Snowflake Summit 2026 | Craig Kerstiens, Snowflake & Jon Sweet & Nickhil Tekwani, OpenGov

Clips
More from Snowflake Summit 2026

Craig Kerstiens

Director of Snowflake Engineering, Postgres

Snowflake

Jon Sweet

SVP of Operations

OpenGov

Nickhil Tekwani

Senior Manager of Applied Ai

OpenGov

play_circle_outline OpenGov AI-powered public service platform for government ERP, budgeting, procurement, 311, infrastructure.

play_circle_outline Snowflake Acquires Crunchy Data: Embedding Enterprise Postgres to Unite Transactions and Analytics, Cutting ETL Latency

play_circle_outline Building company knowledge graph on Postgres combining structured and unstructured data.

play_circle_outline Agent-first AI interactions needing performant context and semantic layer for real-time use.

play_circle_outline Postgres community contributions: open-sourcing pg_lake enabling Postgres to work with data lakes.

Info

Craig Kerstiens, Snowflake & Jon Sweet & Nickhil Tekwani, OpenGov

Craig Kerstiens

Director of Snowflake Engineering, Postgres Snowflake

Jon Sweet

SVP of Operations OpenGov

Nickhil Tekwani

Senior Manager of Applied Ai OpenGov

search

Dave Vellante

>> Welcome back to San Francisco everybody. I'm Dave Vellante. For Rebecca Knight, we're here at Moscone South. This is day two of our deep coverage, live coverage of Snowflake Summit 2026, or seventh year covering Snowflake Summit. Craig Kerstiens is here. He's the Director of Engineering for Postgres at Snowflake. Nickhil Tekwani is Senior Manager of Applied AI at OpenGov. And Jon Sweet is the SVP of operations at the same company, OpenGov guys. Thanks for coming on.

Craig Kerstiens

>> Thanks for having us.

Jon Sweet

>> Thanks for having us.

Nickhil Tekwani

>> Thanks for having us.

Dave Vellante

>> Jon, for those who don't know, OpenGov, you guys do some pretty cool stuff, ERP, cloud-based software. Help us understand why you started the company and what's it all about.

Jon Sweet

>> Yeah, absolutely. OpenGov's been around for about 12 years now. We serve 2,000 state and local government customers across the United States. And our public service and digital service offerings are actually touching about one in three people in the United States at this point. So over the past 12 years, have spread out a pretty broad footprint of how we can help deliver more effective and accountable government to US citizens and constituents. So where we're at right now is we have an AI-powered public service platform that can do everything from budget and procurement workflows for your internal finance accounting procurement teams to everything from providing 311 services and management of public infrastructure. And what we're super excited about is building alongside government customers. In government, it's all about trust. It's about establishing trust with your constituents and your community members. And so when we think about building software for our customers and building internal tooling, we're really focused on that as well. And we're super excited to be here today to talk to you about it.

Dave Vellante

>> Now, why would the governments around the country use your product versus some of the brand name, the big companies that everybody knows about? What's your unique value prop to them?

Jon Sweet

>> Yeah, great question. So not to go too far back, but there is ultimately a number of unique requirements for how things like government accounting works, for example. And so government accounting workflows need to adhere to specific regulations, and so our software has been really purpose built for governments. And so that gives us not only a leg up in how tailored our software solutions are out of the box, but we then pair that with real tight subject matter expertise internally when we go into implementations and partner with these customers. And we think that gives us a really unique advantage for ultimately meeting governments where they are right now as they're going through a technical transformation evolving from systems that are 20 or 30 years old.

Dave Vellante

>> All right, great. And Nickhil, before I get to the AI story, I want to ask Craig. You got Postgres in your title.

Craig Kerstiens

>> Yep.

Dave Vellante

>> Okay. Thought you guys were the database. Come on. Why are we talking about Postgres? No, I'm kidding. So take us through the logic of the strategy around Postgres. You guys made an acquisition last year around transactions. Of course, Postgres transaction database came out as the alternative, open source alternative to Oracle. That's how it sort of made its name. Obviously, very well respected, reliable. Take us through the strategy there.

