How a New AI Architecture Fixes Hallucinations and Builds Trust!

Duncan Riley (00:01.927)

Hello and welcome back to Building Momentum, the show where we pull back the curtain on the exciting and often chaotic world of building a successful technology business. I'm Duncan, your host for the show, where every episode we bring you the stories and strategies of people who have been in the trenches, conquering churn, scaling their team and building products that people and businesses love. In this episode, we'll be chatting with John, the founder and CEO of Data Squared, pioneering next generation artificial intelligence.

that transforms how organizations understand and leverage their data. A US military veteran and from the United States Air Force and a New York University London School of Economics and Harvard Aluminus, John's experience lies in developing transparent, explainable AI solutions for high reliability sectors, including defense, intelligence, energy, and critical infrastructure. His global experience spans multiple continents,

including innovative technology implementations in Scotland, Nigeria, South Africa, Brazil, the UK, Thailand and Australia. John, it's great to have you on the show. Thank you for joining us.

Jon Brewton - CEO Data² (01:13.038)

Yeah, thanks for having me. I look forward to kind of getting into the discussion today and telling you a little bit about our story and, you know, kind of learning what you guys are up to and just hearing about all the crazy stuff happening in AI at the moment.

Duncan Riley (01:27.325)

Brilliant, brilliant. Well, thanks for obviously joining the show. Really, maybe what we'll start with is what's data squared and what's the big problem out there that you're trying to solve in the world?

Jon Brewton - CEO Data² (01:41.614)

Yeah, I think a couple of different things. So Data Square is company we founded in September of 2023. So we're a young company. We haven't been around that long. The team has a couple of key themes, at least in the founding cohort and the folks that are working for us right now. We either served in the US military in some capacity. We have a Navy SEAL, an Air Force vet, a Marine Corps vet, and an Army vet. We worked in critical infrastructure industry like mining or oil and gas.

And we studied together in some capacity, whether that was at NYU, London School of Economics, Harvard, we all kind of went to school together and we studied new and innovative ways to sort of use technology to grow businesses. And after all that learning, we thought, it'd probably be interesting to give this a shot ourselves. And so we ended up starting a company. And I'd say at the beginning of the company, we really didn't know what we wanted to do. We know that we wanted to use technology. We knew that we wanted to

try to help people build solutions and do really interesting stuff. We boiled it down to do cool stuff with great people. And that's kind of the founding ethos that we started with. But at the end of the day, we use some of our background and our understanding of the industries that we wanted to target, primarily defense and intelligence, and then specifically the energy industry. And some of our technical experience with

scaling different solutions, commercializing technology and doing different things to try to understand the balance between what the market really needed and this really cool and interesting, potentially transformative new technology from a generative AI perspective that was just hitting the market. And we ultimately settled on, we wanted to figure out a way to limit hallucinations in application, scale systems.

that could integrate a lot of different intelligence and information together because enterprise offerings and different companies that work across enterprise sectors, especially in the government, really differentiated, bifurcated systems infrastructure. Nothing is sort of connected. And we wanted to figure a smart way to sort of connect these things together. And we wanted to see if we could create a system that could give people explainability in a systematic way.

Jon Brewton - CEO Data² (04:01.982)

A lot of the systems that you'll see today from foundation models like, know, Claude or open AI suffer from hallucinations and they suffer from black box sort of wiring. You you can't really see how things are happening under the hood and why they're happening and you can't really decipher how information was leveraged and what that means for the answers that you get. And that was kind of going back to the very beginning of

you know, Chat GPT coming out, we had a couple of months or runway to understand, you know, how these systems work. And then we try to use our history and the things that we had done, you know, working for companies like Chevron and BP and other big mining entities around the world to try to solve this problem. And so we created the review platform. And so data squares review platform is the only fully explainable AI system.

that you can deploy in high reliability, high stakes sectors. It offers complete traceability, auditability, zero hallucination based outcomes. And ultimately we wanted to make sure that we could deliver something that was transformative, that met the moment associated to the new technology and what it meant for its potential scale and commercialization. And ultimately trying to take advantage of the value proposition that you could get if you properly design systems around these capabilities.

Duncan Riley (05:19.827)

So you've obviously mentioned some pretty strong sectors there, know, defence and oil and gas. So could you apply data squared across most industries or would you say that it really works well when you can't see under the hood and you need to aggregate that data and bring it to life essentially? Or would you say that for our audience, if they're watching this, can it be applied to multiple sectors?

