Chris Child, Snowflake & Ben Kus, Box

News
More from Snowflake Summit 2025

Christopher Child

VP of Product, Data Engineering

Snowflake

Ben Kus

Chief Technology Officer

Box

Three insights you might have missed from theCUBE’s coverage of Snowflake Summit

Agentic AI is rewriting the rules of enterprise data engineering, turning passive infrastructure into intelligent systems that act, adapt and automate at scale.Gone are the days when platforms simply stored and queried data. Today’s data stacks are being rearchitected to support agentic workflows — systems where metadata, governance and orchestration layers work in concert to guide machine reasoning. Snowflake Inc.’s embrace of open formats such as Apache Iceberg signals a broader shift: Enterprises are building environments where intelligence is embedded into the fabric of data itself

The AI use case for unstructured data: How Box and Snowflake make content actionable

The race is on to turn unstructured data into real business value — and AI is finally making it possible.For years, documents, contracts and images sat buried in disconnected systems; now they’re finally being recognized for the value they hold. Generative AI and smarter data pipelines are helping organizations transform this raw content into actionable insights. The goal isn’t just to automate, it’s to unify fragmented data ecosystems and make better decisions, faster. As platforms evolve to bridge unstructured and structured data, businesses are rethinking what their content can actually do, according to Ben Kus (pictured, left), chief technology officer of Box Inc

Info
Transcript

Chris Child, Snowflake & Ben Kus, Box

Christopher Child

VP of Product, Data Engineering Snowflake

Ben Kus

Chief Technology Officer Box

In this interview at Snowflake Summit 2025, Box CTO Ben Kus and Snowflake VP of Product, Data Engineering Chris Child join theCUBE’s Dave Vellante and John Furrier live on the show floor to unpack the new Snowflake Openflow partnership. The conversation spotlights how Box’s rich store of unstructured content – from documents to video – can now flow directly into the AI Data Cloud, unlocking seamless analytics and AI development alongside enterprise-grade governance.

Kus explains why customers have struggled to extract insights from content siloed in ... Read more

Chris Child, Snowflake & Ben Kus, Box

>> Hello everyone, and welcome back to theCUBE's live coverage of the Snowflake Summit 2025 here at the Moscone Center in San Francisco. I'm your host, Rebecca Knight. This is our last segment of the day, and we've saved the best for last, I would like to introduce our last guest of the day, Chris Child, VP of Product, Data Engineering at Snowflake. Welcome, Chris.

>> Thank you, Rebecca. Glad to be here.

>> And Ben Kus, Chief Technology Officer at Box. Thank you so much for coming on theCUBE, Ben.

>> Thanks for having me.

>> I'm going to actually start with you, and we're going to start with the basics in terms of what are the kinds of data that Box customers typically work with and what have been their challenges in terms of unlocking insights from that data?

>> Yeah. If you think about Box, we are like the unstructured data platform that many of our customers rely on to store almost all of their unstructured data. And when we say unstructured data, we're thinking about things like key documents, things like contracts, things like their plans, their records, videos, images just across the different industries, many different examples. And one of the challenges has always been that this data was incredibly important and useful. I mean, in most organizations, it's the vast majority of your data so it's like 90% of your data is typically unstructured compared to the 10% that is structured. But usually the 90% people, it required a person to go sit down and actually use it. And unlike some of the enhancements over the last 10 plus years in data, unstructured data didn't have a technology revolution that helped people automate anything associated with it because it was too human oriented. Now, with generative AI, it unlocks a lot of the idea that generative AI was born on unstructured data, of course. It can not only help you understand it, but then also help you do things like pull data, extract data from it, answer questions about it, and so on. So this idea of generative AI, helping you understand and find insights in your data is really a critical thing that many companies are looking at now. Most enterprises are thinking about how can I get better use out of some of my most valuable assets to do things like just interact with them better in addition to pulling out the structured data so that they could then use it and tools like Snowflake and get all the benefits of structured data associated with it.

>> Okay, so Chris, I want to ask you, for the people who may not be aware with it, what exactly is Snowflake Openflow and what problem is it trying to solve?

>> Openflow is a new set of capabilities that we just announced really today that we're super excited about, that are about helping you bring data from anywhere into Snowflake so that you can do things with it. This applies to structured data, you can connect it to databases, to SaaS applications really to anything, pull the data out generally in a change data streaming fashion and stream that data into Snowflake to use. But it also works with unstructured data, which is a really big advancement. And as Ben was saying, unstructured data has for the first time really gotten unlocked by large language models just really in the last few years. We're seeing a ton of our customers want to bring together the structured data that they have, who are their customers, what are they doing, what activities have they done with their unstructured data? Maybe that's the contracts that they've signed with the customer, maybe it's some images and videos that they have about that. It can be all kinds of different things. And in the past, you've kept these in two different systems. With Openflow now, you can bring the data from Box or other unstructured places along with your structured data, bring it together in Snowflake and start getting value out of it and using it in new and really exciting ways.