Craig Kerstiens

>> Yeah. So I think the key is just bringing the world's most popular transactional database into the trusted AI data cloud. Postgres is very, very well-loved by developers, very, very feature rich in its capabilities. But historically, that lived in a separate silo from your analytics. And one of my colleagues joked that these are different teams that never show up in the same room, because I don't think they like each other. The ones managing your analytics and one managing your transactions. And yet, these have to coexist. How do you get your data from your transactional system into your analytical one to drive business value? Today, that is hung up with latency and complicated ETL pipelines. One of my colleagues joked that he thinks the E in ETL stands for evil, which I thought was incredibly accurate, because it's a necessary evil. We have to get data from our transactional into our analytical system to drive the business and to drive value, yet it's duct tape and WD-40 and glue. And with the acquisition of Crunchy Data last year, bringing in a enterprise grade trusted Postgres to unify data and analytics under one platform so that you can now start to do more with AI.

Dave Vellante

>> And just a point of clarification. So you got Crunchy, you made that acquisition. Several years ago, you announced Unistore. Snowflake struggled getting it out, but then now it's out. And so you've got ... How do those all fit together?

Craig Kerstiens

>> Yeah. So I think you now have choice, right? And I think of Unistore and hybrid tables, that's very much unified within Snowflake. So you can think about Snowflake on the analytics side where you have hybrid tables and you have interactive tables. I'm sorry, you have choice now within Snowflake where we're very specialized on the analytics, like that unified data platform. But in parallel, Postgres is its own ecosystem with its own community of developers that love it, its own rich set of extensions and ecosystem and tooling. And together, these both exist and you have choice of when do you want to use which.

Dave Vellante

>> All right, Nickhil, everybody wants to talk about AI. We've waited all this time to bring you into the conversation. So you've got a decades plus platform. New AI versus ML comes in. How are you thinking about applying AI? What's your mission? Take us through the strategy there.

Nickhil Tekwani

>> 100%. So when I think about our AI strategy as a company and really the stuff that we're trying to build, I think it all comes back to a lot of what Craig said, which is that we don't want to operate in silos. Our governments do not operate in silos, so there's no reason that the software that supports them should operate in silos. So really, what we're trying to do is build, or the use case that is really interesting to us with Postgres is we're actually trying to build a company knowledge graph using unstructured and structured data into on place. And when we think about the AI strategy, it all comes back to the data that funds that. So we need a simplified architecture. We need one place where we can store all of that. We need a comfortable developer experience that can support that. And we really want to make sure that we are supporting, like Craig said, all kinds of use cases from the most transactional level data to the analytical level data for these AI workflows.

Dave Vellante

>> And you are building that knowledge graph on top of Postgres, is that right? And it's sort of a purpose built?

Nickhil Tekwani

>> That's correct. We're purpose building the knowledge graph using our first-party and third-party data directly on top of Snowflake Postgres.

Dave Vellante

>> So knowledge graphs have been, they're not new, but everybody now is talking about knowledge graphs. And they've been popular in things like security for years. I think part of that is you get this wonderful expressiveness, but if you wanted to query the knowledge graph, you had to go back to these sort of arcane approaches. That's all changed. Why has that changed? Why is it so popular today? Why the ascendancy today of knowledge graph?

Nickhil Tekwani

>> That's a great question. So to us at OpenGov, it's all about injecting the right context at the right time. So when you think about the knowledge graph and having, again, all of your data from hundreds of different data sources in one place, I think why knowledge graph becomes so interesting now is that as an agent interacting, right? Previously, we had humans that maybe had to interact with this context. Now, as an agent that needs to interact with this content at runtime, it needs to be performant. We need to lower the amount of hubs. We need to not deal with age-old E for evil ETL architecture and pipelines to support that. We need to make sure that we're supporting an agent-first interaction model on top of this context and performant knowledge graphs such as the one we're building on Postgres, on Snowflake Postgres is 100% one of the best ways to do that.

Dave Vellante

>> And I just think just hearing this conversation, I think about bringing transactions and analytics together, as you were saying, Craig. I like to think of like Uber for all. I have riders, drivers, locations. I have the transaction that has to take place and all happens in real-time. And nirvana would be if we can actually run our systems like that. So how does that apply to governments? Do they want that real-time transaction and analytics system to where it's not strings that databases understands, but it's people, places, things, activities, processes? Is that where you're headed?

Jon Sweet

>> Short answer is absolutely. And if you pay attention to what's happening in state and local government, you'll actually see that there's a number. There's a real early adopter curve that's taking place in government the same way you see it taking place in the private sector. And so if you look at the city of Boston, they're very much in the early adopter curve of that. And so you have a CIO in Boston who's basically stood up and said, "We want that vision. We think we can basically blaze that trail for other governments to follow in our footsteps." And so again, we're excited to be working with them and trying to figure that out.