Jon Brewton - CEO Data² (05:46.638)

So I'd say when we started, we started with the thesis that we were going to apply it to oil and gas. And the initial proof of concept that we built and the initial systems that we developed were all for oil and gas, really well abandonment use cases. So as you had a well on production, its usable life goes down. Now it's just sort of taking money and not really productive from an economic or really hydrocarbon's perspective. You want to decommission that asset.

That's usually a very tricky process because these things produce in some instances for up to 50 years. And so we wanted to try to reconcile that history. And it just so happens that the way that we design this system is really applicable to anything. It's not applicable to any given industry. It's not applicable to any domains, not applicable to any use case. It can be applied to basically anything. But there are sort of tenets of operation that we try to at least communicate to customers the way that these systems work best.

we use knowledge graphs as a fundamental part of this equation, really work best in areas where you're trying to map networks or systems and the value of how those systems are connected and why individual pieces are connected in certain ways is really getting down into the decision analysis that you would do to create value in these areas. So think of things like

cybersecurity, supply chain, engineering workflows, any sort of financial fraud investigations or market mapping exercises that you'd like to do, any personalized sort of medicine applications or like new drug development. These rely on lot of disparate pieces of information being aggregated in a way that's tangible and traceable so that you can understand how to optimize that network at one point in time.

So that was really our focus is like, if we're going to re-engineer something, well, we need a lot of data. If we're going to manage a network or an infrastructure of networks, how do we connect these things together? And graphs happen to be a great way to do that. And so they work best where the connections in your data matter and you're trying to model across sort of a value chain, but it's really applicable to anything. mean, use cases, you could apply it to a general chat bot functionality that you're trying to build.

Jon Brewton - CEO Data² (08:10.83)

you know, graphs as an underlying part of the equation allow you to really understand how things are connected, why they're connected and what that means for how you can optimize anything. And it really captures sort of this rich interconnected nature of the data. And so if you're thinking about where to apply it, I think it's really fit for purpose in those areas where the networks matter, the connections in the data matter, and you're trying to reconcile large volumes of

disconnected information to come up with some answer or way to optimize a system. So preventative maintenance is another great example.

Duncan Riley (08:47.315)

Yeah, absolutely. was going to say, within these large organizations, who do you think gets the most value from data squared? Who's kind of in it and benefiting from all this aggregated data to be able to visualize it? What sort of type of avatar would you say?

Jon Brewton - CEO Data² (09:03.296)

Yeah. Yeah, look. Yeah. Well, the personas are kind of different. The personas that you would sell to versus the personas of the users are really in many instances vastly different. A lot of the stuff that we curate works really, really well in sort of a asset management layer of thinking. So if you're an executive chief operating officer and you're running a large sort of footprint of disconnected and bifurcated assets and you want to understand how to

model within this network and toggle this network on an economic basis. You know, that's the people that we're building systems for, but very often those are not the people that are actually executing the work to either implement those changes or even find the areas of opportunity for those changes. And so we're really selling to personas that are head of analytics, head of IT, chief operating officers, chief financial officers, kind of that lane.

But we're building solutions for the people that are working these things on a day-to-day basis. So we're building it at an engineering level or a financial analyst level or intelligence analyst level, the people that need to really engage with the data. One of the things that we sort of have as a foundational tenant for what we built is that, you know, people see the world in rows and columns. And we like to really sort of use the underlying data model infrastructure that we build around knowledge graphs to reveal

in sort of a hidden way, a dimension and a landscape of really rich connected insights. And we use those to sort of reason to and from. And that allows us to get a couple of things that really, really important out of the system. One, it allows us to complete the transparency angle that we're really searching for and the traceability angle. We know what data we pass these systems. We know what data are used by these systems. What is...

turned in the answers, how that aligns with the fact basis that we provide it and in the overall database that we're curating from. We understand how we can grade those answers on a subject matter expertise basis, given the problem that we're trying to solve. And graphs really sort of allow us to do that, but it also sort of strips away this whole black box element of AI. Now we understand how the AI is working, why it's doing what it's doing, and we can start to toggle and lever these things in our favor in the background. So that's ultimately, you know,

Jon Brewton - CEO Data² (11:26.562)

kind of the places that we're playing, people we're selling to and how we're trying to set the system up. But it's really trying to present a really, really simple user interface to people that is in natural language that allows them to really derive sort of very unique, but traceable and tangible outcomes from now connected sort of information.

Duncan Riley (11:48.755)

It sounds brilliant, you know, for those watching who are constantly evaluating loads of data sets and data points coming in at a fast moving pace and an ever changing environment. It sounds something like this would be fantastic for them. Obviously, you're still a young company to certain extent. I always like to ask this question, but what's been your biggest challenge so far?

