In this interview from GraphTalk San Francisco, Chad Cloes, staff software engineer at Intuit, joins theCUBE's John Furrier to discuss how graph technology is becoming the connective layer that makes enterprise AI more accurate and explainable. Cloes explains that as large language models become commodities, the real differentiator is proprietary context and how data relates within an organization. He details Intuit's Security Knowledge and Insights Platform, built to replace sprawling SQL joins with graph relationships that inherently model connections between resources, projects and people, supporting compliance needs like SOX and IRS 7216.
Cloes breaks down how graph data science brings explainability to security attribution, giving auditors and developers clear lineage into how conclusions are reached. He shares how the shift to graph-based context helped Intuit drive mean time to remediate from days down to seconds, powered by automated queries with humans retained in the loop for escalation and judgment calls. Cloes also walks through how Intuit built an MCP server on top of its graph platform, letting developers query project context in natural language rather than manually stitching together data from disparate tools. From abstracting away underlying vendors like Wiz and Databricks to enabling domain-specific sub-graphs that connect into larger enterprise knowledge structures, Cloes makes the case that graph is fast becoming a prerequisite for production AI rather than an optional layer.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems.
Thanks for confirming your account. Now you can access theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems with this email address.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Data + AI: Turning Data Into Knowledge for Autonomous Systems. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Chad Cloes, Intuit
In this interview from GraphTalk San Francisco, Chad Cloes, staff software engineer at Intuit, joins theCUBE's John Furrier to discuss how graph technology is becoming the connective layer that makes enterprise AI more accurate and explainable. Cloes explains that as large language models become commodities, the real differentiator is proprietary context and how data relates within an organization. He details Intuit's Security Knowledge and Insights Platform, built to replace sprawling SQL joins with graph relationships that inherently model connections between resources, projects and people, supporting compliance needs like SOX and IRS 7216.
Cloes breaks down how graph data science brings explainability to security attribution, giving auditors and developers clear lineage into how conclusions are reached. He shares how the shift to graph-based context helped Intuit drive mean time to remediate from days down to seconds, powered by automated queries with humans retained in the loop for escalation and judgment calls. Cloes also walks through how Intuit built an MCP server on top of its graph platform, letting developers query project context in natural language rather than manually stitching together data from disparate tools. From abstracting away underlying vendors like Wiz and Databricks to enabling domain-specific sub-graphs that connect into larger enterprise knowledge structures, Cloes makes the case that graph is fast becoming a prerequisite for production AI rather than an optional layer.
>> Welcome back to theCUBE here in San Francisco for Neo4j's Graph Talk San Francisco. They have an event in New York City, which will be there as well in New York City in September. It's an event where all the practitioners, insiders, and also the innovators get together to talk about the innovations around graph databases, ontologies, but also the future of the infrastructure and data structures in AI that's really flourishing things into production. We're starting to see evidence now of the data stack, how it fits into the AI equation. Chad Cloes is here. He's a senior software staff engineer at Intuit. Those guys are well known for their open source work and innovation. Chad, thanks for coming on. Appreciate it.
Chad Cloes
>> Glad to be here. Thanks for having me.
John Furrier
>> First of all, love Intuit. I have friends who work there. Great Silicon Valley company, but also hat tip to Intuit for really being an end user contributor to open source. Linux Foundation, very active. And so props to you and your team.
Chad Cloes
>> Yeah, we're proud of what we've done and we're glad to be part of the community.
John Furrier
>> All right. So we're here at the Neo4j Graph Talk. This is where I've seen this early in the early agent side, agent infrastructure, where the real experts come together to share what they're working on. And graph databases and Neo4j in particular have had recent success just financially and on the business side, but also the standards bodies, you start to see graphs, a key ingredient data structure with knowledge graphs in the AI equation. Explain what's the rationale behind this? What's your perspective? Why is it happening? Is it just the moment in time, it's perfect timing, everything's clicking, or is there something else going on?