Dave Vellante

>> That's interesting, because I live outside of Boston and I remember it's probably about 10 years ago now I was invited to a mayoral meeting and they were trying to make Boston a smart city. You guys know Boston. It's laid out as a mess. The really big challenge they were taking on. So I'm glad to hear that they're leaning in as an early adopter of this. When I think about, again, coming back to this idea of a knowledge graph in the city, what kind of workloads and applications are these cities and governments driving? How is it enabling them to provide better service to customers? Can you talk to that?

Nickhil Tekwani

>> Yeah, that's a great question. And I'll say that's a plane that we're building as we fly it alongside our customers. I think the use cases that we're really finding value in as a first pass is internally taking all the product data, customer data and internal data and public data that we have and we're actually powering workflows that allow us to serve our customers better. So in the case where you may be a customer success manager or you may be a professional services implementation consultant and you may be able to take previously unstructured or disparate data and you are actually able to proactively flag when a customer might be experiencing pain points in their implementation and get ahead of the problem instead of be behind the problem. And that's one way that we're already realizing value from the knowledge graph is being able to push that proactively and serve our customers in an even better way and meet them where they are.

Dave Vellante

>> So Craig, explain how you guys partner, what role Snowflake plays generally but specifically with your sort of product area.

Craig Kerstiens

>> Yeah. So I think Postgres is well-loved by developers, right? It's a great foundation, it's solid, it's stable. It's 30 years old, right? What technologies are we talking about that are 30 years old? Usually, that's pretty boring in sale, but it's this great foundation for developers. Developers want Postgres. But as you were saying before, these are latency and pipelines, they're fragile and brittle and they're just painful. And we're not able to drive value when we have those latency and pipelines, when these are breaking all the time and we're maintaining them. And in the age of AI where we want to go faster, that latency of data has a massive impact, right? And so if we can get that from Postgres, your transactional system into your analytical system, then we can start to drive a lot of value. And that's really where we're focused is simplifying that experience. And in your analytical system, you want to get in there and extract value out of it. Your transactional system, you want it to be boring. I don't want to think about it. I want it to just work. And so my goal is the best day is when you don't hear from us and we don't hear from you because your database is doing exactly what you need and you're busy driving your AI agents to add value for your customers and for that local community on top of this data. So the best thing we can do is make this simple and easy and just fade away into the background.

Dave Vellante

>> And your product's cloud based, right?

Jon Sweet

>> Correct.

Dave Vellante

>> And you're not running on-prem, or can you?

Jon Sweet

>> We have some certain legacy offerings that are on-prem, but predominantly all our customers are cloud-based.

Dave Vellante

>> Yeah. And I know Postgres runs up that, so let's talk cloud.

Jon Sweet

>> Yeah.

Dave Vellante

>> So assuming you've got cloud infrastructure, let's choose AWS, whatever, it doesn't matter. And I'm envisioning this new sort of software stack that's emerging and I'd love to get your sort of input on this. So we're not throwing away the old. We've got to have the determinism of the existing ERP systems and we want to bring intelligence into that, and we can't just slap it on. We can't just plug into a vector database and expect amazing results. It's like a nice demo. Now, okay, what do we do with it? Okay. So you've got the analytics system, you've got the transaction systems, you make sure those work. You've got this sort of system of engagement, I'll call it, which used to be maybe the BI layer. Now, you're talking to your data, text to SQL, and I'm learning from that. And then at the very top, you've got this agent. You got the LLMs, but then they have this agent control framework. And in the middle, you've got this new thing, I'm going to call it the system of intelligence. And you guys are sort of poking at that with your context, ontology, words like that. And it's all about harmonizing the data, bringing in business process, underlying application knowledge. Of course, metadata, technical, operational metadata, all that munged together. So now, I have this intelligent digital twin, if I will, of my enterprise. Okay, this is a little futuristic, I know, but it seems to me that this is the promise of AI. So what I just described, is it a reasonable roadmap for what your customers are thinking about? Is it something that's technically feasible? How do you think about that?

Jon Sweet

>> Yeah, I think it's probably what every company's thinking about right now is the short answer. Most companies I talk to have an information management problem candidly. And so AI just takes that problem and blows it through the roof is the short answer. You're going to spend tens of thousands, hundreds of thousands and millions of dollars getting really poor results using AI if you do not have a trusted semantic layer and a trusted data layer. And the more you can put that in a single place, the more you're going to simplify the technical solution, so you can take teams that are actually focused on engineering solutions for internal teams and external customers.

Dave Vellante

>> But you've got to evolve that, right? So you're going to build your semantic layer, you're building a semantic layer. The LLM vendors, we'll see. They probably want to get into the act. So as technologists, how do you think about that? You just, "Hey, just build the best product that our customers love"? Is that sort of where you start? I mean, how do you sort through all that?

