This episode explores knowledge graphs and contextual artificial intelligence for robotics and agentic workflows. Jordi Spranger of Spranic, founder and chief executive officer, appears on the Neo4j AI Luminary Series to discuss knowledge graphs, semantic maps and physical AI. Spranger explains how semantic maps and knowledge graphs supply the contextual intelligence robots and agentic systems need to operate reliably in unstructured environments. They emphasize graph grounding, provenance and the transition from constrained factory settings to free-roaming real-world deployments. The episode is produced by theCUBE Research with hosts John Furrier and Gabe Olave guiding the discussion.
Spranger states that graphs provide the metadata and provenance that ground AI and reduce hallucinations, enabling safer, auditable decision-making for digital agents and physical robots. They recommend beginning knowledge-layer work at the proof of concept level and prioritizing open standards and connectivity with existing platforms such as Databricks and Snowflake. This conversation highlights graph grounding, provenance, knowledge layers and data infrastructure considerations for deploying autonomous systems in complex real-world environments.
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
Jordi Spranger, Spranic | AI Luminaries with Neo4j
This episode explores knowledge graphs and contextual artificial intelligence for robotics and agentic workflows. Jordi Spranger of Spranic, founder and chief executive officer, appears on the Neo4j AI Luminary Series to discuss knowledge graphs, semantic maps and physical AI. Spranger explains how semantic maps and knowledge graphs supply the contextual intelligence robots and agentic systems need to operate reliably in unstructured environments. They emphasize graph grounding, provenance and the transition from constrained factory settings to free-roaming real-world deployments. The episode is produced by theCUBE Research with hosts John Furrier and Gabe Olave guiding the discussion.
Spranger states that graphs provide the metadata and provenance that ground AI and reduce hallucinations, enabling safer, auditable decision-making for digital agents and physical robots. They recommend beginning knowledge-layer work at the proof of concept level and prioritizing open standards and connectivity with existing platforms such as Databricks and Snowflake. This conversation highlights graph grounding, provenance, knowledge layers and data infrastructure considerations for deploying autonomous systems in complex real-world environments.
Jordi Spranger, Spranic | AI Luminaries with Neo4j
Jordi Spranger
Founder & CEOSpranic
search
John Furrier
>> Palo Alto studio connecting Silicon Valley to Wall Street. I'm John Furrier, host of theCUBE here with Gabe Olave, my co-host. Hello, I'm John Furrier, host of theCUBE here in our NYSE theCUBE Studios, of course, our Palo Alto studio connecting Silicon Valley to Wall Street. This is our Neo4j AI Luminary Series. We talk to the leaders who are making it happen in the AI era around data graphs. Knowledge graphs, all the innovation around data that's feeding into the AI is really the key value we're seeing as AI continues to accelerate at the infrastructure layer and the performance gets better and better, faster. The AI needs the data, and that's the purpose of this interview series. Jordi Spranger, founder and CEO of Spranic, is here. Welcome to the Luminary Series and congratulations. You're an AI Luminary.
Jordi Spranger
>> Thank you so much. Nice to be here.
John Furrier
>> First of all, anyone who uses knowledge graphs is a Luminary. First of all, talk about what you guys do because this series explores the notion of graphs, knowledge graphs, context in the spirit of where the AI action is, which is how do I get the data fast, relevant at the right time at the right place?
Jordi Spranger
>> Yeah, right. So we're taking the approach from the physical AI side. And right now, robotics and physical AI is making huge jumps. But where they are lacking and where they are being underserved is this context. And I think also robots deserve the right context at the right time across domains. And this is the problem that we're tackling. And we can see that this works really well for digital agents already. And then now we're bringing this to the physical world.
John Furrier
>> the AI knowledge layer has been discussed. We've been covering, data for about 17 years back and going back to the old Hadoop days, but Now you hear ontology is being kicked around. That's been around for a while, knowledge graphs, but there's a lot of data going in. What is the value of having that graph from your experience, people who are learning, what is the main value?