You know, has it been building the product? Has it been acquiring some of those customers? What have you really faced on this? This has a big challenge for you.

Jon Brewton - CEO Data² (12:27.032)

Yeah, I think challenges manifest in different ways, right? And the building of a high reliability, high confidence, hallucination free, explainable AI platform comes with its own challenges, right? There's a lot of technical complexity, a lot of sort of underlying work that you need to do from a research and development perspective, make sure that you can do this right. As far as we know right now, we're the only people in the world that can do this. So, you know, that, that kind of speaks to the R and D effort that we took to sort of build this system.

But I'd say the bigger issues that we've really had to manage more effectively come in the form of just sort of counter narrative positioning. know, like there's an article in the Wall Street Journal very recently. I believe the title of this article was something along the lines of beyond hallucination prevention, the power of connected data systems. it's, know, quoting people from OpenAI and Claude and,

just a variety of different large companies. Microsoft is another company that's quoted in this. And the sort of tagline that this starts with is you can't stop an AI chat bot from sometimes hallucinating or giving misleading or mistaken answers to a prompt or even making things up. But there are some things you can do to limit the amount of faulty information. And so we just stop right there. Like we can stop.

AI chat bots from hallucinating, giving misleading information and giving mistaken answers to, you know, even poorly prompted questions. We can stop them from making things up. And that's really counter narrative. At the end of the day, we have had to go on sort of an education campaign to talk to folks about what is possible. And you know, like our mission statement as a company is we want to, and we exist to discover a better way. And I think

That's really part and parcel to the overall message and the capabilities we're trying to bring to the market. You're not stuck with this legacy systems, even though they're two years old, you're not stuck with the inefficiencies of the systems. There is a better way to do this. And like our vision as a company is to build a world where people and machines make smarter decisions together. And at the end of the day, we think we can do that, but we have to sort of convince people that the things that we do are even possible.

Jon Brewton - CEO Data² (14:45.61)

as a byproduct of like sort of this counter narrative positioning that we have. So what is an overwhelming amount of information that comes in from the likes of Sam Altman and Elon Musk and a variety of other people that have very large megaphones that they use to pass their messages.

Duncan Riley (15:03.382)

Yeah, absolutely. was going to, yeah, I mean, it's pretty, it's hard to comprehend sometimes, you know, when you're dealing with such large bits of information and how do you make it relevant to that person at that time? Now, John, you've obviously had some great experience, you being in the military, working with some of your current colleagues as well, you mentioned sort of.

maybe Seal and whatnot. Seems like a high performing team with a high performing output. But I'm keen to find out about yourself and guess like the early days and maybe like your journey kind of like leading up to this because you just don't get to become CEO of Data Squared and building awesome product. There's a journey behind it and people are probably wondering how did we get to this point? it's, yeah, I'd love to find out a little bit more about that.

Jon Brewton - CEO Data² (15:36.514)

Yeah.

Jon Brewton - CEO Data² (16:00.846)

Yeah, no problem. Look, we started the company, what was it, 20 months ago or whatever that was. So not that long ago, but obviously the path to starting DataSquared started well before that. You've read in my bio or at least the intro, it's kind of lived and worked in a lot of different places. And I've lived and worked on a lot of different problems and a lot of different sort of scale of different problems.

I think everything we do in life sort of leads us to find ways that we can hit easy buttons and try to change the trajectory of our personal efficiency and the things that we're asked to do on a pretty routine and recurring basis. Engineering's no different. I started out in the military, I didn't know what I wanted to do. I got out of the military, ended up hiring into this training program with BP for an oil and gas operations role. That shortly years later turned into an engineering role.

A couple of years later, I'm working at Chevron. I stopped this blowout on a well. I meet the president of our company and he's like, you should come work for me. So I end up, you know, very shortly after that, going to work for him and then start taking a bigger picture view of the business of oil and gas through an engineering lens. And then it was all about optimizing different business units. So I travel around the world to different places.

and try to figure out ways to increase their profitability, decrease their waste, and then overall just change the efficiency curves of these businesses. Now to do that, you need specialized expertise and understanding of how data is connected, why it's connected, how you generate value from these different silos of information.

And so I did like portfolio rationalization around different applications that we would use and why. And then I got into the commercialization of technology. And then I got into trying smart ways to use ML processes and AI and trying to scale these things internally. And that led me to going back to school where you meet sort of different people and you sort of new perspectives are built for how you can do different things. ultimately I think it was Chevron.

