Simplify The Entire Data Development Lifecycle with Snowflake CoCo

This video was created and contributed by, Snowflake. 

Overview

Snowflake CoCo is described as the industry's first data-native AI coding agent. It works with built-in awareness of Snowflake's data, compute, governance and operational semantics. The demo follows a team building a live VIP fan dashboard that combines Salesforce sales data with streaming fan activity, and shipping it in days rather than weeks. From one natural-language prompt, CoCo Desktop builds the app and deploys it to Snowflake. When pipeline steps break, CoCo is opened inside the running app. It scans task history, traces the dependency chain, writes the fix and verifies the full pipeline run. Normally that kind of diagnosis means hours of troubleshooting across several engineering teams. The architecture behind it needs no code changes. Salesforce connects through zero-copy integration, and an existing Kafka producer feeds a Snowflake data stream into a managed Iceberg table that Horizon Catalog governs. That gives one governed copy of the data and no duplication. With Snap and Ask, users can select part of a chart and ask a question directly in the app. In the demo, that surfaced a localized drop in fan engagement.

Key points

• Build from a prompt: Describe the app in natural language and CoCo builds and deploys it to Snowflake.
• Diagnose and fix pipelines: Finds the root cause, writes the fix and verifies the run in one request, replacing hours of cross-team troubleshooting.
• Data-native awareness: Understands Snowflake's data, compute, governance and operational context.
• One governed data copy: Zero-copy Salesforce integration and Kafka-fed data streams land in a Horizon-governed managed Iceberg table, with no duplication.
• Snap and Ask: Select a chart segment and ask a question in the app, with no separate SQL worksheet.
• Faster time to value: Shortens the path from business idea to secure, trusted implementation.