Jordi Spranger
>> So the thing is, you have these, structured tables, but if you look in the real world, and this is also where AI and robotics is going, especially we're going from very constrained factory environments with robots to very unstructured free environments in the real world. And the real world is connected as a graph. So we are all on planet Earth and on Earth there's like cities and countries and in the countries there's different factories, et cetera. So everything is connected in the real world. And that's why if you want to make robots really use context for working in the real world, graphs seem to be the real natural way to go.
John Furrier
>> Talk about what you guys are working on right now. What's your company doing? How are you using graphs? What's the upside?
Jordi Spranger
>> Yeah, so essentially the idea came because before I was doing this, I was doing the deployments for Boston Dynamics in the DACH region through a technical consultancy. And the problem that I found to be reoccurring is that you have these problems of missing context to do new deployments for robots. So if you have a Boston Dynamics Spot, that's the robotic dog, and you want to do predictive maintenance, etc., these use cases, you need to have a map and then you need to plan it out. And to redo this, it would take essentially the same effort every single time. And what it is missing is this context so that you can agentically or semi-automatically redeploy these robots for new tasks relatively quick. And that is really a crucial and critical point to scale AI to millions and billions of robots that don't need a robotics engineer every time it breaks down or needs to be replanned or repurposed for a new process and so on.
John Furrier
>> What was that context that you saw in your problem statement? Was it the data wasn't available, was hard to get, latency was too long. What was the issue?
Jordi Spranger
>> So the issue is that it doesn't really exist right now. And what I mean by that is for businesses, you need Excel data and that's readily available in the right format to work with it. But for robots, maps are a really big part of robotics. So a map is really mapping out the environment around it. What exists, what doesn't exist. Google Maps is a great example because Google Maps has all the streets and intersections already mapped. This is the map and it serves the purpose of navigation in robotics, especially free-roaming robots like humanoids. They need Google Maps on steroids to be really free from a statically planned task and can replan across a lot of tasks that a humanoid could do. And for that, this context, a map that is understandable and has semantic value, it's incredibly important. And that's lacking right now.
John Furrier
>> Talk about the impact of say, AI hallucinations or not having the data because the relationship maps also are in there. You have relationships and context in a graph. So it's almost like picking your brain. You have access to that information. Talk about what happens if you don't have a graph? What happens when you have the graph with AI?
Jordi Spranger
>> Yeah, sure. So I think what we can see also in general with agents and agentic workflows is if you just give them like a table and you don't give them the structure to navigate information landscapes, then they will be stuck to either hallucinate. That's the bad thing you don't want, or they will just not be able to do the task efficiently. And what you need is with this grounding of information and this navigation for agents to find the relevant information, you need this in the systems. And this is where the graphs really provide this metadata layer information that helps you to navigate information and use the right one at the right time.
John Furrier
>> Yeah, and AI is going to help that speed. You mentioned agents and robotics are two use cases that pop out a lot. I want to get your thoughts on this because on robotics, it's very hardcore. Safety is a huge concern. Getting everything accurate is really paramount. You cannot have inaccuracies and you can't have hallucinations. On the agent side, people have been more loose about it, but then they realize, wait a minute, I can't have transactions that are going to be inaccurate. So the accuracy becomes super important. Now you're starting to see people really focus on this. What's your reaction? How do you explain the accuracy piece and how graphs can apply because, fast and loose agents is one thing and unsafe robotics is another. No one wants that.