Nickhil Tekwani

>> Yeah, I think it's really going back to becoming a trusted partner for your customer. So when we think about governments and the people that we serve, the way in which we serve them is inherently secondary to the impact that we bring to them as a community. So we talk about bringing that value. So it's less about what is the technology layer that is supporting that intelligence for them, but it's about what is the impact of having that intelligence for our customers, both internally and for our government customers. I think it all comes back to that impact.

Dave Vellante

>> Now, are you guys, are you big committers to Postgres? What's your standing in the community? How do you play there?

Nickhil Tekwani

>> From a use case perspective, at least what we're looking at with the knowledge graph and these context layers, we are investing very heavily into Snowflake Postgres, but it's still to be determined how prevalent that'll be across the company.

Dave Vellante

>> Okay. And you guys, how do you work with the Postgres community? Are you committers? Are you-

Craig Kerstiens

>> We have a number of committers at Snowflake. I think Postgres is very interesting in that one company owns it, right?

Dave Vellante

>> Right.

Craig Kerstiens

>> It's defied the test of time as an open source database that this community just continues to evolve. We open source pg_lake and contributed that back in November, which turns Postgres into this ability to work with your data lake that normally transactional databases can't do. We have a number of committers on staff contributing directly to open source Postgres. We're betting big on Postgres. I joke that last year was the year of Postgres and the year before was the year of Postgres. And we're like five years on and the year of Postgres as the transactional database of the year, and we're very much wanting to continue advancing Postgres the open source database as that foundation and as that transactional layer. And then again, how do we tie those two together as one unified data platform? What we're seeing is enterprises don't want 20 vendors in this for all. They want one unified platform for simplicity. We have a new model every month now that we've got to now swap out. And is it better? Is it worse? When do we use it? But the key is now you have model choice with the data with context and you can move faster as a business with a lot of simplicity then.

Dave Vellante

>> And just to put a finer point on it, I want to make it clear, you're not forking Postgres.

Craig Kerstiens

>> Nope.

Dave Vellante

>> Right? You're adding value to Postgres. Yeah.

Craig Kerstiens

>> Yeah, we're contributing back. We are giving it away as open source, contributing, invested in that because we believe that's the right long-term bet.

Dave Vellante

>> So last question, thinking ahead. We're here at Snowflake Summit 2027. What would you like to be able to say? I'll go around the horn. And Craig, I'll give you the last word. What would you like to be able to say a year from now that you're not able to say today? Or maybe in this world, this crazy world, it's three months from now, but let's say a year from now.

Jon Sweet

>> I'll take the lead, I guess. So I alluded to it before, but I believe the problem we're trying to solve right now inside of OpenGov, which is at its core, I think an information management problem and our solution is a knowledge graph that takes something akin to a traditional intranet, which is internal offering of data, blending with all of the raw data we have across the enterprise application stack, bringing that together in one place to form a source of truth for humans and agents to do their job more effectively. We want to solve that problem and then we want to take that solution and make sure we're helping governments across America solve that problem for their constituents.

Dave Vellante

>> Great. Thank you.

Jon Sweet

>> And we think solving it for ourselves, like many other companies are going to try and solve it. We think until we know how to do that for ourselves with, obviously, Snowflake is a core part of that, we're not going to be effective in advising and partnering with governments as they go on that same journey.

Dave Vellante

>> Nickhil?

Nickhil Tekwani

>> Yeah. I don't think I could have put it better than Jon did. I think it goes back to being able to be the leaders and not just show ... When I think about a year from now, I want to be able to have shown we went from point A to point B and show that playbook and use that playbook to drive value for our customers and not just have it be a theoretical or something hidden to us, but something that we can in the open source manner and with that same spirit behind it, show and guide our partners and guide our customers in realizing the same value.

Dave Vellante

>> Craig, bring us home and answer either in the context of OpenGov or sort of generally for your Postgres initiative.

Craig Kerstiens

>> Yeah. I think generally, the number one thing we hear from our customers is it just works like, "This is the Postgres I wanted and now I have this unified platform." So for me a year from now, what I'm most excited about is more stories like this of hearing from our customers the time I spend with them, the interesting use cases, it's always fascinating. Before we came on, we were talking about one of our customers that is a snake surveyor in Canada that uses the geospatial capabilities. And so two days a week, he's out in the field surveying and catching snakes, and then he's using PostGIS and Postgres to catalog that. How do you get that job? And so for me, it's building that solid foundation as a product that gets out of your way and lets you drive that business outcome and benefit for users. So I just hope I have more stories just like that a year from now.