Chad Cloes
>> Well, I think it's piggybacking on all the inertia surrounding AI and how everybody's trying to make it relevant and capitalize on the ecosystem and/or the buzz and the bubble around AI in general. I think graphs already were very relevant from a, "How do I make my data valuable? How do I make it useful? How do I monetize and/or democratize that data?" And we have found at Intuit that the graph layer gives that contextualization. And oh, by the way, it's a byproduct. The byproduct of it is you can hand all that context that you've built with the graph to your LLM. What we found is that LLMs in general are all commodity now. You turn to one, one's a little bit ahead of the other. What differentiates your use of it is the context and/or the data and how it's related in your ecosystem so that when you ask the questions, it has the information that is relevant to your company.
John Furrier
>> Yeah. Philip and I were just riffing about graphs and talking about, he's like, "It's like 2D, 3D. It's like black and white in color." And I look at it simply as, well, there's grinding database work tables and schemas, and then there's freedom. He went more black and white in color. I went more grinding, freedom. What is your use case? Where did you take particular attention to on the graph insight Intuit? Was it security data? Explain your use case or where you leaned in and drilled into it.
Chad Cloes
>> So for our team specifically, we're with a SKIP team, Security Knowledge and Insights Platform. So the platform that we built was all around taking the silos of data and tools, security tools, or data swamps, data lakes, even if they were healthy, and taking that data and contextualizing and connecting it so that... The funny thing that Phil talked about in his talk is the 120 line SQL joins, right? Or we had some that were as long as 2,000. And the irony in those SQL statements is you're recreating the relationships between the tables and the silos in SQL. Well, graph, you don't have to do that. It's just inherent. I think Emil can be quoted as, "The relationships are first-class citizens. The relationship between your data is almost more important than the data itself." And we found that in that journey of connecting the data together, it slip streams straight into AI. And we predicted about a year ago that graph and/or our platform was going to be key to our AI adoption, and it truly has.
John Furrier
>> And it worked out that way.
Chad Cloes
>> It did.
John Furrier
>> Results did. What specifically are you doing? What's the outcome? What was the result?
Chad Cloes
>> Well, so our platform, it's all in service of doing security attribution. What resource is owned by what project, what team? How is it related to other things? For instance, we have to do SOX compliance or we have to do... 7216 is a dictate by the federal government for tax data. So having the relationships between the data and where that data is being used is relevant and is paramount for us to be able to track and make available to the company so they can make use of it.
John Furrier
>> So you're mapping resource to projects, people, places, things, those kinds of things. That's great for audits, but I mean, explainability is a huge issue right now on AI. Oh, sure, there's an upside there on explainability.
Chad Cloes
>> Yeah. So we actually have a property that said that we use graph data science or GDS that we explain, and the property is explainability. So when we do an attribution of a resource to a resource or a project or a person, it has that explainability. It says, "Well, this is the lineage that we manifested from our graph data science insight." And we give that to the customer so that they have that data as well on top of it.
John Furrier
>> Yeah. I mean, that's the holy grail right there. I mean, people love lineage, but when you actually have it inherently in the system as a first class citizen. Talk about production and reliability because you're seeing a lot of AI projects, some make it the production, some get it right, some don't. They get stuck in POC purgatory, some say, or just kind of just sit there on the one yard line. Can you push it in? And they don't make it. There's no governance. The identity's screwed up. When you're talking about security accountability, I mean, you can't screw that up. I mean, there's audits, there's also penalties.
Chad Cloes
>> That's right. Yes.
John Furrier
>> There's a lot of pressure. How did you get it into production? What's the secret sauce? How did you nail the reliability?
Chad Cloes
>> So if you're talking about reliability specifically, we do leverage the reliability of Aura. And knock on wood, we've been flawless in that regard. With regards to the reliability of the data, the reliability of the data evolves as you learn. And one of the things that we've found is that we don't know what we don't know. And we democratize and we just try and make it visible because the term I always use is, I don't make the news, I just report it. If we bring the data together and we make it available to people, then they can make their own determinations. And from a security perspective, bringing things to light really is the way to solve those hard issues.