Jon Brewton - CEO Data² (18:09.644)

that led me around the world. And it was Chevron that led me to do my MBA. And it was during those sort of years doing my MBA with different people that you start to find things that you're really interested in and data analytics and in general optimization was a big part of that. But, you know, that led me to Harvard, which does some post-grad work there. And ultimately is all focused on

how to integrate technology and smart ways into your businesses and how to change the efficiency curves of your businesses with commercialized approach, the scale technology implementation. And so that led me to Quantium, who is at least in Australia and the Anzac region, the premier machine learning AI solutions development firm across Australia and New Zealand. They work in America too. And I happen to be lucky enough to work

Harvard and said, I've post grad worked with the chairman of Quantium and got an opportunity to come over and work on an industrial venture with Quantium. And Quantium was an amazing company because it kind of opened my eyes to the possibilities that were really in front of me, thinking about things in different ways, understanding different approaches to solving problems, a scale platform approach to development. it's all these quivers, these arrows that you throw in your quiver, like

You know, it just, it's additive. And at some point you go like, I've cultivated this knowledge. I understand the things that I'm passionate about. understand how to do things well, how not to do things well. And the only limiting factor to starting a company at that point is your own sort of risk tolerance and drive. you know, it's like Quantium was an amazing company. I wish I had more time at the company. but at that point it was like,

I just had this desire to do something on my own. had worked in big corporations for a while. And so what do you do whenever you decide to call, like open your own company, you call your friends and you're like, Hey, do you want to start a company with me? So that's, ultimately, you know, how we got to where we are. And, you know, I think our experience really informed what we were doing. I think we found out very, very early on when we started the company.

Duncan Riley (20:13.523)

Yeah.

Jon Brewton - CEO Data² (20:28.514)

that if you apply graphs to generative AI interfaces, especially GPT models and transformer models in general, that you can change the efficiency curves of what you can get out of these systems and you can increase the transparency. Now graphs alone don't solve the whole problem, but they work as a fantastic grounding mechanism. And semantic graphs in general give you the ability to understand how to transition from flat text documents and how they work in structure and how they work with these systems.

Duncan Riley (20:33.586)

Yeah.

Jon Brewton - CEO Data² (20:58.68)

to a multimodal interface and how you can build these multimodal information repositories that really harmonize a lot of sort of really important data, but disconnected data. And we figured this out super early on. We started to do testing and we were like, okay, like if we can find a really smart way to generate graphs quickly and efficiently, we can really do something interesting here. And, you know,

That's kind of how we got here. It's like everything sort of leads to something else. The military gives you discipline, that discipline gives you the ability to get interesting jobs. You go back to school, like you do work as an engineer, you figure out ways to optimize that system, travel around the world, you find different ways to use technology, you go to school again, you find sort of new networks of people. And then all of a sudden, like you're running a company that is doing something really interesting, novel, and different than a lot of the folks that are doing sort of similar stuff.

Duncan Riley (21:57.242)

Yeah, definitely. mean, John, I speak to quite a few people and I think definitely what you're doing is probably one of the most unique things that I've not heard of before. that's why I was really excited to get you on the show and just sort of unearth where it's going, where's Data Squared going? And that's where I'm really excited to ask you, what's on the horizon? Because obviously, I guess AI is still in its infancy to a certain degree.

But every week is forever evolving and as a CEO, you must be thinking, how do we keep up with this? And especially when it's a large part of your product. Maybe it's a difficult question to ask, but what's on the horizon for data squared? Where do you see it going?

Jon Brewton - CEO Data² (22:45.826)

Yeah, look for data square, we've got a lot of activities. We're really focused on a couple of key vertical markets. It's defense, intelligence, engineering, finance, and then healthcare applications. We're really active in a couple of those places right now.

I think, where do I see data square going? The fun part about what we built is we built it through our own sort of trials and tribulations of trying to implement technology and really large scale enterprise companies that are really differentiated. So we built a platform that is, as far as I'm concerned, the most flexible, agile, and amenable platform to scale deployment ever.

It's completely cloud agnostic. It can be containerized, deployed in air-gapped environments. It's fully compatible with any LLM. It doesn't matter what data type you throw at it. It can be integrated with any existing systems. All we need to know is how to get to it. So can you stand up an API endpoint that we can access? And can you use somebody in your system or your company to tell us what's important and why it's important? So as long as we can define those things, we can connect to any legacy deployed systems.

It really is something that allows us to dream of limitless ways to apply it and limitless places to put it. But the best thing about it is it can work as the fundamental sort of platform approach for how you do work in these environments or.

It can really be something that sits alongside next to or on top of existing legacy systems that you're just trying to use in a different way. You're trying to take the sweat equity that you put into those systems and reprocess it into an environment that lets you use it at scale with other systems that you have deployed in your company. And I think like it's really the sky's the limit. We have a world-class partner network. We're tied up right now and really tangible go-to-market.

