Jim Jacob of Cummins Inc., executive director of digitalization, advanced analytics and artificial intelligence, appears in the Neo4j Luminary Series at theCUBE in Palo Alto. Jacob outlines how knowledge graphs, ontologies and large language models unify disparate datasets across supply chain, tariffs, logistics and product domains. They emphasize the role of graph databases in enterprise data strategy and enterprise data architecture to enable reliable inference and decision support.
theCUBE Research frames the discussion with John Furrier of theCUBE and video hosts, focusing on graph adoption, agentic AI patterns and enterprise data architecture. Jacob describes a tariff use case where graph-backed analysis produces enterprise-wide visibility in about 20 days versus months before graph adoption. They argue that autonomous graphs combined with LLMs and agent frameworks provide guardrails, faster context retrieval and scalable contextual AI for business users while dramatically reducing manual data wrangling.
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
Jim Jacob, Cummins | AI Luminaries with Neo4j
Jim Jacob of Cummins Inc., executive director of digitalization, advanced analytics and artificial intelligence, appears in the Neo4j Luminary Series at theCUBE in Palo Alto. Jacob outlines how knowledge graphs, ontologies and large language models unify disparate datasets across supply chain, tariffs, logistics and product domains. They emphasize the role of graph databases in enterprise data strategy and enterprise data architecture to enable reliable inference and decision support.
theCUBE Research frames the discussion with John Furrier of theCUBE and video hosts, focusing on graph adoption, agentic AI patterns and enterprise data architecture. Jacob describes a tariff use case where graph-backed analysis produces enterprise-wide visibility in about 20 days versus months before graph adoption. They argue that autonomous graphs combined with LLMs and agent frameworks provide guardrails, faster context retrieval and scalable contextual AI for business users while dramatically reducing manual data wrangling.
Executive Director - Digitalization, Advanced Analytics and Artificial IntelligenceCummins Inc.
search
John Furrier
>> (INTRO)Hello, I'm John Furrier, host of theCUBE. We are here in our Palo Alto studios for the Neo4j Luminary Series. Folks that are using AI and expanding the data estates and bringing in intelligence of the highest quality. Jim Jacob is here, AI and advanced analytics executive. Jim, thanks for coming on theCUBE. Appreciate it.
Jim Jacob
>> Thank you a lot, John. Looking forward to the discussion.
John Furrier
>> So, one of the things we've been covering in this series is really the rise of graph databases. We're hearing, ontology is being talked a lot about. It's not like it's new, but it is popping into the mainstream and really the focus of AI software. And that's all the buzz. Obviously, the infrastructure, supercomputing capabilities. Now with these large AI factories, as the infrastructure continues to expand, we're seeing a massive shift in how applications are accelerating. So, that's the main topic. So I want to get your thoughts on what you're working on there because we're seeing the data estate kind of not refactoring, but evolving when graphs are playing a big role in that, knowledge graphs. every keynote these days you're hearing knowledge graphs. Is it a new thing? Do I have to go away from something else? How do I integrate it in? What does it actually do? Challenges associated with it. Talk about what you're working on and then we'll kind of dig into because we're seeing this being an acceleration from pilot to production. For a lot of the AI-native applications.
Jim Jacob
>> Yeah, thanks, John. I'd have a few notes about the company that I work with, and that's an important piece in this discussion. As you introduced me, I work for Cummins, and Cummins is a large industrial company that generates power for transportation as well as data centers. So we are closely associated in the— as an industry mode or power for those segments. So we are actively engaging obviously with the AI era, so to say. But I think to the point that you made on graph technology, I personally don't believe that it has rooted in enough. And that's, for me personally, that's a shame because of the real capability it brings to bear. That said, I used to work for 3M before and I started looking at Graph probably around 2012 or so. And then from then on, I've just fallen in love with it and I really like it. I can break down why that's the case when you have a little bit more questions, but I just want to give you the introduction to say for large industrials that have tenured for decades, they're not digital natives. There are, disparate datasets, the mergers and acquisitions that went in, diversification that went in, the changes the company in a constant state. It's a constant state of change for companies like that, and especially for non-digital natives. These companies are going to benefit so much with something like Graph when they understand what they can do for disparate data systems as they build their enterprise context.
