JOIN US @ SNOWFLAKE SUMMIT 2025!
A yellow arrow pointing to the right.
Stipo Josipovic
Director of Product Management
Greg Owen
Software Engineer
June 3, 2025

Launching Cortex AISQL: A New Chapter For Snowflake And Sigma

June 3, 2025
Launching Cortex AISQL: A New Chapter For Snowflake And Sigma

Unstructured data has always been out of reach for traditional analytics. Contracts, images, PDFs—anything that didn’t live in a table was stuck in a separate workflow, handled manually, or ignored altogether. That changes with the launch of Snowflake’s Cortex AISQL.

Now, teams can query raw text, documents, and images using standard SQL. And with Sigma, that capability becomes part of your everyday workflow—just files in and answers out.

The human bottleneck for unstructured data? Solved

BI tools have always operated on the assumption that data is clean, structured, and neatly arranged in rows and columns. But when it comes to unstructured data, gaining insight has traditionally required human intervention—to read documents, classify images, or manually extract the signal.

This process is slow, limits productivity, and doesn’t reflect how teams actually work. We call this the human bottleneck. 

Some of the most important decisions rely on data traditional BI tools can’t access—compliance PDFs, images, annotated docs, product specs, screenshots, receipts. Cortex AISQL changes that by allowing Snowflake to process unstructured data directly, while Sigma serves as the interface to centralize unstructured data into your warehouse and then analyze it. By pairing Snowflake’s AI engine with Sigma’s interface, you can scale human context without increasing your workforce. Want to process 5,000 vendor contracts? Classify images by content? Extract key terms from unstructured text? 

You can do all of that directly in your existing stack—with built-in governance, high-performance, and speed. You can upload files into Sigma, analyze them with Snowflake’s LLMs, and use the results like any other dataset. No external tools, no translation layers, no friction. 

What used to take a team now takes a query

This isn’t just about adding AI. It’s about automating what people already do with data—without the manual work.

Think legal records. Housing deeds, rulings, contracts—documents that are dense, hard to parse, and full of critical information. Before you needed a person to read and interpret each individual record, but now you can run those same files through AI that’s seen thousands of examples, and get answers instantly. What used to be a tedious, scale-constrained, and error-prone process becomes an automated workflow. And it’s not limited to legal. It applies across teams, industries, and formats.

For example, financial services companies are some of the most heavily regulated organizations in the world, and with that comes mountains of documentation—standard operating procedures, compliance protocols, internal legal docs. They need something simple: the ability to ask questions against those documents like they would a database.

With Snowflake’s Cortex AISQL, they’re able to upload documents in Sigma directly and analyze them in real time. They can build live workflows that include both structured data and the real-world context that’s always been missing. All of this lives in Sigma and runs on Snowflake.

This is how you stop making decisions based on what's easy to analyze—and start making them based on This is how you stop making decisions based on what's easy to analyze—and start making them based on the right information—which has likely been excluded all along.

A tool you already know how to use

So why has unstructured data been such a pain for BI tools? The answer is simple: technology couldn’t handle it. Only humans could. But with Cortex AISQL, that manual review process gets replaced by something that scales—LLMs that can actually read and process the content, just like a person would.

In Sigma, this process is straightforward. You upload a file and it goes into a governed cloud bucket you control. Snowflake’s Cortex AISQL points directly at that bucket, processes the file using an LLM, and returns structured results. Sigma takes that data and displays it in real-time, directly in the workflow of your dashboard or data app.

There’s no need for middleware, reformatting, or special pipelines—just upload files and get answers. That simplicity is exactly what makes this solution not just powerful, but usable.

It also unlocks new types of workflows by allowing you to attach real evidence to operational data. Reviewing insurance claims? Drop in receipts, police reports, or photos. Investigating a financial adjustment? Upload the PDF with full context. The result is governed, centralized, and integrated with your ecosystem.

We’ve seen customers in wealth management who used to email files back and forth for audit purposes. Now, those same files are directly part of the app, creating a more efficient and reliable system with a better UI. 

Your stack just got a whole lot smarter

One of the key advantages of Cortex AISQL is that Snowflake designed it to work like standard SQL from the start. That means no new syntax, no bolt-on tools, and no friction. Just structured access to unstructured data, at scale.

Sigma makes that functionality usable—fast. Teams can work with the full spectrum of their data using SQL or our formula language, all in one place, without changing how they work.

For anyone looking to combine structured and unstructured data into real workflows, this is the stack that makes it possible. The manual bottlenecks are removed, and unstructured data is now fully integrated into the analytics workflow.

THE STATE OF BI REPORT