Jordi Spranger
>> Yeah, absolutely. wrong helpdesk support agent that gives hallucinations is one thing. But in robotics, if you have physical force with people, that is of course a huge danger and you need reliability, you need security. And there are new standards developed right now. So this is a very new field where you need to say, okay, what are the standards that need to be set to make humanoids and other mobile robots safe to operate? And there's a lot of layers to this. So this is on the hardware layer, sensor layer, algorithm layer, and information retrieval layer as well, right? And the one thing that has been working really well for agents so far to reduce that, because at the end AI is always probabilistic, but to reduce it to a safe level is usually a grounding mechanism where it says also when you use ChatGPT, et cetera, it needs to link back to the main source, make it traceable so that you can analyze later on and say this is something reasonable or not, or here we have to adjust. And this grounding is also something that we do for maps. So if it says I will go XYZ location and I will do XYZ, it needs to be grounded in a real map asset of data. And this grounding and these harnesses is something that is an open field, but we're addressing it.
John Furrier
>> I like that angle on that. I want to ask you about context and reasoning because you got to have the grounding. Okay, great. Now you have agents and AI that's going to be more than just search retrieval. There's actual reasoning going on. So context becomes super valuable. What's the angle on graphs and context and reasoning memory? These are big discussions.
Jordi Spranger
>> Yes. So that's the beauty of it all. At the end of the day, a lot of domains have to come together and graphs are really good at connecting domains. They say Here's some information, here's another information located, connect these if you need the relationship and need to leverage it for a task. And this is something that has to account for robots that act in a huge variety of different scenarios. They need to have the right information sources, provenance as well, and work in an auditable way, in a safe way like you mentioned. And this is something where graphs provide the flexibility to connect data sources that are needed and remove data sources that are not needed.
John Furrier
>> Jordi, that brings up a good point. I talk to people all the time, I got so much data, I got a data lake over here, a little Snowflake, I got Databricks, mostly enterprise customers. But when you have a lot of data, and certainly if you have new data coming in, how should an architect think about their existing data platforms and how should they think about the combination or the intersection of a knowledge layer?
Jordi Spranger
>> Yeah, so I think these things can work in harmony together, right? So there are certain data formats and certain data solutions. You said Databricks and Snowflake, open formats, and they work really well. The question is just how do we connect this coherently to serve a bigger enterprise landscape? And for this, you always have to go the route that nowadays things need to be connected. You can have specialized solutions, but these solutions need to be connectable to the rest of your data ecosystem. So if you ask me what is the number one thing, I think it's open standards, it's connectivity, and it's the guardrails along the way so that you can make this safe and not a wild landscape.
John Furrier
>> You know, we love AI. It's very dynamic, great personalization, but autonomy has been the big discussion. Robotics, we see the cars, robotaxis, autonomous systems have to have the ability to execute, especially in robotics and the edge. What are your thoughts on the autonomous piece of it? How fast can we get there? Are we already there? What's your view there?
Jordi Spranger
>> So I think autonomy is a given in certain constrained domains. So it really goes, you probably know the autonomous driving level 1, 2, 3, 4. Yeah. And these are kind of levels that are already achieved or they are close to being achieved. The autonomy is not that much the problem anymore, I believe. The question is, where can we deploy autonomy and where is the risk-benefit analysis in favor of actually doing it also from a compliance and safety and regulatory aspect? And especially in autonomous driving, the risk is quite high because you have these tons with high velocity that can kill people, of course. And so I think autonomy is here. That's my take on it. I think autonomy will come to also less known places than self-driving cars. It can come in your house to clean your flat in some low-risk environment. And I really want to see it there first because that can give us a baseline of how well it works, how we can evaluate it, gives regulatory frameworks time to test it and then roll it out to more high-risk areas.
John Furrier
>> What's your advice to folks who are kind of stuck in a pilot phase with the project or haven't yet got a knowledge graph, trying to get the context right. What's your advice? How do I implement, say, Neo4j or implement this knowledge layer? What would you suggest? We've had many examples where people hit a breakthrough literally overnight after implementing knowledge graphs.
