Transcript
Snow Music is preparing for a massive global tour and wants to ensure that their VIP fans, the biggest drivers of revenue, have an absolute epic experience. To do that, we need a VIP fan command center, a similar application that merges Salesforce data with live fan activity. And we need to ship it in days and not weeks. Let's jump into Coco to get started. So I'm standing here in Snowflake Coco desktop, which is purpose-built for heavier development work, writing complex logic, connecting data, and overall data management. Here, I'll prompt Coco using natural language to build our VIP fan command center. Here's my prompt. Build a Streamly dashboard to monitor our Snow Music tour data pipeline with live streaming metrics and engagement analytics. And then deploy to Snowflake. And just like that, Coco gets to work. Coco immediately begins building code based on a deeply context-aware understanding of our environment. Now, for the purpose of this demo and our time constraints, I've already built and deployed this app using Coco. I'm going to navigate straight over to the running Simlet app in Snowsite and show you how we can use Coco to diagnose and fix a broken pipeline live. Here's a live fan command center application running in Snowsite. Everything looks incredible, but looks like we've hit a couple of roadblocks. The app is highlighting a couple of broken pipeline steps. In the old world, diagnosing a stale data pipeline across live apps, CRMs, and other data sources would take hours of troubleshooting across multiple engineering teams. But Coco is here to help, meeting builders where they are to solve real complex challenges. All right, so here we go. My fan app events data pipeline tasks failed. Can you diagnose the root cause, fix it, and verify the full pipeline run successfully? We use Coco desktop to build, and now I can open Coco right here in Snowsite to investigate the broken pipeline in context of our live app. While Coco is thinking, scanning our task history, checking failed runs, and tracing the dependency chain, here's what's happening under the hood. In the background, Coco is analyzing an incredibly elegant modern data stack. With just that single prompt we used earlier, it successfully managed connections to Salesforce via zero-copy integration alongside streaming data from the VIP fan app. We can see that the live streaming data refreshing in our app here in every few seconds. The ingestion of that live fan data is coming in via Snowflake data stream. We simply activated it by connecting our existing Kafka producer to a data stream topic with zero code changes. And the live data materializes into a native Snowflake managed iceberg table almost immediately. Because the fan app data lives in a managed iceberg table, it is cleanly governed through Horizon Catalog. Horizon automatically registers this iceberg table in the universal catalog, making it instantly accessible to any engine. One data copy, fully governed, zero duplication. This is exactly why Coco is a category-defining tool. It is the industry's first truly data-native AI coding agent, meaning it operates with a deep, intrinsic awareness of Snowflake's data, compute, governance, and operational semantics. And look at that, Coco has already generated the precise code fix and verified the downstream dependencies. Now you can see that the pipeline status cards are green. Coco turns what used to be an exhausting multi-team engineering escalation into a simple, intent-driven interaction, safely accelerating our time to production with high accuracy and trust. Let me show you Snap and Ask. Now that the data is flowing perfectly, I'm noticing a massive drop in fan engagement in San Francisco. So instead of jumping into a separate worksheet to write a complex SQL, I can use Snap and Ask directly inside our app. By making the selection on the chart and asking a direct question of the data using Snap and Ask, Coco instantly understands the exact data segment I'm referencing to surface business insights. Now this is only possible because our live fan app activity joined with Salesforce records in real time. Proving that Coco doesn't just confirm data movement, it deeply understands business context. Right here, Coco instantly identifies app activity and live fan engagement plummeting. It highlights this specific localized dip, giving us a real-time warning so our local team can step in and check on the venue's fan experience immediately. What you've just seen is the power of the industry's first truly data-native AI coding agent. Coco redefines how data and developer teams work, seamlessly interacting across the entire modern data stack. This demo showed how we can turn application development, complex data engineering, and advanced analytics into a simple, natural language interaction, and how Coco dramatically compresses the distance between a creative business idea and secure, trusted implementation. With that, SoulFlate Coco empowers builders to optimize workflows and speed time-to-value faster than ever before. www.soulflate.com