Dave Vellante

>> From AI to snake charmers. Guys, thanks so much for coming in theCUBE. Really appreciate your time.

Nickhil Tekwani

>> Thank you.

Craig Kerstiens

>> Thank you.

Dave Vellante

>> All right. Thank you for watching. Keep it right there. This is Dave Vellante. For Rebecca Knight, we're here at Snowflake Summit 2026 Live from Moscone. We'll be right back right after this short break.

Craig Kerstiens, Snowflake & Jon Sweet & Nickhil Tekwani, OpenGov

search

Dave Vellante

>> Welcome back to San Francisco everybody. I'm Dave Vellante. For Rebecca Knight, we're here at Moscone South. This is day two of our deep coverage, live coverage of Snowflake Summit 2026, or seventh year covering Snowflake Summit. Craig Kerstiens is here. He's the Director of Engineering for Postgres at Snowflake. Nickhil Tekwani is Senior Manager of Applied AI at OpenGov. And Jon Sweet is the SVP of operations at the same company, OpenGov guys. Thanks for coming on.

Craig Kerstiens

>> Thanks for having us.

Jon Sweet

>> Thanks for having us.

Nickhil Tekwani

>> Thanks for having us.

Dave Vellante

>> Jon, for those who don't know, OpenGov, you guys do some pretty cool stuff, ERP, cloud-based software. Help us understand why you started the company and what's it all about.

Jon Sweet

>> Yeah, absolutely. OpenGov's been around for about 12 years now. We serve 2,000 state and local government customers across the United States. And our public service and digital service offerings are actually touching about one in three people in the United States at this point. So over the past 12 years, have spread out a pretty broad footprint of how we can help deliver more effective and accountable government to US citizens and constituents. So where we're at right now is we have an AI-powered public service platform that can do everything from budget and procurement workflows for your internal finance accounting procurement teams to everything from providing 311 services and management of public infrastructure. And what we're super excited about is building alongside government customers. In government, it's all about trust. It's about establishing trust with your constituents and your community members. And so when we think about building software for our customers and building internal tooling, we're really focused on that as well. And we're super excited to be here today to talk to you about it.

Dave Vellante

>> Now, why would the governments around the country use your product versus some of the brand name, the big companies that everybody knows about? What's your unique value prop to them?

Jon Sweet

>> Yeah, great question. So not to go too far back, but there is ultimately a number of unique requirements for how things like government accounting works, for example. And so government accounting workflows need to adhere to specific regulations, and so our software has been really purpose built for governments. And so that gives us not only a leg up in how tailored our software solutions are out of the box, but we then pair that with real tight subject matter expertise internally when we go into implementations and partner with these customers. And we think that gives us a really unique advantage for ultimately meeting governments where they are right now as they're going through a technical transformation evolving from systems that are 20 or 30 years old.

Dave Vellante

>> All right, great. And Nickhil, before I get to the AI story, I want to ask Craig. You got Postgres in your title.

Craig Kerstiens

>> Yep.

Dave Vellante

>> Okay. Thought you guys were the database. Come on. Why are we talking about Postgres? No, I'm kidding. So take us through the logic of the strategy around Postgres. You guys made an acquisition last year around transactions. Of course, Postgres transaction database came out as the alternative, open source alternative to Oracle. That's how it sort of made its name. Obviously, very well respected, reliable. Take us through the strategy there.

Craig Kerstiens

>> Yeah. So I think the key is just bringing the world's most popular transactional database into the trusted AI data cloud. Postgres is very, very well-loved by developers, very, very feature rich in its capabilities. But historically, that lived in a separate silo from your analytics. And one of my colleagues joked that these are different teams that never show up in the same room, because I don't think they like each other. The ones managing your analytics and one managing your transactions. And yet, these have to coexist. How do you get your data from your transactional system into your analytical one to drive business value? Today, that is hung up with latency and complicated ETL pipelines. One of my colleagues joked that he thinks the E in ETL stands for evil, which I thought was incredibly accurate, because it's a necessary evil. We have to get data from our transactional into our analytical system to drive the business and to drive value, yet it's duct tape and WD-40 and glue. And with the acquisition of Crunchy Data last year, bringing in a enterprise grade trusted Postgres to unify data and analytics under one platform so that you can now start to do more with AI.

Dave Vellante

>> And just a point of clarification. So you got Crunchy, you made that acquisition. Several years ago, you announced Unistore. Snowflake struggled getting it out, but then now it's out. And so you've got ... How do those all fit together?