John Furrier
>> You know, one of the things that comes up a lot is that, we've heard this before in other data layer kind of conversations pre-AI, break down the silos. Your vertically specialized databases. Now with AI, you're seeing the need for horizontal scale. You want to access as much data as possible, but yet get the domain-specific content and context for any GPU or system. And so you started to see people thinking, well, we got data lake, we got some Snowflake, got some Databricks, and they're thinking, how do I get data into the system faster because the data in will produce the intelligence. So talk about this aspect of it, because this is where we're starting to see different sources of data, different data types kind of binding together kind of at a runtime.
Jim Jacob
>> Yeah, well, that question alone, John, I can go multiple directions. I'm so excited with this whole thing. I'm in the tribe of trying to offer contextual solutions for business problems that the company has. And there are many, many companies that are trying to get into this era are challenged with this. But I'll tell you the fundamental thought. When you say, as an AI person or a technology person, in a nutshell, if you say knowledge graph, the listener is going to look at it as a technical tool. And they probably won't contextualize the value of what a knowledge graph brings in. So the example I use always, and if I may, is that Imagine you're driving down the street. I do this with many of the business folks that I work with, and it kind of clicks really well. And that is, imagine you're driving down the road and you go to office every day. You've been doing that for a long while. In your periphery, you have collected a lot of information in that drive, and that creates the disparate datasets that is really, though it's not cognizant to you, it's in your brain in some way or the other. And then in a water cooler conversation, somebody says, well, it's a yellow building on 35. And you instantly know where that building is, what it was, and suddenly it becomes a part of your memory, of your knowledge. Now you've created a knowledge map, just like it does it in the brain. That's what graph technology does. And as soon as you do that, now when you pass it the next time, you're looking at it, you're collecting even more data about it, and you know exactly what you can get from there because your interaction in the water cooler told you that. Anyway, my big win for using graph technology is exactly that disparate datasets getting connected, vertical or horizontal, being able to figure out how context can be built with disparate systems with graph. Once that happens, It's a liftoff. It's a liftoff in a scale that people have not seen before. Now, when you apply AI into it, now you can imagine it's on steroids. So, users, super users who start tapping into this capability and say, look, I need a couple of knowledge graphs. It's exciting when they say that once they've gone through that curve, they're seeing so much value out of it.
John Furrier
>> You know, one of the things that comes up when I hear you talking is that you think about agents, and we're talking about agentic and all the discussions around multi-step reasoning. It's a term in the AI world. They're saying, okay, it's not just a chatbot prompt response. It's a lot of steps happening. And if you think about a graph, you got a node and an arc. You have connections and context like the brain. This actually is where the magic is, in my opinion. So I want you to expand on that because When you have low latency requirements like feeding the right piece of data at the right time and then adding that periphery to it, the context, this is where the graph shines because, the AI likes to traverse, right? And they want to do it fast. So if you're an agent or you're an application, getting that right data at the right time fills in the picture. Explain that piece of it and why that applies well to the agentic layer or the AI systems themselves, because it's kind of compatible. a neural network is a graph, basically. So you have the computer science integration, but also the relevancy and the reliability.
Jim Jacob
>> Right. No, I think you said it well. I think autonomous graphs are probably going to be the way of the future. I think the combination, triangular combination of graph technology Autonomous capability with AI and ontological representation of your enterprise context is going to give you very, very powerful information for the simple things that you just said. Number one, you have datasets in multiple locations. You have raw material datasets, you have tariff datasets, you have country datasets, logistics, supply chain, and the whole nine yards. The connection to them Fortunately, with very sophisticated language models, it can decipher, it can figure it out generally. But when you have an ontology, you've already hardcoded in some way your enterprise context for that agent. So that agent's able to stick to a guardrail that you want it to stick into and then create the information without hallucinating, which is a big problem for companies who want accuracy. I always look at it, John, in two sets. You have data for analytics, for decision support and decision control, which is very important. That's how you kind of manage your business. And then there is data for accuracy and reporting. Those are two universes that maybe stay together, but they are sometimes different. Characteristics are different. You'd not go to Wall Street with maybe 15%, right? You want to go to Wall Street with, well, 15 to 17%, and you hit inside that range and you need the data to be able to support that. But when you are saying you're trying to infer from the information you have whether a certain product line will work in a certain market and how much investment you need to make for promoting into that market, inference with analytics is going to be very powerful. So, once you make that change and then you say, well, I have datasets that I'm going to use my language, my native language, whatever that may be, translate it into deep language through an LLM. Now I have a translator in the middle, which I didn't have before. I have a translator in the middle who would understand the data much better than if I try to write macros in Excel and try to do jiggery-pokery with Power BI and all that. None of that is required because of the LLM capability. But then you've got to have the graph lined up. If you don't have graph technology, where it can channel the agents in the right places, get the right information, your inference may be good, but it won't be as good as it should be or could be. And that's where I think the opportunity is. And that structure, that layering process of raw data, graph technology, ontologification of it, if that's even a word, and then creating the context makes way more sense.
