AI is changing the way data engineers work. With Snowflake Cortex Code (CoCo), developers get an AI-powered assistant directly within their Snowflake environment to help write code, explore data, and build data pipelines and models. But how far can you take AI-assisted development? And how do you make sure AI agents don’t execute unrestricted queries across your data?
In this webinar, we explore Cortex Code from development to governance through an end-to-end demonstration. We start by ingesting raw data from Azure into Snowflake and use CoCo to explore the data, identify potential data quality issues, and uncover relevant business opportunities and use cases. From there, we build a Medallion architecture with Bronze, Silver, and Gold layers, before transforming the data into a dimensional model with facts and dimensions ready for analytics and BI.
But AI-assisted development also introduces new risks. A poorly scoped prompt can trigger expensive queries, scan massive amounts of data, and result in unexpected credit consumption. That is why the second part of the webinar focuses on putting guardrails around Cortex Code and AI agents.
We explore how actions can be intercepted before execution, how queries can be scoped and constrained, and how a wrapper agent can provide an additional layer of control. We compare native capabilities within Snowflake with external orchestration through tools such as Claude Code or GitHub Copilot, and cover practical techniques for monitoring, governance, query efficiency, and cost control.
During a live demo, we show how a wrapper agent can intercept an unsafe, unscoped query and turn it into a controlled and efficient alternative.
By the end of the session, you will understand not only how Cortex Code can accelerate data development, but also how to introduce the right controls to keep AI-assisted development safe, predictable, and manageable. This naturally leads into the FinOps challenges that we will explore further in our second Snowflake webinar.
