounding is critical because a technically plausible but contextually wrong data transformation can quickly become a business or compliance risk.

The CIO question: colleague or risk multiplier?

Snowflake’s bet is that enterprise AI will be won where governed data already lives. That puts it in direct competition with Microsoft Copilot, Salesforce Agentforce and Databricks-style agentic platforms, but with a different centre of gravity: proximity to trusted business data rather than productivity apps or CRM workflows. We have already explored why the next AI wave is about trusted action, why process context matters, and how AI agents turn platforms into governance infrastructure.

Still, CIOs should be cautious. 

Poor data governance will make these AI colleagues confidently wrong at scale. Autonomous workflows raise accountability questions. And another agent layer may reduce tool sprawl or simply rename it. Snowflake has made the product case; now enterprises must prove the operating model.