John Furrier
>> You know, Jim, I loved your example about driving the car and recognizing that building. It's a data point, you grab it. It might not be encoded into the system, but as businesses, it's encoded in your brain in your example, but businesses have been trying to encode their intelligence in databases. We see that all the time, structured databases, and then you got a lot of unstructured data. Could be Slack conversations, could be anything. You start to see this evolution of, the computer brain of the company. Okay. And so how do you view encoding that into AI? Because we've heard semantic layers. Okay. That might be some business process, but getting to meaning kind of at a runtime application or an agent will do tasks. They'll want to recall as much as they can. They're going to want to have as much context. Talk about that dynamic because in the data world, Semantic layers have been around for a while, harmonization layers. These things have been discussed, but they were kind of viewed with the SaaS side. Okay, take that through. As you're an architect and look at this, I'm thinking, okay, how do I lay this out? Take us through how a business should think and how you think about laying out the definitions. What is the business? What is the IP? And where graphs fit in, because It's not about replacing anything. It's really adding into it and kind of bringing things together. Again, like you said, there's disparate data sets everywhere. How does that layer in? Explain that piece of it.
Jim Jacob
>> Yeah. Okay. Well, thanks for that question, John. It's a very, very interesting way that things are evolving right now. And your question has exact premise for that. Let me take an example, a real example of tariffs, right? Tariffs have been, in everybody's mind lately. And I think that it's a big challenge for companies that have, hundreds of millions of components that you want to bring in from multiple countries, multiple suppliers, and then trying to figure out how that tariff actually works, orchestrating and organizing that in a complex environment is really tough. So, in our case, when we attack that, we had raw material data, the whole genome of raw material we had already structured. So, that was a great thing. But That was not all. You needed to have supplier information. You needed to have invoice information. You needed to know where the customer was because not just buying, but where you assemble it, where you transport it, the logistics for it, and then where the customer resides. All that had tariff components in it. So if you were to say, hey, there was a new tariff that came in from the administration, 25% for X country, and we are using those components in, ABC product and it's going to a customer somewhere else. The tariff implications there are complex. So when you have something like that, Graph works like magic because it knows where to connect each of these components into. It knows the language translation into machine language that the user is saying, hey, my classification is these 5 digits. I wanna know the tariff effect. So that when I'm invoicing the customer, I know exactly what to uplift and downgrade as required. Graph does that by connecting all those different disparate datasets together, representing it to you. So that then now, not only do you have information as to what the effects are, it also gives you this real good trick. Oh, by the way, and that's a big thing for business guys. That is, you went in for a problem, But you come out with an, oh, by the way, there is also a tariff exposure on these other 3 elements that you didn't look at. You didn't look at border taxes, you didn't look at custom duty. All that also has a tariff implication in it. That's what the AI component brings into it, becomes much more richer. So I love the, oh, by the way, element, but I think on top of it, for you to be able to do something like, and I think this is another thing that do something like that fast. Because when you have a tariff change and you have millions of dollars of commerce happening inside before the next tariff change happens, you need to know exactly where your money is. So yes, so that's the way we do it in our environment, and it's working brilliantly.
John Furrier
>> Well, that's a great example. And I can imagine too that there's a lot of other data you can add into there. Who's involved? There's some emails going around, maybe some Slack chats, some chats. So All that unstructured data around it would be great additions. Total acceleration. Love that example. The question I have for you is, what would it be? What would it have been like pre-Graph? What would be the task? A schema change? You miss it? What kind of work would go into making that happen at the speed and the accuracy of it without the Graph? What would you have to do? Explain the slog or the work required to implement or even have capability.