Jordi Spranger
>> Yes. So I've seen this exact situation you described on the robotics level. Companies are extremely excited about robots, mobile robots. They buy them, they do one POC, and then they figure out it doesn't work as well, or it's quite difficult to reuse it for something else, and they get tired of this. And the real reason behind that is because you're missing a data layer. You're missing a layer of context for robots, something that you connect your processes to. That's also where graph databases come into play. And I think what you need to be aware of is you need to make the foundations, you need to make the infrastructure, and you can do that also at a POC level. That is not a big deal. So graph databases work well on a small and larger scale. You just have to be aware that they are part of the infrastructure. And if you build without them, you might not get as far as you think.
John Furrier
>> Yeah, and they certainly can accelerate. And they are infrastructure. Jordi, thanks so much for coming on the Neo4j AI Luminary Series. Appreciate your insight and thanks for sharing.
John Furrier
>> Thank you so much, John.
John Furrier
>> All right. Cheers. I'm John Furrier with theCUBE. This is our Luminary Series. People working with knowledge layers are achieving success in the AI era as the compatibility of graphs and AI work very, very well together. We're seeing a lot more results of people moving the needle faster with graphs, knowledge layers, and better context. I'm John Furrier, your host. Thanks for watching.
Jordi Spranger, Spranic | AI Luminaries with Neo4j
search
John Furrier
>> Palo Alto studio connecting Silicon Valley to Wall Street. I'm John Furrier, host of theCUBE here with Gabe Olave, my co-host. Hello, I'm John Furrier, host of theCUBE here in our NYSE theCUBE Studios, of course, our Palo Alto studio connecting Silicon Valley to Wall Street. This is our Neo4j AI Luminary Series. We talk to the leaders who are making it happen in the AI era around data graphs. Knowledge graphs, all the innovation around data that's feeding into the AI is really the key value we're seeing as AI continues to accelerate at the infrastructure layer and the performance gets better and better, faster. The AI needs the data, and that's the purpose of this interview series. Jordi Spranger, founder and CEO of Spranic, is here. Welcome to the Luminary Series and congratulations. You're an AI Luminary.
Jordi Spranger
>> Thank you so much. Nice to be here.
John Furrier
>> First of all, anyone who uses knowledge graphs is a Luminary. First of all, talk about what you guys do because this series explores the notion of graphs, knowledge graphs, context in the spirit of where the AI action is, which is how do I get the data fast, relevant at the right time at the right place?
Jordi Spranger
>> Yeah, right. So we're taking the approach from the physical AI side. And right now, robotics and physical AI is making huge jumps. But where they are lacking and where they are being underserved is this context. And I think also robots deserve the right context at the right time across domains. And this is the problem that we're tackling. And we can see that this works really well for digital agents already. And then now we're bringing this to the physical world.
John Furrier
>> the AI knowledge layer has been discussed. We've been covering, data for about 17 years back and going back to the old Hadoop days, but Now you hear ontology is being kicked around. That's been around for a while, knowledge graphs, but there's a lot of data going in. What is the value of having that graph from your experience, people who are learning, what is the main value?
Jordi Spranger
>> So the thing is, you have these, structured tables, but if you look in the real world, and this is also where AI and robotics is going, especially we're going from very constrained factory environments with robots to very unstructured free environments in the real world. And the real world is connected as a graph. So we are all on planet Earth and on Earth there's like cities and countries and in the countries there's different factories, et cetera. So everything is connected in the real world. And that's why if you want to make robots really use context for working in the real world, graphs seem to be the real natural way to go.
John Furrier
>> Talk about what you guys are working on right now. What's your company doing? How are you using graphs? What's the upside?
Jordi Spranger
>> Yeah, so essentially the idea came because before I was doing this, I was doing the deployments for Boston Dynamics in the DACH region through a technical consultancy. And the problem that I found to be reoccurring is that you have these problems of missing context to do new deployments for robots. So if you have a Boston Dynamics Spot, that's the robotic dog, and you want to do predictive maintenance, etc., these use cases, you need to have a map and then you need to plan it out. And to redo this, it would take essentially the same effort every single time. And what it is missing is this context so that you can agentically or semi-automatically redeploy these robots for new tasks relatively quick. And that is really a crucial and critical point to scale AI to millions and billions of robots that don't need a robotics engineer every time it breaks down or needs to be replanned or repurposed for a new process and so on.
