Snowflake Summit 2024 Recap

The Hex team just got back from Snowflake Summit 2024, which was, according to new Snowflake CEO Sridhar Ramaswamy, “the largest gathering of AI and Data professionals in the world”! After walking the conference floor and talking to attendees, it’s clear that the buzz was swinging towards the AI end of that spectrum — the majority of data vendor booths sported shiny new AI taglines; every talk had “AI” in the title; last year’s slogan of “The World of Data Collaboration” was replaced by “Building the Future Together with AI and Apps.” Snowflake is going all in on AI.

But there’s an interesting subtlety in how Snowflake thinks about AI, or specifically what they’re calling “Enterprise AI.” I’d argue that their AI vision (which is well, visionary) has a real sense of… reality and the simple truth of what enterprises need to see from AI vendors.

There’s a lot more to dive into later. For now, here’s a brief recap of the event and a few thoughts on “Enterprise AI.”

AI Partner of the Year — What Does it Mean? 

We’re proud to have been selected as Snowflake’s AI Partner of the Year, marking the second year in a row that Hex has been chosen as a Partner of the Year. Especially with so much focus on AI, it’s exciting to have our commitment to integrating with and powering the AI data cloud recognized in this way.

Whew, that paragraph had a lot of buzzwords. What does it actually mean to be AI Partner of the Year? Why’d they choose us? It’s simple: Hex is the easiest way to work with Snowflake data for AI and ML features.

Since the first public release of Snowpark in 2022, Hex has been the easiest and most powerful tool for using Snowpark and Snowpark ML in data analysis. With hundreds of shared customers, we’ve seen this first-hand. The key is in our “easy button” for Snowpark, which puts all the power of Snowpark literally just one click away: keep using your familiar data workspace, write the same familiar SQL or Python code, and seamlessly add in powerful ML capabilities that run at warehouse scale. 

And last week at Summit, Snowflake released new updates to Cortex AI: a library of Generative AI services from RAG and search to custom fine-tuning for AI models. These complement the traditional ML functions and services of Snowpark, and just like we do with Snowpark, Hex plugs directly into Cortex with zero extra configuration and a fully native integration. You can keep working with your data in exactly the efficient way you’re accustomed to and effortlessly add in powerful generative AI capabilities where they make sense.

This is a really similar philosophy to how we’ve built our own Magic AI tools. AI needs to be there when you need it, and get out of your way when you don’t. 

For Hex, this means our Magic AI tools are completely invisible to your development workflow until you summon them, in which case they can take action throughout your workspace to fix errors, edit code, and carefully write semantically accurate SQL queries.

For Snowflake, this means that you interface with complex new technologies like LLMs and massively scalable ML infrastructure in the same, simple way you’d access any data: with a SQL query. 

Together, this means there’s no faster way to bring generative AI (or traditional ML!) into your data pipelines and apps than using Hex with Snowflake’s AI and ML features. We’re proud to have received a recent round of funding from Snowflake to continue to enhance this vision and to have been officially recognized by Snowflake as a leading AI partner. We can’t wait to see what you build with these tools. 

Supporting insights fast lanes

While we all continue to prepare for bigger changes in AI (more on that later), we know that data teams are still tackling endless amounts of requests and are working on increasing the speed of analysis. We’re glad to hear from customers like Toast that Hex is helping them build a “data insights fast lane,” where the data team can run advanced analytics and be a data “proving ground” before sharing data upstream. At an organization with a 5,000+ employee base, data teams need a tool that lets them do advanced analysis fast and drive trusted insights across the org to keep up with day-to-day goals. Their data-insights fastlane in Hex was so successful, Toast’s internal usage of Hex has grown 4x in the last year (collaboration!). 

Enterprise AI

This year, the banner theme at Snowflake Summit was: The Era of Enterprise AI. And AI was certainly everywhere, but the official Snowflake flavor is actually subtly different than the overall AI fervor sweeping the entire world. Reading between the lines of their carefully crafted marketing and piecing together quotes from sessions with data leaders, an interesting message appears:

AI promises to be this magic, do-anything tool that can be applied to almost any problem to do things faster and more efficiently. It sounds like a silver bullet for businesses big and small. But getting this new technology deployed in ways that have real business impact is going to be way harder than most people in the AI space freely admit — especially at large enterprises.

"For the first time, everyone in the organization can talk to their data in fluid, natural language… But, here’s the issue. The bar for AI use in enterprise is much higher than in consumer AI. Consumer AI is not ready for business use."

- Snowflake CEO Sridhar Ramaswamy

“How do we move at the speed that execs are expecting, but with these technologies that are not proven yet?”

- Darlene Newman, Head of Innovation and AI at DTCC during an executive panel

These two quotes, one from a CEO selling AI and one from a Head of Data trying to use it, sum things up pretty neatly. “Not proven yet” is essentially synonymous with “not enterprise ready.” Enterprises are under immense pressure from execs and shareholders to jump on the hot new AI train (or at the very least to not get left behind when it leaves the station). But to have real value, you need to integrate this unproven new tech into the deepest, most critical and secure parts of your data stack — and that’s both really hard, and really scary. Companies are going to be slow, and hesitant to do that.

So basically, Snowflake’s offering to do it for them! With this whole “Era of Enterprise AI,” Snowflake wants to take a complex, infinitely tuneable technology (LLMs) and package it up into a simple, efficient, secure platform that it’s easy for big companies to adopt — or perhaps seamlessly integrated into a platform they’ve already adopted 😉.

And this is essentially what they’ve already done so successfully with data warehousing. The key feature of Snowflake is arguably that it doesn’t have as many bells and whistles and things to worry about as competitors. No indexes or partitions to configure, nothing to carefully tune and tweak and monitor. It’s mostly ready to use out of the box, on whatever shape and size of data you can throw at it. 

From the Snowflake documentation: 

• ​​There is no hardware (virtual or physical) to select, install, configure, or manage.

• There is virtually no software to install, configure, or manage. 

• Ongoing maintenance, management, upgrades, and tuning are handled by Snowflake. 

Imagine the LLM version of that, but tack on a few more bullet points about how there’s no data movement / egress necessary, no additional transformations, and no need to learn new programming languages or frameworks, and I’d reckon you’ll get something close to the Snowflake Enterprise AI vision.

Some of this is still a little fuzzy, and we’re reading between the lines somewhat. But the message from data leaders is clear: “We are having a tough time moving as fast as we’d like to with this new technology” and the reply from Snowflake is sounding something like “We’re going to make it dead easy for you.” 

And of course as AI partner of the year, we’re excited for Hex to be a key part of the way that Snowflake helps enterprise data teams get real value from AI, both generative and traditional.