John Furrier
>> So when you talk about authorization, governance, these are the table stakes we're hearing a lot in a lot of the AI conversations, especially around AI safety and whatnot, what's the timing impact that graphs give you? Is it weeks to days, weeks to months, months to weeks? What's the scope of some of the benefits in terms of when you look at the outcomes that you get from, say, figuring out what the root cause of that or this was?
Chad Cloes
>> Yeah. So one of the main tenants of what we attempt to do is we attempt to drive down MTTR, which is meantime to remediate. And you can't do that if you're spending all your time figuring out where or who or what it's connected to. And when we started, it would take days to do analysis on how things were connected. You have to log into seven different things. You had to have three different people with different credentials. Once you start drawing that data in and connecting it in a way that's relevant, you again, democratize the data such that it takes it from days to seconds.
John Furrier
>> Days to seconds.
Chad Cloes
>> We can actually run API calls against our system that immediately manifests this endpoint is associated with this project, which is associated with these resources, which is associated with these people.
John Furrier
>> Is there a human loop there or is there agents doing that? Because seconds means it's automation there.
Chad Cloes
>> Yeah, that's right. Yeah, so we do. We have automated systems. They're constantly querying and then manifesting the results to the consuming systems.
John Furrier
>> So the humans are curating or watching and managing, "Okay, keeping an eye on it." So the human is in the loop in the sense of, are they doing what they need to do?
Chad Cloes
>> Yes. So one of the systems that we use is the human will prompt and say, "Hey, I need information about these sets of assets or these endpoints." And then that data will get presented back to the person. We do have automated systems that say, "Okay, if it meets this criteria, this criteria is something that needs to be escalated and then it just automatically escalates.
John Furrier
>> You said Aurora. You guys use-
Chad Cloes
>> The Aura. The Neo4j Aura .
John Furrier
>> Okay. Not Aurora AWS?
Chad Cloes
>> Nope. Aura.
John Furrier
>> Okay. I want to make sure I clarify that.
Chad Cloes
>> Yeah.
John Furrier
>> All right. So this brings up the whole data ecosystem discussion. API cloud era was easy. Connect to an API, all good. But when you start getting into data transfer with agents and systems, the relationships of the vendors change. So how should people think about their data ecosystem platforms when they start bringing knowledge graphs in? It's a data inclusive environment. It's open.
Chad Cloes
>> Yeah.
John Furrier
>> Yeah.
Chad Cloes
>> So one of the things that we have prided ourself on is if you pull that data into, we use Neo4j, pull that data in a graph, you can abstract your system from whatever tool that the data lives in, Wiz or Databricks or an S3 file or any of that stuff. You create an abstraction layer that allows you to not be so reliant on the underlying tool or vendor, if that's what you're asking.
John Furrier
>> Yeah. So you can just ingest it.
Chad Cloes
>> Yeah.
John Furrier
>> And then manifest. Okay.
Chad Cloes
>> That's right.
John Furrier
>> In your world in security, root cause is a big deal. I know there's a lot of compliance too involved in if something happened, reports got to get spit out, all kinds of forensics happen. How much are you involved in that? Is that included in? Is it a full suite of like, "Okay, just agents go to town, press a button, metaphorically speaking"? Take us through.
Chad Cloes
>> Yeah. So on the security operations side, that is not our team. Our team is more about making the data available, ingesting it, contextualizing it. And then the SOC uses it for their type of forensics. We're definitely getting drawn into those areas. And one of the challenges that we're going to have to face is some of that data can be sensitive and so we need to segregate and/or make it available. In all of our cases from an AI perspective, there is human in the loop. Now like I said before, if it meets certain criteria, then it can be automatic.
John Furrier
>> So they're your customer basically. They're the customer, in a sense.
Chad Cloes
>> Well, yeah.
John Furrier
>> User.
Chad Cloes
>> Well, our customers range from just the developers saying, "Hey, I inherited this project. I don't know anything about it. Give me all the context associated with this GitHub repo or this endpoint or this asset or where are the AWS accounts?" So it really enables just-
John Furrier
>> All right, take me through that. I think that's a really good instructive. Let's just say I inherit the project. I need to look at, "Okay, open up the book. Here's my GitHub." So what plugs in? So what happens? What's my interface? What does it look like? Take me through that play by play.