John Furrier
>> What was that context that you saw in your problem statement? Was it the data wasn't available, was hard to get, latency was too long. What was the issue?
Jordi Spranger
>> So the issue is that it doesn't really exist right now. And what I mean by that is for businesses, you need Excel data and that's readily available in the right format to work with it. But for robots, maps are a really big part of robotics. So a map is really mapping out the environment around it. What exists, what doesn't exist. Google Maps is a great example because Google Maps has all the streets and intersections already mapped. This is the map and it serves the purpose of navigation in robotics, especially free-roaming robots like humanoids. They need Google Maps on steroids to be really free from a statically planned task and can replan across a lot of tasks that a humanoid could do. And for that, this context, a map that is understandable and has semantic value, it's incredibly important. And that's lacking right now.
John Furrier
>> Talk about the impact of say, AI hallucinations or not having the data because the relationship maps also are in there. You have relationships and context in a graph. So it's almost like picking your brain. You have access to that information. Talk about what happens if you don't have a graph? What happens when you have the graph with AI?
Jordi Spranger
>> Yeah, sure. So I think what we can see also in general with agents and agentic workflows is if you just give them like a table and you don't give them the structure to navigate information landscapes, then they will be stuck to either hallucinate. That's the bad thing you don't want, or they will just not be able to do the task efficiently. And what you need is with this grounding of information and this navigation for agents to find the relevant information, you need this in the systems. And this is where the graphs really provide this metadata layer information that helps you to navigate information and use the right one at the right time.
John Furrier
>> Yeah, and AI is going to help that speed. You mentioned agents and robotics are two use cases that pop out a lot. I want to get your thoughts on this because on robotics, it's very hardcore. Safety is a huge concern. Getting everything accurate is really paramount. You cannot have inaccuracies and you can't have hallucinations. On the agent side, people have been more loose about it, but then they realize, wait a minute, I can't have transactions that are going to be inaccurate. So the accuracy becomes super important. Now you're starting to see people really focus on this. What's your reaction? How do you explain the accuracy piece and how graphs can apply because, fast and loose agents is one thing and unsafe robotics is another. No one wants that.
Jordi Spranger
>> Yeah, absolutely. wrong helpdesk support agent that gives hallucinations is one thing. But in robotics, if you have physical force with people, that is of course a huge danger and you need reliability, you need security. And there are new standards developed right now. So this is a very new field where you need to say, okay, what are the standards that need to be set to make humanoids and other mobile robots safe to operate? And there's a lot of layers to this. So this is on the hardware layer, sensor layer, algorithm layer, and information retrieval layer as well, right? And the one thing that has been working really well for agents so far to reduce that, because at the end AI is always probabilistic, but to reduce it to a safe level is usually a grounding mechanism where it says also when you use ChatGPT, et cetera, it needs to link back to the main source, make it traceable so that you can analyze later on and say this is something reasonable or not, or here we have to adjust. And this grounding is also something that we do for maps. So if it says I will go XYZ location and I will do XYZ, it needs to be grounded in a real map asset of data. And this grounding and these harnesses is something that is an open field, but we're addressing it.
John Furrier
>> I like that angle on that. I want to ask you about context and reasoning because you got to have the grounding. Okay, great. Now you have agents and AI that's going to be more than just search retrieval. There's actual reasoning going on. So context becomes super valuable. What's the angle on graphs and context and reasoning memory? These are big discussions.