Jim Jacob
>> Yeah, well, John, I have been actively trying to forget the days that— Scar tissue. But I'll answer the question, and this is a really important piece, and I try to explain this to some of the business executives when I get connected with them. Imagine doing something like this in Microsoft Excel. I'm not saying Excel is bad by any means. It's a fantastic tool. It was life-changing for mankind, I think, Excel and the whole suite. But I think it would have been laborious. Laborious, number one. You would've had single instances of datasets for each business segment, and it wouldn't have had the capacity or the capability to handle it at scale. And then on top of all that, you won't be able to get what I said, oh, by the way, there's these other elements involved. That's what intelligence brings in. I would say for us, When the tariff issue came up, for us it was a 20-day, 3-week exercise to get enterprise-wide tariff visibility for decision-making and decision support. I would contend it would have taken months, if not 6 to 7 months, to be able to organize that in a meaningful, usable way that can be applied on the fly if we didn't have or if we didn't use Graph. I must tell you though, John, that AI greatly enhanced that exercise too. That's a great thing. So you got to give credit where credit is. But I think, like I told you, the triangular effect is its benefits in the orders of magnitude that we have not seen before. You couldn't have done it unless you built something. Imagine going through multi-million lines of tables to figure out all the information where it is, making those connections, driving APIs. You can't do it with anything that is, that's not, without graphs. So
John Furrier
>> yeah, I teed that up for you because I think this is where I think the breakthrough is with graphs because every use case is subjective to the situation. In this case, tariffs. Complex. There's a lot of data wrangling, a lot of human work involved, grinding out the tables, understanding all the data. And by the way, what— how fresh is the data? Where is it? Did you miss anything? So there's a lot of opportunities to get things wrong in the manual process. And of course, agents love to traverse graphs. So, the what does it mean and oh, by the way are great, elements of the AI. And now with the surge of these decision engines that are emerging, you're starting to see the dots connect. Reasoning turns into decisions. So you're starting to see analytics, reasoning, decision-making, all kind of coming in. So the work you do with the graphs have downstream benefits. Share your thoughts on that and how you view that, because what's right around the corner is not just agents, fleets of agents, decision-making, not queries, scenarios. What if the tariffs change? Things like this are going to be coming more and more. Share your vision on this because this is where I think the acceleration breakthrough goes even further.
Jim Jacob
>> Yeah, I think you said it well too. my hope is that one day a supply chain planner or a finance analyst or a strategic planner will be able to do their job or practice their vocation without having to worry about how am I going to get the data assembled? What am I going to do if I don't have the elements of data? I remember when I got into Cummins, I had my initial interviews and I was talking to the CFO of Cummins at the time. And, I said, in my previous company, people would spend 40, 50% of the time on wrangling data to do their day job. And he said, you're wrong. That doesn't happen in Cummins. And obviously, I'm getting into a new company, I want to impress him, and I was bothered about it. So, towards a couple more questions, I said, I was really bothered about that question. what did you mean by it? He said, it's about 70 to 80%. No one admits it. So, I said, that's an opportunity for me. So, I would love to reduce that. So, my dream, I think my hope is that a practitioner in any function, sourcing, whatever you call it, can do their job without having to worry about anything. Like a Google search, you just give them one line and they say, well, I'm looking for this. Boom, everything happens. For that to happen, these assumptions should be in the right place. Foundational elements data for inference should be ready in an inferenciable fashion. And that is done through graph and ontologies. That is, it's a fundamental table stakes needs for companies to be able to do that. It is no question. I've done this. I've had this game for a decade now. I started in 2011, at 3M, and we've been doing this now for a long while. And I can tell you, without the capability of graph, you're not going to be able to get that. Now, there are companies that are still running around trying to figure out tables and master data and all of the traditional stuff that is '80s, '90s. And I'm like, no, that is a long way home. it's not going to get you there. So, the fact is that the day we get to the point where this is so habitual where we are able to get the information we require for decision-making or for reliable context. When the days are gone, when you don't have to spend time figuring out how the datasets are done, that's when we have come through in this whole era where we can accelerate rapidly. I think that era is at hand because of AI, and I think it's happening. But, believe you me that without something like Graph, without something like ontologies that bring that together, I think it'll be a big fight for companies to put in.