>> Okay, so this really is pretty exciting in the sense of what companies can do with all of its vast institutional knowledge, which as you said, it could be contracts, it could be PowerPoint decks, and it could be the actual structured data that we're all familiar with. Then what made Box want to team up with Snowflake on Openflow?

>> Yeah, so one of the key things that we do at Box is we have made this data extraction capability part of our product offering, something we call Box AI. And so in many ways, the way that it works is you just basically give a schema and say, "I want Box to pull out this data with these columns and be able to have it the AI figure out your information." For instance, like contracts that you mentioned, we have customers who have a million contracts and they want to go through and figure out who signed it, what's the effective date, what are the key terms, and so on and so on. And so then we show them how to do this and then immediately their first question is, "This is great, this is what I've always wanted to, what do I do with the data though?" And so at Box we're like, "Okay, well, you can sort and filter and do basic stuff with it." But they're like, "No, no, I need to then combine it with other data. I need to be able to do advanced machine learning operations on it. I need to be able to do advanced analytics. I need to make it." And they'll point out, they're like, "I don't even want Box to do this, even if it could because I use Box, my unstructured data store, but I need to be able to move it somehow and make it kept in sync with your structured data store." And so then this Openflow announcement is just exactly what customers are asking for because it just makes it very easy for them to say, "Now that I've unlocked the structured data out of my unstructured data, then I can then operate on it," in the ways that they have grown accustomed to using a very powerful platform.

>> Chris, this is a big launch, this is a big partnership. How does it fit in with Snowflake's broader strategy of helping customers, as Ben was just saying, work with their data no matter where it lives?

>> So really, what we hear from our customers is that they have data in a lot of different places and they want to bring it together to make better business decisions. That's ultimately our goal and our strategy. Whether that data lives in an on-premise database underpinning your ERP, we announced a partnership also with Oracle where we can pull data out of those types of situations or it lives in somewhere like Box or it lives in Snowflake itself. We want you to be able to join that all together to be able to answer questions quickly and efficiently. And so Openflow for us came out of a lot of feedback we got from customers that getting data out of these systems, particularly on-premise systems, unstructured data systems and others, it was hard. It required managing infrastructure, it required stitching together a lot of different tools. And so with Openflow, we're breaking that barrier down and making all of that data accessible to Snowflake so that you can now ask questions of it in a way that brings it all together and hopefully helps you make better decisions in a much faster way than you've ever been able to before.

>> Okay. Ben, let's talk about this. How can give me a use case here of a customer who is now able to work with unstructured data in this way, making better, smarter, faster decisions, and how it can help AI unlock many more insights?

>> There's a ton of different examples, and many of them are things that used to require these manual efforts along the way. I had talked to a customer yesterday and they had all of these client files. This is a financial services firm, and they worked with their clients, high net worth clients. They had all this information about each of their clients, and that they constantly shared with them, they constantly talked about. But many times, when the financial advisor sat down to talk to their individual, the client, they had to go reread and remember all these different pieces. They sometimes hire some sort of specific process to go in and try to take the data together and figure out a way to create some assets for their client. But in the world of being able to first take the data out of it, the structured data out of it, like the boxes, data extraction, they could then feed the data into a structured data store and then be able to do things where they're able to, let's say, do high-scale data process to say, these are the kind of clients that we use and they have these can predict some things that the customer might be interested in or they can try to go through and be able to figure out if there's something missing about this client compared to their other customers and that kind of thing, you don't really do if you have a variety of unstructured data, it's the kind of thing you do when you're looking at data holistically, which is exactly the kind of thing that customers want to do with their structured data.

>> It's the kind of work that we do all every day and the average knowledge worker does every day. That's tedious and time-consuming and that isn't the kind of higher-order thinking that you're talking about there with being able to make recommendations and see where there might be gaps, absolutely.

>> Correct.

>> Okay. So this unstructured data opportunity is massive, but it's really been a tough nut to crack so far. So how does Openflow help unlock these insights more quickly and get it flowing into AI models?

>> It's really, Openflow helps get the data flowing into Snowflake, and then that allows you to unlock that data with things like our Cortex tools. So Cortex allows you to bring a bunch of different large language model-based tools in to ask questions about specific documents similar to what Ben's been describing, but also ask questions across a wide variety of documents. You can ask things like similar to the example that you were just giving. Let's say I've got thousands of clients and I want to ask questions about, "Hey, which are my clients who have this type of clause in their contract? Which are my clients where we've actually updated their contract within the last three years?" If you haven't gone through ahead of time and structured that data, those types of questions have been really, really hard to ask. And then it gets more complicated where you say, "Hey, I want to actually join that together. I'm really interested in the clients who have updated their contract within the last three years and have earned me at least this much money." Some of that structured data that lives in Snowflake, some of that's contract data that lives in Box, asking that type of question in the past required a person to go read through all of those contracts and annotate what was happening. So now, with the power of Cortex plus Openflow to bring that data together, I can now actually type that question in and get a list of the customers who meet that criteria, and then go iterate and move a lot faster than I have been able to before.