Chad Cloes
>> Again, one of the things that we're proud of is that we have created this platform. And on top of it, we've created an API. And that GraphQL API lends itself nicely to creating MCP and skills. So in the AI world, we have created an MCP server on top of our data ingest and our APIs that we can get out within minutes. And then the MCP server, you build the tools that are relevant to the developer. And then every developer nowadays is using skills and/or MCP servers in their AI context. And you basically ask the AI, you give the AI, "This is the MCP server that has all the context associated with our company. Tell me about this particular piece." And then it will then spider out and it uses the MCP server that uses Neo4j behind the scenes to give that context so that the developer can just ask naturally.
John Furrier
>> "Give me a summary. I'm going to take my bike ride home. Give me a podcast."
Chad Cloes
>> Right. Exactly. Exactly.
John Furrier
>> That was a dream scenario eight years ago. Now it's happening.
Chad Cloes
>> It is.
John Furrier
>> All right. What's the coolest thing if you had to explain to a friend, tell me about the graphs. Why should I do it? What's the motivation? What's your experience with it? People seem to be jazzed up about graphs. GraphRAG seems to be getting that good of results. What's the pep talk or motivational speech to give someone? Because I think once people see it and touch it and taste it, they don't really go back.
Chad Cloes
>> Yeah. So I think Emil said once you see graphs and you start thinking your graphs, you see graphs everywhere. The discovery I think is probably the coolest thing. Being able to say, "Well, I don't know what I don't know. I'm just going to start ingesting it and it's going to evolve over time." And that evolution in that journey really is the fun part of being able to just stitch stuff together and see it grow over time. It really is a-
John Furrier
>> And the alignment with AI is pretty fantastic. And what's not really reported well, and I want to get your reaction on this, Ed Moore's comment, is that knowledge graphs align beautifully with computer science. If you look at AI, it really is a testament of the most alpha computer science wave we've hit since structured programming and punch cards went away. And graphs can recurse. You can run through them superfast. It's aligned with the neural network philosophy of AI and DeepMind and all the tools and all the great work that came out of the past decade. What's your reaction to that? I mean, that's probably the best motivation. It's how people think and work basically.
Chad Cloes
>> Yeah. I think the AI and the LLMs... I don't know if this analogy works or not, but I'm going to use it. AI and LLM, they're commodity. You're just buying them off the shelf now. Graph is the Nordstrom experience of that data, right? You put your graph in the context of what you're trying to do or your company or your specific use case. Even down to the personal level, like I think John was talking about in his talk, is he uses it for health. It really is the way to take that commodity and make it relevant and make it real for you.
John Furrier
>> Yeah, I mean LLMs are not the product anymore. They're just a part of the input, the system. It's funny, we were at the AMD event. We was doing a live stream there today too. And we were commenting about, "Oh, who's got the best GPU?" The real game is, "Is NVIDIA going to win? Is AMD? Intel?" Everyone, Cerebras is the new hot IPO." And I said to Dave, I said, "To me, it's like whoever can, in the fastest way possible, redefine computing to fit the way people think and work." And that's because that's the utility that's growing bottoms up from AI, which is it's not a GUI anymore. So it's not like SAS. It's a natural language interface. So the experience is going to dictate to the stack, which is growing. Every user's going to want it. So that changes the entire nature of this data stack completely and also changes the relationship for computing architecture.
Chad Cloes
>> Yeah. Frankly, I don't even care who wins because I'm going to be pulling data. I'm going to be pulling the data into a graph, making it relevant and real for me. And it's going to be ChatGPT today. It's going to be Amazon Bedrock tomorrow. It's going to be whatever. It doesn't really matter because I'm going to be able to contextualize and use it in the most important way.
John Furrier
>> Well, this is the best thing about it. General intelligence is the internet. They already got that covered. But if you look at a company like Intuit and others, there's domain expertise. There's specialty intelligence there. That's the real value. So there's going to be a lot of inference going on. So you don't need the mega models to do inference and cool things.