Craig Kerstiens

>> Yeah. So I think you now have choice, right? And I think of Unistore and hybrid tables, that's very much unified within Snowflake. So you can think about Snowflake on the analytics side where you have hybrid tables and you have interactive tables. I'm sorry, you have choice now within Snowflake where we're very specialized on the analytics, like that unified data platform. But in parallel, Postgres is its own ecosystem with its own community of developers that love it, its own rich set of extensions and ecosystem and tooling. And together, these both exist and you have choice of when do you want to use which.

Dave Vellante

>> All right, Nickhil, everybody wants to talk about AI. We've waited all this time to bring you into the conversation. So you've got a decades plus platform. New AI versus ML comes in. How are you thinking about applying AI? What's your mission? Take us through the strategy there.

Nickhil Tekwani

>> 100%. So when I think about our AI strategy as a company and really the stuff that we're trying to build, I think it all comes back to a lot of what Craig said, which is that we don't want to operate in silos. Our governments do not operate in silos, so there's no reason that the software that supports them should operate in silos. So really, what we're trying to do is build, or the use case that is really interesting to us with Postgres is we're actually trying to build a company knowledge graph using unstructured and structured data into on place. And when we think about the AI strategy, it all comes back to the data that funds that. So we need a simplified architecture. We need one place where we can store all of that. We need a comfortable developer experience that can support that. And we really want to make sure that we are supporting, like Craig said, all kinds of use cases from the most transactional level data to the analytical level data for these AI workflows.

Dave Vellante

>> And you are building that knowledge graph on top of Postgres, is that right? And it's sort of a purpose built?

Nickhil Tekwani

>> That's correct. We're purpose building the knowledge graph using our first-party and third-party data directly on top of Snowflake Postgres.

Dave Vellante

>> So knowledge graphs have been, they're not new, but everybody now is talking about knowledge graphs. And they've been popular in things like security for years. I think part of that is you get this wonderful expressiveness, but if you wanted to query the knowledge graph, you had to go back to these sort of arcane approaches. That's all changed. Why has that changed? Why is it so popular today? Why the ascendancy today of knowledge graph?

Nickhil Tekwani

>> That's a great question. So to us at OpenGov, it's all about injecting the right context at the right time. So when you think about the knowledge graph and having, again, all of your data from hundreds of different data sources in one place, I think why knowledge graph becomes so interesting now is that as an agent interacting, right? Previously, we had humans that maybe had to interact with this context. Now, as an agent that needs to interact with this content at runtime, it needs to be performant. We need to lower the amount of hubs. We need to not deal with age-old E for evil ETL architecture and pipelines to support that. We need to make sure that we're supporting an agent-first interaction model on top of this context and performant knowledge graphs such as the one we're building on Postgres, on Snowflake Postgres is 100% one of the best ways to do that.

Dave Vellante

>> And I just think just hearing this conversation, I think about bringing transactions and analytics together, as you were saying, Craig. I like to think of like Uber for all. I have riders, drivers, locations. I have the transaction that has to take place and all happens in real-time. And nirvana would be if we can actually run our systems like that. So how does that apply to governments? Do they want that real-time transaction and analytics system to where it's not strings that databases understands, but it's people, places, things, activities, processes? Is that where you're headed?

Jon Sweet

>> Short answer is absolutely. And if you pay attention to what's happening in state and local government, you'll actually see that there's a number. There's a real early adopter curve that's taking place in government the same way you see it taking place in the private sector. And so if you look at the city of Boston, they're very much in the early adopter curve of that. And so you have a CIO in Boston who's basically stood up and said, "We want that vision. We think we can basically blaze that trail for other governments to follow in our footsteps." And so again, we're excited to be working with them and trying to figure that out.

Dave Vellante

>> That's interesting, because I live outside of Boston and I remember it's probably about 10 years ago now I was invited to a mayoral meeting and they were trying to make Boston a smart city. You guys know Boston. It's laid out as a mess. The really big challenge they were taking on. So I'm glad to hear that they're leaning in as an early adopter of this. When I think about, again, coming back to this idea of a knowledge graph in the city, what kind of workloads and applications are these cities and governments driving? How is it enabling them to provide better service to customers? Can you talk to that?