John Furrier
>> Well, Jim,
Jim Jacob
>> I must also tell you, I'm sorry, John, I must also tell you that I feel proud that we are ahead of this game because we started about 5 years ago building that foundation with Graph.
John Furrier
>> Well, congratulations. And we really appreciate what you guys do. We need more power for these data centers and these systems. That are producing the intelligence are going to be determined by who can harness that intelligence, and that's driven by data. So, 18 months out, we're going to see a whole different organizational role of the business operating model of companies, and data will drive that. And what you just said, I think encapsulates the answer to that question, which is the future. You got to get your data in line, and leaders are looking to do that. I really appreciate you sharing what you guys do and your commentary. Appreciate it.
Jim Jacob
>> Thank you, John. I sure hope so. And looking forward to connecting again.
John Furrier
>> Well, congratulations on being a luminary in our series. Again, the future is graphs and this is part of a bigger system of data that feeds the AI intelligence. The intelligence produced from that data will generate the opportunities for companies. I'm John Furrier, host of theCUBE. Thanks for watching.
>> (INTRO)Hello, I'm John Furrier, host of theCUBE. We are here in our Palo Alto studios for the Neo4j Luminary Series. Folks that are using AI and expanding the data estates and bringing in intelligence of the highest quality. Jim Jacob is here, AI and advanced analytics executive. Jim, thanks for coming on theCUBE. Appreciate it.
Jim Jacob
>> Thank you a lot, John. Looking forward to the discussion.
John Furrier
>> So, one of the things we've been covering in this series is really the rise of graph databases. We're hearing, ontology is being talked a lot about. It's not like it's new, but it is popping into the mainstream and really the focus of AI software. And that's all the buzz. Obviously, the infrastructure, supercomputing capabilities. Now with these large AI factories, as the infrastructure continues to expand, we're seeing a massive shift in how applications are accelerating. So, that's the main topic. So I want to get your thoughts on what you're working on there because we're seeing the data estate kind of not refactoring, but evolving when graphs are playing a big role in that, knowledge graphs. every keynote these days you're hearing knowledge graphs. Is it a new thing? Do I have to go away from something else? How do I integrate it in? What does it actually do? Challenges associated with it. Talk about what you're working on and then we'll kind of dig into because we're seeing this being an acceleration from pilot to production. For a lot of the AI-native applications.
Jim Jacob
>> Yeah, thanks, John. I'd have a few notes about the company that I work with, and that's an important piece in this discussion. As you introduced me, I work for Cummins, and Cummins is a large industrial company that generates power for transportation as well as data centers. So we are closely associated in the— as an industry mode or power for those segments. So we are actively engaging obviously with the AI era, so to say. But I think to the point that you made on graph technology, I personally don't believe that it has rooted in enough. And that's, for me personally, that's a shame because of the real capability it brings to bear. That said, I used to work for 3M before and I started looking at Graph probably around 2012 or so. And then from then on, I've just fallen in love with it and I really like it. I can break down why that's the case when you have a little bit more questions, but I just want to give you the introduction to say for large industrials that have tenured for decades, they're not digital natives. There are, disparate datasets, the mergers and acquisitions that went in, diversification that went in, the changes the company in a constant state. It's a constant state of change for companies like that, and especially for non-digital natives. These companies are going to benefit so much with something like Graph when they understand what they can do for disparate data systems as they build their enterprise context.
John Furrier
>> You know, one of the things that comes up a lot is that, we've heard this before in other data layer kind of conversations pre-AI, break down the silos. Your vertically specialized databases. Now with AI, you're seeing the need for horizontal scale. You want to access as much data as possible, but yet get the domain-specific content and context for any GPU or system. And so you started to see people thinking, well, we got data lake, we got some Snowflake, got some Databricks, and they're thinking, how do I get data into the system faster because the data in will produce the intelligence. So talk about this aspect of it, because this is where we're starting to see different sources of data, different data types kind of binding together kind of at a runtime.