Chad Cloes
>> Well, even within the company, we find that there are silos of domain experts. Inside those sub-domains, you're going to have the smaller graphs and the smaller graphs are going to connect to larger graphs. And it really does-
John Furrier
>> I think that's where the models fit in. So again, it's funny, we predicted this three years ago, that the power law of models and the power law at the head is the popular ones. That's general. But as you move down, open source and specialty comes in. And again, mixture of experts is just mixture of domains. And that's where I think graphs highlights. Chad, great conversation. We're going to probably go another hour on graphs. Again, but the strategic importance of graphs is it's compatible with the future architecture of where AI's going with easy results and benefits that you can get out of data.
Chad Cloes
>> Yeah. In a lot of ways, I think it's a prerequisite.
John Furrier
>> Yeah. Great job. Thanks for coming on. I'm John Furrier with theCUBE here. Neo4j Graph Talk San Francisco. This is where all the top engineers and practitioners get together to share their results and best practice. And also new ways to use graphs, how to build that AI data layer, semantic layer, ontology. The new data engineering is going to be abstracted away with agents and of course the architecture and the data structures. Our all changing graphs are at the center of it. We're doing our best to bring that coverage to you. Thanks for watching.
>> Welcome back to theCUBE here in San Francisco for Neo4j's Graph Talk San Francisco. They have an event in New York City, which will be there as well in New York City in September. It's an event where all the practitioners, insiders, and also the innovators get together to talk about the innovations around graph databases, ontologies, but also the future of the infrastructure and data structures in AI that's really flourishing things into production. We're starting to see evidence now of the data stack, how it fits into the AI equation. Chad Cloes is here. He's a senior software staff engineer at Intuit. Those guys are well known for their open source work and innovation. Chad, thanks for coming on. Appreciate it.
Chad Cloes
>> Glad to be here. Thanks for having me.
John Furrier
>> First of all, love Intuit. I have friends who work there. Great Silicon Valley company, but also hat tip to Intuit for really being an end user contributor to open source. Linux Foundation, very active. And so props to you and your team.
Chad Cloes
>> Yeah, we're proud of what we've done and we're glad to be part of the community.
John Furrier
>> All right. So we're here at the Neo4j Graph Talk. This is where I've seen this early in the early agent side, agent infrastructure, where the real experts come together to share what they're working on. And graph databases and Neo4j in particular have had recent success just financially and on the business side, but also the standards bodies, you start to see graphs, a key ingredient data structure with knowledge graphs in the AI equation. Explain what's the rationale behind this? What's your perspective? Why is it happening? Is it just the moment in time, it's perfect timing, everything's clicking, or is there something else going on?
Chad Cloes
>> Well, I think it's piggybacking on all the inertia surrounding AI and how everybody's trying to make it relevant and capitalize on the ecosystem and/or the buzz and the bubble around AI in general. I think graphs already were very relevant from a, "How do I make my data valuable? How do I make it useful? How do I monetize and/or democratize that data?" And we have found at Intuit that the graph layer gives that contextualization. And oh, by the way, it's a byproduct. The byproduct of it is you can hand all that context that you've built with the graph to your LLM. What we found is that LLMs in general are all commodity now. You turn to one, one's a little bit ahead of the other. What differentiates your use of it is the context and/or the data and how it's related in your ecosystem so that when you ask the questions, it has the information that is relevant to your company.
John Furrier
>> Yeah. Philip and I were just riffing about graphs and talking about, he's like, "It's like 2D, 3D. It's like black and white in color." And I look at it simply as, well, there's grinding database work tables and schemas, and then there's freedom. He went more black and white in color. I went more grinding, freedom. What is your use case? Where did you take particular attention to on the graph insight Intuit? Was it security data? Explain your use case or where you leaned in and drilled into it.