Nickhil Tekwani

>> Yeah, that's a great question. And I'll say that's a plane that we're building as we fly it alongside our customers. I think the use cases that we're really finding value in as a first pass is internally taking all the product data, customer data and internal data and public data that we have and we're actually powering workflows that allow us to serve our customers better. So in the case where you may be a customer success manager or you may be a professional services implementation consultant and you may be able to take previously unstructured or disparate data and you are actually able to proactively flag when a customer might be experiencing pain points in their implementation and get ahead of the problem instead of be behind the problem. And that's one way that we're already realizing value from the knowledge graph is being able to push that proactively and serve our customers in an even better way and meet them where they are.

Dave Vellante

>> So Craig, explain how you guys partner, what role Snowflake plays generally but specifically with your sort of product area.

Craig Kerstiens

>> Yeah. So I think Postgres is well-loved by developers, right? It's a great foundation, it's solid, it's stable. It's 30 years old, right? What technologies are we talking about that are 30 years old? Usually, that's pretty boring in sale, but it's this great foundation for developers. Developers want Postgres. But as you were saying before, these are latency and pipelines, they're fragile and brittle and they're just painful. And we're not able to drive value when we have those latency and pipelines, when these are breaking all the time and we're maintaining them. And in the age of AI where we want to go faster, that latency of data has a massive impact, right? And so if we can get that from Postgres, your transactional system into your analytical system, then we can start to drive a lot of value. And that's really where we're focused is simplifying that experience. And in your analytical system, you want to get in there and extract value out of it. Your transactional system, you want it to be boring. I don't want to think about it. I want it to just work. And so my goal is the best day is when you don't hear from us and we don't hear from you because your database is doing exactly what you need and you're busy driving your AI agents to add value for your customers and for that local community on top of this data. So the best thing we can do is make this simple and easy and just fade away into the background.

Dave Vellante

>> And your product's cloud based, right?

Jon Sweet

>> Correct.

Dave Vellante

>> And you're not running on-prem, or can you?

Jon Sweet

>> We have some certain legacy offerings that are on-prem, but predominantly all our customers are cloud-based.

Dave Vellante

>> Yeah. And I know Postgres runs up that, so let's talk cloud.

Jon Sweet

>> Yeah.

Dave Vellante

>> So assuming you've got cloud infrastructure, let's choose AWS, whatever, it doesn't matter. And I'm envisioning this new sort of software stack that's emerging and I'd love to get your sort of input on this. So we're not throwing away the old. We've got to have the determinism of the existing ERP systems and we want to bring intelligence into that, and we can't just slap it on. We can't just plug into a vector database and expect amazing results. It's like a nice demo. Now, okay, what do we do with it? Okay. So you've got the analytics system, you've got the transaction systems, you make sure those work. You've got this sort of system of engagement, I'll call it, which used to be maybe the BI layer. Now, you're talking to your data, text to SQL, and I'm learning from that. And then at the very top, you've got this agent. You got the LLMs, but then they have this agent control framework. And in the middle, you've got this new thing, I'm going to call it the system of intelligence. And you guys are sort of poking at that with your context, ontology, words like that. And it's all about harmonizing the data, bringing in business process, underlying application knowledge. Of course, metadata, technical, operational metadata, all that munged together. So now, I have this intelligent digital twin, if I will, of my enterprise. Okay, this is a little futuristic, I know, but it seems to me that this is the promise of AI. So what I just described, is it a reasonable roadmap for what your customers are thinking about? Is it something that's technically feasible? How do you think about that?

Jon Sweet

>> Yeah, I think it's probably what every company's thinking about right now is the short answer. Most companies I talk to have an information management problem candidly. And so AI just takes that problem and blows it through the roof is the short answer. You're going to spend tens of thousands, hundreds of thousands and millions of dollars getting really poor results using AI if you do not have a trusted semantic layer and a trusted data layer. And the more you can put that in a single place, the more you're going to simplify the technical solution, so you can take teams that are actually focused on engineering solutions for internal teams and external customers.

Dave Vellante

>> But you've got to evolve that, right? So you're going to build your semantic layer, you're building a semantic layer. The LLM vendors, we'll see. They probably want to get into the act. So as technologists, how do you think about that? You just, "Hey, just build the best product that our customers love"? Is that sort of where you start? I mean, how do you sort through all that?

Nickhil Tekwani

>> Yeah, I think it's really going back to becoming a trusted partner for your customer. So when we think about governments and the people that we serve, the way in which we serve them is inherently secondary to the impact that we bring to them as a community. So we talk about bringing that value. So it's less about what is the technology layer that is supporting that intelligence for them, but it's about what is the impact of having that intelligence for our customers, both internally and for our government customers. I think it all comes back to that impact.

