A Guide to Semantic Layers for Snowflake CoCo and CoWork
By Mark Palmer
How Snowflake Semantic Views reach CoWork, CoCo, Power BI, and Excel without leaving the platform.
At Snowflake Summit 26, Snowflake renamed Cortex Code to CoCo and turned Snowflake Intelligence into CoWork, a personal agent for every knowledge worker. The headline is autonomy: agents that write code, run recurring workflows, and answer business questions on their own. But every one of those agents is only as trustworthy as the definitions underneath it.
Snowflake and AtScale also announced Snowflake Semantic Views XMLA Endpoint, powered by AtScale, available in private preview soon. It extends the governed definitions in Snowflake Semantic Views to Microsoft Power BI and Excel through live connections. This post is your guide to how semantics actually work across these new CoCo and CoWork surfaces, and how the same definitions reach the tools your business already lives in.
The pieces
There are five moving parts. Here’s what each does.
CoCo is Snowflake’s coding agent, formerly Cortex Code. It writes SQL, builds pipelines, and now runs autonomous, event-driven workflows. When you ask it a data question, it searches for and queries the relevant semantic view first, falling back to raw tables only if none exists.
CoWork is the personal agent for knowledge workers, formerly Snowflake Intelligence. It answers business questions in plain language, grounded in trusted definitions rather than ad hoc queries against raw tables.
Snowflake Semantic Views are the semantic foundation. They define your metrics and dimensions once, in Snowflake, so that gross margin means the same thing whether CoCo, CoWork, or Cortex Agents are asking. This is where your business logic should live.
The XMLA Endpoint, powered by AtScale, is the new piece. It carries those same Semantic View definitions out to Power BI and Excel through live connections. No mirrored data. No stale extracts. No metric logic rebuilt in a second tool where it quietly drifts out of sync. Data and compute stay in Snowflake.
AtScale Enterprise is the upgrade path. As your semantic sophistication grows, you’ll want to do more. You might want to connect legacy Tableau dashboards or compute in Databricks. You’ll want advanced control over your semantics, including time intelligence, hierarchies, dimensional modeling, advanced calculations, and additional tools, agents, and platforms, while keeping Snowflake at the center. When you’re ready for this, Snowflake and AtScale have you covered.
How it all works together
Picture a retailer’s finance team.
An analyst defines gross margin once, as a Snowflake Semantic Views. That single definition now grounds everything downstream.
The data engineer asks CoCo to build a margin-by-region pipeline. CoCo reads the semantic view and inherits the right logic. The VP of Finance asks CoWork, in plain English, why margin slipped in the Northeast, and gets an answer built on the same definition. The board dashboard, live in Power BI through the AtScale XMLA Endpoint, shows that exact number. The finance team’s Excel pivot, connected via AtScale, gets the same answer.
One definition. Four surfaces. Same answer. And when that team needs rolling 13-month trends and a product hierarchy the embedded views don’t cover, and they also want the numbers in Tableau, and so on, AtScale Enterprise adds all of it without moving the foundation off Snowflake.
Why it matters
For years, companies survived inconsistent numbers because humans patched the gaps. The dashboard said one thing, the spreadsheet said another, and someone in finance reconciled the two before the board meeting. That grace period is over.
“AI did not create the metrics consistency problem. It exposed it,” says AtScale CEO Chris Lynch. “Agents won’t get the benefit of the doubt.”
Ask CoWork for gross margin and ask Power BI the same question. If the two answers disagree, you don’t have an AI problem. You have a semantics problem.
Govern your semantics in Snowflake, give live access from the tools your people already use, and keep compute on the platform. That’s the bet. In the agent era, it’s the only one that holds up.
See it in action at snowflake.atscale.com, or read the full post. AtScale was recognized as a Leader in GigaOm’s 2025 Semantic Layer Radar and serves more Fortune 500 enterprises in production than any other semantic layer.