Chad Cloes
>> So for our team specifically, we're with a SKIP team, Security Knowledge and Insights Platform. So the platform that we built was all around taking the silos of data and tools, security tools, or data swamps, data lakes, even if they were healthy, and taking that data and contextualizing and connecting it so that... The funny thing that Phil talked about in his talk is the 120 line SQL joins, right? Or we had some that were as long as 2,000. And the irony in those SQL statements is you're recreating the relationships between the tables and the silos in SQL. Well, graph, you don't have to do that. It's just inherent. I think Emil can be quoted as, "The relationships are first-class citizens. The relationship between your data is almost more important than the data itself." And we found that in that journey of connecting the data together, it slip streams straight into AI. And we predicted about a year ago that graph and/or our platform was going to be key to our AI adoption, and it truly has.
John Furrier
>> And it worked out that way.
Chad Cloes
>> It did.
John Furrier
>> Results did. What specifically are you doing? What's the outcome? What was the result?
Chad Cloes
>> Well, so our platform, it's all in service of doing security attribution. What resource is owned by what project, what team? How is it related to other things? For instance, we have to do SOX compliance or we have to do... 7216 is a dictate by the federal government for tax data. So having the relationships between the data and where that data is being used is relevant and is paramount for us to be able to track and make available to the company so they can make use of it.
John Furrier
>> So you're mapping resource to projects, people, places, things, those kinds of things. That's great for audits, but I mean, explainability is a huge issue right now on AI. Oh, sure, there's an upside there on explainability.
Chad Cloes
>> Yeah. So we actually have a property that said that we use graph data science or GDS that we explain, and the property is explainability. So when we do an attribution of a resource to a resource or a project or a person, it has that explainability. It says, "Well, this is the lineage that we manifested from our graph data science insight." And we give that to the customer so that they have that data as well on top of it.
John Furrier
>> Yeah. I mean, that's the holy grail right there. I mean, people love lineage, but when you actually have it inherently in the system as a first class citizen. Talk about production and reliability because you're seeing a lot of AI projects, some make it the production, some get it right, some don't. They get stuck in POC purgatory, some say, or just kind of just sit there on the one yard line. Can you push it in? And they don't make it. There's no governance. The identity's screwed up. When you're talking about security accountability, I mean, you can't screw that up. I mean, there's audits, there's also penalties.
Chad Cloes
>> That's right. Yes.
John Furrier
>> There's a lot of pressure. How did you get it into production? What's the secret sauce? How did you nail the reliability?
Chad Cloes
>> So if you're talking about reliability specifically, we do leverage the reliability of Aura. And knock on wood, we've been flawless in that regard. With regards to the reliability of the data, the reliability of the data evolves as you learn. And one of the things that we've found is that we don't know what we don't know. And we democratize and we just try and make it visible because the term I always use is, I don't make the news, I just report it. If we bring the data together and we make it available to people, then they can make their own determinations. And from a security perspective, bringing things to light really is the way to solve those hard issues.
John Furrier
>> So when you talk about authorization, governance, these are the table stakes we're hearing a lot in a lot of the AI conversations, especially around AI safety and whatnot, what's the timing impact that graphs give you? Is it weeks to days, weeks to months, months to weeks? What's the scope of some of the benefits in terms of when you look at the outcomes that you get from, say, figuring out what the root cause of that or this was?
Chad Cloes
>> Yeah. So one of the main tenants of what we attempt to do is we attempt to drive down MTTR, which is meantime to remediate. And you can't do that if you're spending all your time figuring out where or who or what it's connected to. And when we started, it would take days to do analysis on how things were connected. You have to log into seven different things. You had to have three different people with different credentials. Once you start drawing that data in and connecting it in a way that's relevant, you again, democratize the data such that it takes it from days to seconds.
John Furrier
>> Days to seconds.
Chad Cloes
>> We can actually run API calls against our system that immediately manifests this endpoint is associated with this project, which is associated with these resources, which is associated with these people.
John Furrier
>> Is there a human loop there or is there agents doing that? Because seconds means it's automation there.
Chad Cloes
>> Yeah, that's right. Yeah, so we do. We have automated systems. They're constantly querying and then manifesting the results to the consuming systems.
John Furrier
>> So the humans are curating or watching and managing, "Okay, keeping an eye on it." So the human is in the loop in the sense of, are they doing what they need to do?