Dave Vellante

>> Now, are you guys, are you big committers to Postgres? What's your standing in the community? How do you play there?

Nickhil Tekwani

>> From a use case perspective, at least what we're looking at with the knowledge graph and these context layers, we are investing very heavily into Snowflake Postgres, but it's still to be determined how prevalent that'll be across the company.

Dave Vellante

>> Okay. And you guys, how do you work with the Postgres community? Are you committers? Are you-

Craig Kerstiens

>> We have a number of committers at Snowflake. I think Postgres is very interesting in that one company owns it, right?

Dave Vellante

>> Right.

Craig Kerstiens

>> It's defied the test of time as an open source database that this community just continues to evolve. We open source pg_lake and contributed that back in November, which turns Postgres into this ability to work with your data lake that normally transactional databases can't do. We have a number of committers on staff contributing directly to open source Postgres. We're betting big on Postgres. I joke that last year was the year of Postgres and the year before was the year of Postgres. And we're like five years on and the year of Postgres as the transactional database of the year, and we're very much wanting to continue advancing Postgres the open source database as that foundation and as that transactional layer. And then again, how do we tie those two together as one unified data platform? What we're seeing is enterprises don't want 20 vendors in this for all. They want one unified platform for simplicity. We have a new model every month now that we've got to now swap out. And is it better? Is it worse? When do we use it? But the key is now you have model choice with the data with context and you can move faster as a business with a lot of simplicity then.

Dave Vellante

>> And just to put a finer point on it, I want to make it clear, you're not forking Postgres.

Craig Kerstiens

>> Nope.

Dave Vellante

>> Right? You're adding value to Postgres. Yeah.

Craig Kerstiens

>> Yeah, we're contributing back. We are giving it away as open source, contributing, invested in that because we believe that's the right long-term bet.

Dave Vellante

>> So last question, thinking ahead. We're here at Snowflake Summit 2027. What would you like to be able to say? I'll go around the horn. And Craig, I'll give you the last word. What would you like to be able to say a year from now that you're not able to say today? Or maybe in this world, this crazy world, it's three months from now, but let's say a year from now.

Jon Sweet

>> I'll take the lead, I guess. So I alluded to it before, but I believe the problem we're trying to solve right now inside of OpenGov, which is at its core, I think an information management problem and our solution is a knowledge graph that takes something akin to a traditional intranet, which is internal offering of data, blending with all of the raw data we have across the enterprise application stack, bringing that together in one place to form a source of truth for humans and agents to do their job more effectively. We want to solve that problem and then we want to take that solution and make sure we're helping governments across America solve that problem for their constituents.

Dave Vellante

>> Great. Thank you.

Jon Sweet

>> And we think solving it for ourselves, like many other companies are going to try and solve it. We think until we know how to do that for ourselves with, obviously, Snowflake is a core part of that, we're not going to be effective in advising and partnering with governments as they go on that same journey.

Dave Vellante

>> Nickhil?

Nickhil Tekwani

>> Yeah. I don't think I could have put it better than Jon did. I think it goes back to being able to be the leaders and not just show ... When I think about a year from now, I want to be able to have shown we went from point A to point B and show that playbook and use that playbook to drive value for our customers and not just have it be a theoretical or something hidden to us, but something that we can in the open source manner and with that same spirit behind it, show and guide our partners and guide our customers in realizing the same value.

Dave Vellante

>> Craig, bring us home and answer either in the context of OpenGov or sort of generally for your Postgres initiative.

Craig Kerstiens

>> Yeah. I think generally, the number one thing we hear from our customers is it just works like, "This is the Postgres I wanted and now I have this unified platform." So for me a year from now, what I'm most excited about is more stories like this of hearing from our customers the time I spend with them, the interesting use cases, it's always fascinating. Before we came on, we were talking about one of our customers that is a snake surveyor in Canada that uses the geospatial capabilities. And so two days a week, he's out in the field surveying and catching snakes, and then he's using PostGIS and Postgres to catalog that. How do you get that job? And so for me, it's building that solid foundation as a product that gets out of your way and lets you drive that business outcome and benefit for users. So I just hope I have more stories just like that a year from now.

Dave Vellante

>> From AI to snake charmers. Guys, thanks so much for coming in theCUBE. Really appreciate your time.

Nickhil Tekwani

>> Thank you.

Craig Kerstiens

>> Thank you.

Dave Vellante

>> All right. Thank you for watching. Keep it right there. This is Dave Vellante. For Rebecca Knight, we're here at Snowflake Summit 2026 Live from Moscone. We'll be right back right after this short break.