Jon Brewton - CEO Data² (24:39.566)

relationships with AWS, Microsoft, Dell, WWT, Neo4j, Nvidia. These are really large companies that are happy to lend us their credibility and help us sell alongside them. So the scale of the business hopefully goes up, but we can really apply it to just about anything. And ultimately, I think the next step is just for people to understand what is possible.

Like again, we're trying to show people there is a better way. We're trying to build a world where people and machines can make smarter decisions together. And I think it's more about compare and contrast, getting ourselves in front of people so that they understand. There is a way to eliminate hallucinations completely. There is a way to get explainable systems. There is a way to connect my legacy architecture, infrastructure and systems. There is a way to do this in a really smart, low cost, efficient manner so that you can start to use what you have.

right in front of you already, but in a smarter way. Like we like to tell the people that we talk to, look, you're not missing data. It's, it's, it's holistically impossible that you're missing data. Everybody's sort of drowning in data. There's lots of data exhausts, but you are missing is the connections in your data that exists and matter right now. And like, you have to reprocess that environment. You have to sort of rethink how you use that information to really find those connections, but those connections are sitting there and we want to be the critical unlock for people.

We want to make sure that they can actually figure out how to use their data more effectively, use their existing systems architecture in really smart and efficient ways, and then help people sort of grow into applying analytics and very specifically generative AI at scale in their businesses today. We've done really, really well so far in the oil and gas industry. We're next moving into the intelligence industry at scale. And then after that sort of finance, but

Duncan Riley (26:28.349)

Yep, really.

Jon Brewton - CEO Data² (26:31.746)

Yeah, there's sky's the limit. I think we're doing something different and interesting. you know, like even, you know, very recently it was talking to people in Australia and the folks that I know sort of at Quantium about the ways that you can sort of toggle, you know, in combination, the things that they're doing with the things that we do to create a better value proposition for the clients that, everybody's already working with. So I think it's just about, you know, fluency, getting out in front of people and trying to help them understand what's possible.

Duncan Riley (26:59.015)

Yeah. And for those that are watching right now and are thinking, yeah, I need to speak to John, how do they go about having a conversation with data?

Jon Brewton - CEO Data² (27:08.61)

Yeah.

It's really easy. We're a low tech, high tech company, meaning you can just reach out to me via email and I'll do my best to try to get back to you. But look, we have a website. It's data2.ai. There is a sort of form you can fill out there to let us know that you're interested in having a conversation. You can hit us up on LinkedIn. It's at data2us is the handle, but data square.

as a company, you can look me up at John Bruton, J-O-N-B-R-E-W-T-O-N, and you can email us at the company at contact, so contact at data2.ai. And so that's how you can get in touch with us. We're happy to have a conversation with you and explore what's possible.

Duncan Riley (27:55.951)

Excellent, excellent. And we'll leave all those details down in the comments as well so everyone can access that if they're watching on YouTube. But John, it's been an absolute pleasure having you on the show. I've been wanting to get you on for a while and what you've sort of explained today has been excellent and there's a lot of value in there. And I think for those that are watching, it's just thinking about how they can apply data squared to.

their organization, how can they bring their data to life and start getting some meaningful insight into their own data. Like you said, everyone's drowning in data, it's just all about how you access it. And DataSquare seems like the solution in the market right

Jon Brewton - CEO Data² (28:39.054)

Yeah.

Jon Brewton - CEO Data² (28:43.944)

One last note before we jump off. In our commercialized applications that we've deployed just in the last year, we've created around $39 million of tangible value for the people that we're building and deploying solutions for. So this isn't just some sort of fairy tale, let's apply AI and see what we can do with it. This is generating real business value. It's being very targeted in doing so.

That's what's sitting on the table for everybody out there. So if you're interested, please don't hesitate to reach out. We can tell you how we can get there.

Duncan Riley (29:15.571)

That's great. And also, I guess if you've got ERP systems or CRMs and you've got data in there that your teams are not utilizing, presume data squared would be an answer for that, being able to use natural language processing to pull that out and visualize it.

Jon Brewton - CEO Data² (29:34.775)

Absolutely.

Duncan Riley (29:35.923)

Brilliant. Well, John, thank you very much. And that brings us to the end of another episode. Thank you so much for tuning in, whether you've been watching along with us on YouTube or listening on Spotify. We truly appreciate you spending your time with us. If you enjoyed what you heard and saw today, please don't forget to hit that like button and subscribe so you don't miss out on future episodes. We've got some exciting things coming up. Until next time, speak soon.