Jim Jacob
>> Yeah, well, that question alone, John, I can go multiple directions. I'm so excited with this whole thing. I'm in the tribe of trying to offer contextual solutions for business problems that the company has. And there are many, many companies that are trying to get into this era are challenged with this. But I'll tell you the fundamental thought. When you say, as an AI person or a technology person, in a nutshell, if you say knowledge graph, the listener is going to look at it as a technical tool. And they probably won't contextualize the value of what a knowledge graph brings in. So the example I use always, and if I may, is that Imagine you're driving down the street. I do this with many of the business folks that I work with, and it kind of clicks really well. And that is, imagine you're driving down the road and you go to office every day. You've been doing that for a long while. In your periphery, you have collected a lot of information in that drive, and that creates the disparate datasets that is really, though it's not cognizant to you, it's in your brain in some way or the other. And then in a water cooler conversation, somebody says, well, it's a yellow building on 35. And you instantly know where that building is, what it was, and suddenly it becomes a part of your memory, of your knowledge. Now you've created a knowledge map, just like it does it in the brain. That's what graph technology does. And as soon as you do that, now when you pass it the next time, you're looking at it, you're collecting even more data about it, and you know exactly what you can get from there because your interaction in the water cooler told you that. Anyway, my big win for using graph technology is exactly that disparate datasets getting connected, vertical or horizontal, being able to figure out how context can be built with disparate systems with graph. Once that happens, It's a liftoff. It's a liftoff in a scale that people have not seen before. Now, when you apply AI into it, now you can imagine it's on steroids. So, users, super users who start tapping into this capability and say, look, I need a couple of knowledge graphs. It's exciting when they say that once they've gone through that curve, they're seeing so much value out of it.
John Furrier
>> You know, one of the things that comes up when I hear you talking is that you think about agents, and we're talking about agentic and all the discussions around multi-step reasoning. It's a term in the AI world. They're saying, okay, it's not just a chatbot prompt response. It's a lot of steps happening. And if you think about a graph, you got a node and an arc. You have connections and context like the brain. This actually is where the magic is, in my opinion. So I want you to expand on that because When you have low latency requirements like feeding the right piece of data at the right time and then adding that periphery to it, the context, this is where the graph shines because, the AI likes to traverse, right? And they want to do it fast. So if you're an agent or you're an application, getting that right data at the right time fills in the picture. Explain that piece of it and why that applies well to the agentic layer or the AI systems themselves, because it's kind of compatible. a neural network is a graph, basically. So you have the computer science integration, but also the relevancy and the reliability.
Jim Jacob
>> Right. No, I think you said it well. I think autonomous graphs are probably going to be the way of the future. I think the combination, triangular combination of graph technology Autonomous capability with AI and ontological representation of your enterprise context is going to give you very, very powerful information for the simple things that you just said. Number one, you have datasets in multiple locations. You have raw material datasets, you have tariff datasets, you have country datasets, logistics, supply chain, and the whole nine yards. The connection to them Fortunately, with very sophisticated language models, it can decipher, it can figure it out generally. But when you have an ontology, you've already hardcoded in some way your enterprise context for that agent. So that agent's able to stick to a guardrail that you want it to stick into and then create the information without hallucinating, which is a big problem for companies who want accuracy. I always look at it, John, in two sets. You have data for analytics, for decision support and decision control, which is very important. That's how you kind of manage your business. And then there is data for accuracy and reporting. Those are two universes that maybe stay together, but they are sometimes different. Characteristics are different. You'd not go to Wall Street with maybe 15%, right? You want to go to Wall Street with, well, 15 to 17%, and you hit inside that range and you need the data to be able to support that. But when you are saying you're trying to infer from the information you have whether a certain product line will work in a certain market and how much investment you need to make for promoting into that market, inference with analytics is going to be very powerful. So, once you make that change and then you say, well, I have datasets that I'm going to use my language, my native language, whatever that may be, translate it into deep language through an LLM. Now I have a translator in the middle, which I didn't have before. I have a translator in the middle who would understand the data much better than if I try to write macros in Excel and try to do jiggery-pokery with Power BI and all that. None of that is required because of the LLM capability. But then you've got to have the graph lined up. If you don't have graph technology, where it can channel the agents in the right places, get the right information, your inference may be good, but it won't be as good as it should be or could be. And that's where I think the opportunity is. And that structure, that layering process of raw data, graph technology, ontologification of it, if that's even a word, and then creating the context makes way more sense.