Jordi Spranger
>> Yes. So that's the beauty of it all. At the end of the day, a lot of domains have to come together and graphs are really good at connecting domains. They say Here's some information, here's another information located, connect these if you need the relationship and need to leverage it for a task. And this is something that has to account for robots that act in a huge variety of different scenarios. They need to have the right information sources, provenance as well, and work in an auditable way, in a safe way like you mentioned. And this is something where graphs provide the flexibility to connect data sources that are needed and remove data sources that are not needed.
John Furrier
>> Jordi, that brings up a good point. I talk to people all the time, I got so much data, I got a data lake over here, a little Snowflake, I got Databricks, mostly enterprise customers. But when you have a lot of data, and certainly if you have new data coming in, how should an architect think about their existing data platforms and how should they think about the combination or the intersection of a knowledge layer?
Jordi Spranger
>> Yeah, so I think these things can work in harmony together, right? So there are certain data formats and certain data solutions. You said Databricks and Snowflake, open formats, and they work really well. The question is just how do we connect this coherently to serve a bigger enterprise landscape? And for this, you always have to go the route that nowadays things need to be connected. You can have specialized solutions, but these solutions need to be connectable to the rest of your data ecosystem. So if you ask me what is the number one thing, I think it's open standards, it's connectivity, and it's the guardrails along the way so that you can make this safe and not a wild landscape.
John Furrier
>> You know, we love AI. It's very dynamic, great personalization, but autonomy has been the big discussion. Robotics, we see the cars, robotaxis, autonomous systems have to have the ability to execute, especially in robotics and the edge. What are your thoughts on the autonomous piece of it? How fast can we get there? Are we already there? What's your view there?
Jordi Spranger
>> So I think autonomy is a given in certain constrained domains. So it really goes, you probably know the autonomous driving level 1, 2, 3, 4. Yeah. And these are kind of levels that are already achieved or they are close to being achieved. The autonomy is not that much the problem anymore, I believe. The question is, where can we deploy autonomy and where is the risk-benefit analysis in favor of actually doing it also from a compliance and safety and regulatory aspect? And especially in autonomous driving, the risk is quite high because you have these tons with high velocity that can kill people, of course. And so I think autonomy is here. That's my take on it. I think autonomy will come to also less known places than self-driving cars. It can come in your house to clean your flat in some low-risk environment. And I really want to see it there first because that can give us a baseline of how well it works, how we can evaluate it, gives regulatory frameworks time to test it and then roll it out to more high-risk areas.
John Furrier
>> What's your advice to folks who are kind of stuck in a pilot phase with the project or haven't yet got a knowledge graph, trying to get the context right. What's your advice? How do I implement, say, Neo4j or implement this knowledge layer? What would you suggest? We've had many examples where people hit a breakthrough literally overnight after implementing knowledge graphs.
Jordi Spranger
>> Yes. So I've seen this exact situation you described on the robotics level. Companies are extremely excited about robots, mobile robots. They buy them, they do one POC, and then they figure out it doesn't work as well, or it's quite difficult to reuse it for something else, and they get tired of this. And the real reason behind that is because you're missing a data layer. You're missing a layer of context for robots, something that you connect your processes to. That's also where graph databases come into play. And I think what you need to be aware of is you need to make the foundations, you need to make the infrastructure, and you can do that also at a POC level. That is not a big deal. So graph databases work well on a small and larger scale. You just have to be aware that they are part of the infrastructure. And if you build without them, you might not get as far as you think.
John Furrier
>> Yeah, and they certainly can accelerate. And they are infrastructure. Jordi, thanks so much for coming on the Neo4j AI Luminary Series. Appreciate your insight and thanks for sharing.
John Furrier
>> Thank you so much, John.
John Furrier
>> All right. Cheers. I'm John Furrier with theCUBE. This is our Luminary Series. People working with knowledge layers are achieving success in the AI era as the compatibility of graphs and AI work very, very well together. We're seeing a lot more results of people moving the needle faster with graphs, knowledge layers, and better context. I'm John Furrier, your host. Thanks for watching.
John Furrier