Chad Cloes
>> Yes. So one of the systems that we use is the human will prompt and say, "Hey, I need information about these sets of assets or these endpoints." And then that data will get presented back to the person. We do have automated systems that say, "Okay, if it meets this criteria, this criteria is something that needs to be escalated and then it just automatically escalates.
John Furrier
>> You said Aurora. You guys use-
Chad Cloes
>> The Aura. The Neo4j Aura .
John Furrier
>> Okay. Not Aurora AWS?
Chad Cloes
>> Nope. Aura.
John Furrier
>> Okay. I want to make sure I clarify that.
Chad Cloes
>> Yeah.
John Furrier
>> All right. So this brings up the whole data ecosystem discussion. API cloud era was easy. Connect to an API, all good. But when you start getting into data transfer with agents and systems, the relationships of the vendors change. So how should people think about their data ecosystem platforms when they start bringing knowledge graphs in? It's a data inclusive environment. It's open.
Chad Cloes
>> Yeah.
John Furrier
>> Yeah.
Chad Cloes
>> So one of the things that we have prided ourself on is if you pull that data into, we use Neo4j, pull that data in a graph, you can abstract your system from whatever tool that the data lives in, Wiz or Databricks or an S3 file or any of that stuff. You create an abstraction layer that allows you to not be so reliant on the underlying tool or vendor, if that's what you're asking.
John Furrier
>> Yeah. So you can just ingest it.
Chad Cloes
>> Yeah.
John Furrier
>> And then manifest. Okay.
Chad Cloes
>> That's right.
John Furrier
>> In your world in security, root cause is a big deal. I know there's a lot of compliance too involved in if something happened, reports got to get spit out, all kinds of forensics happen. How much are you involved in that? Is that included in? Is it a full suite of like, "Okay, just agents go to town, press a button, metaphorically speaking"? Take us through.
Chad Cloes
>> Yeah. So on the security operations side, that is not our team. Our team is more about making the data available, ingesting it, contextualizing it. And then the SOC uses it for their type of forensics. We're definitely getting drawn into those areas. And one of the challenges that we're going to have to face is some of that data can be sensitive and so we need to segregate and/or make it available. In all of our cases from an AI perspective, there is human in the loop. Now like I said before, if it meets certain criteria, then it can be automatic.
John Furrier
>> So they're your customer basically. They're the customer, in a sense.
Chad Cloes
>> Well, yeah.
John Furrier
>> User.
Chad Cloes
>> Well, our customers range from just the developers saying, "Hey, I inherited this project. I don't know anything about it. Give me all the context associated with this GitHub repo or this endpoint or this asset or where are the AWS accounts?" So it really enables just-
John Furrier
>> All right, take me through that. I think that's a really good instructive. Let's just say I inherit the project. I need to look at, "Okay, open up the book. Here's my GitHub." So what plugs in? So what happens? What's my interface? What does it look like? Take me through that play by play.
Chad Cloes
>> Again, one of the things that we're proud of is that we have created this platform. And on top of it, we've created an API. And that GraphQL API lends itself nicely to creating MCP and skills. So in the AI world, we have created an MCP server on top of our data ingest and our APIs that we can get out within minutes. And then the MCP server, you build the tools that are relevant to the developer. And then every developer nowadays is using skills and/or MCP servers in their AI context. And you basically ask the AI, you give the AI, "This is the MCP server that has all the context associated with our company. Tell me about this particular piece." And then it will then spider out and it uses the MCP server that uses Neo4j behind the scenes to give that context so that the developer can just ask naturally.
John Furrier
>> "Give me a summary. I'm going to take my bike ride home. Give me a podcast."
Chad Cloes
>> Right. Exactly. Exactly.
John Furrier
>> That was a dream scenario eight years ago. Now it's happening.
Chad Cloes
>> It is.
John Furrier
>> All right. What's the coolest thing if you had to explain to a friend, tell me about the graphs. Why should I do it? What's the motivation? What's your experience with it? People seem to be jazzed up about graphs. GraphRAG seems to be getting that good of results. What's the pep talk or motivational speech to give someone? Because I think once people see it and touch it and taste it, they don't really go back.