John Furrier
>> You know, Jim, I loved your example about driving the car and recognizing that building. It's a data point, you grab it. It might not be encoded into the system, but as businesses, it's encoded in your brain in your example, but businesses have been trying to encode their intelligence in databases. We see that all the time, structured databases, and then you got a lot of unstructured data. Could be Slack conversations, could be anything. You start to see this evolution of, the computer brain of the company. Okay. And so how do you view encoding that into AI? Because we've heard semantic layers. Okay. That might be some business process, but getting to meaning kind of at a runtime application or an agent will do tasks. They'll want to recall as much as they can. They're going to want to have as much context. Talk about that dynamic because in the data world, Semantic layers have been around for a while, harmonization layers. These things have been discussed, but they were kind of viewed with the SaaS side. Okay, take that through. As you're an architect and look at this, I'm thinking, okay, how do I lay this out? Take us through how a business should think and how you think about laying out the definitions. What is the business? What is the IP? And where graphs fit in, because It's not about replacing anything. It's really adding into it and kind of bringing things together. Again, like you said, there's disparate data sets everywhere. How does that layer in? Explain that piece of it.
Jim Jacob
>> Yeah. Okay. Well, thanks for that question, John. It's a very, very interesting way that things are evolving right now. And your question has exact premise for that. Let me take an example, a real example of tariffs, right? Tariffs have been, in everybody's mind lately. And I think that it's a big challenge for companies that have, hundreds of millions of components that you want to bring in from multiple countries, multiple suppliers, and then trying to figure out how that tariff actually works, orchestrating and organizing that in a complex environment is really tough. So, in our case, when we attack that, we had raw material data, the whole genome of raw material we had already structured. So, that was a great thing. But That was not all. You needed to have supplier information. You needed to have invoice information. You needed to know where the customer was because not just buying, but where you assemble it, where you transport it, the logistics for it, and then where the customer resides. All that had tariff components in it. So if you were to say, hey, there was a new tariff that came in from the administration, 25% for X country, and we are using those components in, ABC product and it's going to a customer somewhere else. The tariff implications there are complex. So when you have something like that, Graph works like magic because it knows where to connect each of these components into. It knows the language translation into machine language that the user is saying, hey, my classification is these 5 digits. I wanna know the tariff effect. So that when I'm invoicing the customer, I know exactly what to uplift and downgrade as required. Graph does that by connecting all those different disparate datasets together, representing it to you. So that then now, not only do you have information as to what the effects are, it also gives you this real good trick. Oh, by the way, and that's a big thing for business guys. That is, you went in for a problem, But you come out with an, oh, by the way, there is also a tariff exposure on these other 3 elements that you didn't look at. You didn't look at border taxes, you didn't look at custom duty. All that also has a tariff implication in it. That's what the AI component brings into it, becomes much more richer. So I love the, oh, by the way, element, but I think on top of it, for you to be able to do something like, and I think this is another thing that do something like that fast. Because when you have a tariff change and you have millions of dollars of commerce happening inside before the next tariff change happens, you need to know exactly where your money is. So yes, so that's the way we do it in our environment, and it's working brilliantly.
John Furrier
>> Well, that's a great example. And I can imagine too that there's a lot of other data you can add into there. Who's involved? There's some emails going around, maybe some Slack chats, some chats. So All that unstructured data around it would be great additions. Total acceleration. Love that example. The question I have for you is, what would it be? What would it have been like pre-Graph? What would be the task? A schema change? You miss it? What kind of work would go into making that happen at the speed and the accuracy of it without the Graph? What would you have to do? Explain the slog or the work required to implement or even have capability.
