GenAI - Snowflake GenAI Overview - Part I

1. GenAI - Snowflake GenAI Overview - Part I

Snowflake embraces simplicity. That was true when it was first created, and it's even more true now, because over time Snowflake has built out one of the most comprehensive AI platforms in the data world. They've done it in a way that makes AI genuinely accessible to anyone who can write SQL. That's what we're going to cover in the next two videos. I want to be upfront about something before we dive in. AI is moving fast, and Snowflake's AI capabilities are moving with it. The goal of these two videos isn't to make you an expert on any one of these features. The goal is to give you a map of the territory, so that as you go deeper into Snowflake, you know what tools exist and roughly where to reach for them. Let's start with a concept that ties all of this together. Snowflake calls itself the AI Data Cloud. The idea is that AI isn't bolted onto the platform — it's built directly into the same platform that manages your storage, your compute, and your governance. That means when you use Snowflake's AI features, your data never has to leave Snowflake's secure environment. The same access controls, the same role-based permissions, and the same governance policies you've already set up all apply to your AI workloads automatically. That's a bigger deal than it might sound. Now, let's get into the actual product landscape. I find it helpful to think about Snowflake's AI features on a spectrum — ranging from things you can use in seconds with almost no setup, all the way to highly customized AI workflows that take more care to build. We'll work our way along that spectrum across these two videos. On the far left, we have a set of features that are accessible right from the Snowsight UI with essentially no code required. The first is **AI and ML Studio**. This is a no-code interface built into Snowsight that lets you access Snowflake's AI and ML capabilities through a point-and-click experience. If you want to run a forecast on your data or classify some records without writing a single line of SQL, AI and ML Studio is where you go. It's a great entry point for people who are newer to AI or who want to move quickly without worrying about syntax. The next one is **Document AI**. This is a feature that uses a Snowflake-managed model to extract structured data from unstructured documents like PDFs, forms, or contracts. You can point it at a whole folder of documents and get back structured output that lands directly in a Snowflake table. If you ever have a pile of PDFs that you wish you could query like a database, Document AI is exactly what you're looking for. Then there is **Universal Search**. This is a natural-language search bar built into Snowsight that lets you search across your database objects, Snowflake Marketplace listings, Snowflake documentation, and community resources all at once. If you're looking for a table, a dataset, or even just trying to figure out how to do something in Snowflake, Universal Search is a surprisingly powerful starting point — assuming you know what you're looking for. Then there is **Snowflake Copilot**. Copilot is an LLM-powered assistant built into the Snowsight experience. You ask it a question in plain English, and it generates a SQL query for you. You ask a follow-up question or make a suggestion, and it refines the query. It's useful both for people who are newer to SQL and for experienced developers who just want to move faster. Let's talk about what sits underneath all of these features: **Snowflake Cortex**. Snowflake Cortex is the intelligent, fully-managed service that powers most of what you've seen and a lot more. You can think of it as the basket of ready-to-use AI functions and services inside Snowflake. The things we just covered are built on top of Cortex. But Cortex also gives you direct access to AI capabilities from inside your SQL workflows. The Cortex LLM functions are the most direct expression of this. These are functions you can call from inside a `SELECT` statement with no model management and no infrastructure required. Just call the function and get a result. We'll go deep on one of these, a function called `AI_COMPLETE`, in a few lessons. But I want to give you a quick sense of the current lineup: - `AI_SENTIMENT` classifies text as positive, negative, or neutral. - `AI_TRANSLATE` translates between languages. - `AI_EXTRACT` lets you ask natural-language questions of unstructured text. - `AI_CLASSIFY` categorizes content into caller-defined groups. - `AI_PARSE_DOCUMENT` extracts content from files such as PDFs. - `AI_COMPLETE` is the most flexible of all. It lets you send any prompt to any supported model and get a response back. We'll spend a whole video on `AI_COMPLETE` because there is a lot to learn there. One important note: you may see references to a function called `SUMMARIZE` in older Snowflake materials or documentation. `SUMMARIZE` has been deprecated, and Snowflake's recommendation now is to use `AI_COMPLETE` with a summarization prompt instead, which gives you more control anyway.

2. Let's practice!

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.