Chad Cloes
>> Yeah. So I think Emil said once you see graphs and you start thinking your graphs, you see graphs everywhere. The discovery I think is probably the coolest thing. Being able to say, "Well, I don't know what I don't know. I'm just going to start ingesting it and it's going to evolve over time." And that evolution in that journey really is the fun part of being able to just stitch stuff together and see it grow over time. It really is a-
John Furrier
>> And the alignment with AI is pretty fantastic. And what's not really reported well, and I want to get your reaction on this, Ed Moore's comment, is that knowledge graphs align beautifully with computer science. If you look at AI, it really is a testament of the most alpha computer science wave we've hit since structured programming and punch cards went away. And graphs can recurse. You can run through them superfast. It's aligned with the neural network philosophy of AI and DeepMind and all the tools and all the great work that came out of the past decade. What's your reaction to that? I mean, that's probably the best motivation. It's how people think and work basically.
Chad Cloes
>> Yeah. I think the AI and the LLMs... I don't know if this analogy works or not, but I'm going to use it. AI and LLM, they're commodity. You're just buying them off the shelf now. Graph is the Nordstrom experience of that data, right? You put your graph in the context of what you're trying to do or your company or your specific use case. Even down to the personal level, like I think John was talking about in his talk, is he uses it for health. It really is the way to take that commodity and make it relevant and make it real for you.
John Furrier
>> Yeah, I mean LLMs are not the product anymore. They're just a part of the input, the system. It's funny, we were at the AMD event. We was doing a live stream there today too. And we were commenting about, "Oh, who's got the best GPU?" The real game is, "Is NVIDIA going to win? Is AMD? Intel?" Everyone, Cerebras is the new hot IPO." And I said to Dave, I said, "To me, it's like whoever can, in the fastest way possible, redefine computing to fit the way people think and work." And that's because that's the utility that's growing bottoms up from AI, which is it's not a GUI anymore. So it's not like SAS. It's a natural language interface. So the experience is going to dictate to the stack, which is growing. Every user's going to want it. So that changes the entire nature of this data stack completely and also changes the relationship for computing architecture.
Chad Cloes
>> Yeah. Frankly, I don't even care who wins because I'm going to be pulling data. I'm going to be pulling the data into a graph, making it relevant and real for me. And it's going to be ChatGPT today. It's going to be Amazon Bedrock tomorrow. It's going to be whatever. It doesn't really matter because I'm going to be able to contextualize and use it in the most important way.
John Furrier
>> Well, this is the best thing about it. General intelligence is the internet. They already got that covered. But if you look at a company like Intuit and others, there's domain expertise. There's specialty intelligence there. That's the real value. So there's going to be a lot of inference going on. So you don't need the mega models to do inference and cool things.
Chad Cloes
>> Well, even within the company, we find that there are silos of domain experts. Inside those sub-domains, you're going to have the smaller graphs and the smaller graphs are going to connect to larger graphs. And it really does-
John Furrier
>> I think that's where the models fit in. So again, it's funny, we predicted this three years ago, that the power law of models and the power law at the head is the popular ones. That's general. But as you move down, open source and specialty comes in. And again, mixture of experts is just mixture of domains. And that's where I think graphs highlights. Chad, great conversation. We're going to probably go another hour on graphs. Again, but the strategic importance of graphs is it's compatible with the future architecture of where AI's going with easy results and benefits that you can get out of data.
Chad Cloes
>> Yeah. In a lot of ways, I think it's a prerequisite.
John Furrier
>> Yeah. Great job. Thanks for coming on. I'm John Furrier with theCUBE here. Neo4j Graph Talk San Francisco. This is where all the top engineers and practitioners get together to share their results and best practice. And also new ways to use graphs, how to build that AI data layer, semantic layer, ontology. The new data engineering is going to be abstracted away with agents and of course the architecture and the data structures. Our all changing graphs are at the center of it. We're doing our best to bring that coverage to you. Thanks for watching.
John Furrier