Jim Jacob
>> Yeah, well, John, I have been actively trying to forget the days that— Scar tissue. But I'll answer the question, and this is a really important piece, and I try to explain this to some of the business executives when I get connected with them. Imagine doing something like this in Microsoft Excel. I'm not saying Excel is bad by any means. It's a fantastic tool. It was life-changing for mankind, I think, Excel and the whole suite. But I think it would have been laborious. Laborious, number one. You would've had single instances of datasets for each business segment, and it wouldn't have had the capacity or the capability to handle it at scale. And then on top of all that, you won't be able to get what I said, oh, by the way, there's these other elements involved. That's what intelligence brings in. I would say for us, When the tariff issue came up, for us it was a 20-day, 3-week exercise to get enterprise-wide tariff visibility for decision-making and decision support. I would contend it would have taken months, if not 6 to 7 months, to be able to organize that in a meaningful, usable way that can be applied on the fly if we didn't have or if we didn't use Graph. I must tell you though, John, that AI greatly enhanced that exercise too. That's a great thing. So you got to give credit where credit is. But I think, like I told you, the triangular effect is its benefits in the orders of magnitude that we have not seen before. You couldn't have done it unless you built something. Imagine going through multi-million lines of tables to figure out all the information where it is, making those connections, driving APIs. You can't do it with anything that is, that's not, without graphs. So
John Furrier
>> yeah, I teed that up for you because I think this is where I think the breakthrough is with graphs because every use case is subjective to the situation. In this case, tariffs. Complex. There's a lot of data wrangling, a lot of human work involved, grinding out the tables, understanding all the data. And by the way, what— how fresh is the data? Where is it? Did you miss anything? So there's a lot of opportunities to get things wrong in the manual process. And of course, agents love to traverse graphs. So, the what does it mean and oh, by the way are great, elements of the AI. And now with the surge of these decision engines that are emerging, you're starting to see the dots connect. Reasoning turns into decisions. So you're starting to see analytics, reasoning, decision-making, all kind of coming in. So the work you do with the graphs have downstream benefits. Share your thoughts on that and how you view that, because what's right around the corner is not just agents, fleets of agents, decision-making, not queries, scenarios. What if the tariffs change? Things like this are going to be coming more and more. Share your vision on this because this is where I think the acceleration breakthrough goes even further.
Jim Jacob
>> Yeah, I think you said it well too. my hope is that one day a supply chain planner or a finance analyst or a strategic planner will be able to do their job or practice their vocation without having to worry about how am I going to get the data assembled? What am I going to do if I don't have the elements of data? I remember when I got into Cummins, I had my initial interviews and I was talking to the CFO of Cummins at the time. And, I said, in my previous company, people would spend 40, 50% of the time on wrangling data to do their day job. And he said, you're wrong. That doesn't happen in Cummins. And obviously, I'm getting into a new company, I want to impress him, and I was bothered about it. So, towards a couple more questions, I said, I was really bothered about that question. what did you mean by it? He said, it's about 70 to 80%. No one admits it. So, I said, that's an opportunity for me. So, I would love to reduce that. So, my dream, I think my hope is that a practitioner in any function, sourcing, whatever you call it, can do their job without having to worry about anything. Like a Google search, you just give them one line and they say, well, I'm looking for this. Boom, everything happens. For that to happen, these assumptions should be in the right place. Foundational elements data for inference should be ready in an inferenciable fashion. And that is done through graph and ontologies. That is, it's a fundamental table stakes needs for companies to be able to do that. It is no question. I've done this. I've had this game for a decade now. I started in 2011, at 3M, and we've been doing this now for a long while. And I can tell you, without the capability of graph, you're not going to be able to get that. Now, there are companies that are still running around trying to figure out tables and master data and all of the traditional stuff that is '80s, '90s. And I'm like, no, that is a long way home. it's not going to get you there. So, the fact is that the day we get to the point where this is so habitual where we are able to get the information we require for decision-making or for reliable context. When the days are gone, when you don't have to spend time figuring out how the datasets are done, that's when we have come through in this whole era where we can accelerate rapidly. I think that era is at hand because of AI, and I think it's happening. But, believe you me that without something like Graph, without something like ontologies that bring that together, I think it'll be a big fight for companies to put in.
John Furrier
>> Well, Jim,
Jim Jacob
>> I must also tell you, I'm sorry, John, I must also tell you that I feel proud that we are ahead of this game because we started about 5 years ago building that foundation with Graph.
John Furrier
>> Well, congratulations. And we really appreciate what you guys do. We need more power for these data centers and these systems. That are producing the intelligence are going to be determined by who can harness that intelligence, and that's driven by data. So, 18 months out, we're going to see a whole different organizational role of the business operating model of companies, and data will drive that. And what you just said, I think encapsulates the answer to that question, which is the future. You got to get your data in line, and leaders are looking to do that. I really appreciate you sharing what you guys do and your commentary. Appreciate it.
Jim Jacob
>> Thank you, John. I sure hope so. And looking forward to connecting again.
John Furrier
>> Well, congratulations on being a luminary in our series. Again, the future is graphs and this is part of a bigger system of data that feeds the AI intelligence. The intelligence produced from that data will generate the opportunities for companies. I'm John Furrier, host of theCUBE. Thanks for watching.
John Furrier