>> One of the things we're hearing a lot about Ben here at the Snowflake Summit and particularly with Openflow, is about this emphasis on open standards and reducing vendor lock-in. I'm curious your perspective on this strategy and how you make sure that the sensitive data stays within the confines for just that customer.

>> Yeah, so this is critical I think across all data, no matter what type it is, which is this is some of the most valuable assets that an organization has. Even though AI is very powerful, you can't do AI as something if you're going to risk your data security. If it turns into data leakage, then that is unnegotiable by almost all organizations. In order to be able to use AI effectively, you have to have platforms that use AI effectively, and you have to make sure that you have open platforms that let you do things where you can, no company wants to be locked into any one single platform that doesn't actually let you get your data in and out effectively. Something like Openflow not only allows us at Box with our unstructured data platform to move data back and forth with a structured data platform, but also it is an indicator to our customers and to Snowflake customers that we take very seriously that we can keep and hold their data not just to be used inside of our platform environments, but also in the bigger ecosystem. And as AI becomes more and more important, being able to securely use the same data across the ecosystem becomes more and more important because agents are more and more capable. So you have to rely on your secure platforms and bet on the fact that they're going to stay open and then also allow you to do AI on them.

>> You said something there that really resonated with me because we've always thought about, I think ourselves as playing a similar role to Box in that within our different data ecosystems, being able to govern your data and understand your data is one of the strengths that we really bring in the same way that you do. What we've seen now with AI is it's basically impossible to run AI models against data if you don't know what's in there, because now you don't know exactly what the output's going to be and whether you can use it the way that you want, and so you have to build it in. Governance has to exist at the layer of the data that you're bringing in at the layer of all the transformations you're doing, and then you can extend that into the AI models. And if you don't have that, you really can't use it in an enterprise environment at all.

>> Definitely. AI models, they don't keep secrets, and so whatever you present them with, they will happily tell whoever has access. The last thing that you want is for a user who doesn't have access to some, you don't want them to have access to some data to talk to an AI agent or AI model that has access to that data. And in order to solve that, it's a really hard problem. That is exactly the kind of problem that data platforms handle. For many people, they realize this fact ahead of time, but for people who didn't realize this, they often end up in a world where they're surprised at how the AI can quickly, if they use not the right platform, turn into a big challenge for them.

>> Okay. Final question for both of you. I want to throw it forward a little bit. Ben, how does Openflow set the stage for Snowflake's data engineering roadmap?

>> We really think of it as in order for people to get value out of their data, they need to get it somewhere that they can use it and then turn it into a form that's useful. And then you can start to do things like analysis or run AI or build applications on top of it. We've always relied on people, our customers, getting the data into a way that Snowflake can use it on their own. A lot of people built custom code to do that. There's a variety of partners that popped up to help with different pieces of it, but Openflow is really about making it easy and seamless to at scale in a highly governed way with the right rules applied, bring data from those systems into a place where you can do something useful with it. That could be Snowflake, that could be Iceberg, it actually even works where it can stream the data to other places as well. So it's not just us, but we want to make it very easy to get the data moving is the easiest way to think about it, which is where the flow part comes from. Then once you've got it landed, then you need to transform it and uplevel it and combine it with other data sets to turn it into something valuable, which is where the data engineering part comes in. We've got capabilities to do that by writing Python code, by writing SQL. We announced DBT projects, which allows you to use their templating engine on top of it, but it's really about making it easy to get the data moving, turn it into a format that's really useful, so then your teams can start using it to generate insights and really drive your business forward. We see this as honestly a critical part of continuing to make that whole journey a lot easier.

>> And Ben, how about you? What do you see as the biggest opportunity for Box customers to generate those insights?

>> I think for all Box customers, but frankly for every enterprise in the world right now, what they need to do is be able to get their platforms in shape. I was just talking to a customer and they were talking, somebody asked them, "What are your AI challenges?" And they said, "I don't have an AI challenge. I have a data challenge and I need to get all my data and trusted platforms that are secure so that I can then turn around and use the existing capabilities that are coming out for AI right now." And if you have it spread out, siloed, unprotected, then you just can't get the benefits that everybody is seeing emerging from these new powerful, disruptive technologies.

>> Well, Ben, Chris, thank you both so much for a really interesting conversation.

>> Thank you very much.

>> I'm Rebecca Knight, and that wraps up day one of theCUBE's live coverage of the Snowflake Summit. Catch you here tomorrow same time, same place. You're watching theCUBE, the leader in enterprise tech news and analysis.