 Welcome back everyone to SuperCloud Six. I'm John Furrier, your host. This episode's on AI innovators. This entire SuperCloud Six is unpacking the people leading the revolution in AI and also underpinning the infrastructure behind it, the software. Obviously the big way, big inflection point. We have a CUBE alumni here going back to 2012, Andrew Joyner, CEO of HyperScience is here. Andrew, thanks for coming on the program. John, you haven't changed a bit. Thanks for having me. You know, it's been a while. You know, we were just talking before we came on here about the big data revolution in the 2010 timeframe, Hadoop and all that jazz happened. And now you fast forward, that had happened. Now you got turned into spark. Now you got data bricks in the snowflake. You got fintech. Now you got all that machine learning and all that data came to the table and now this whole cloud scale kicks in. Now you have this entirely new inflection point around AI. So generative AI has actually created another renaissance of the big data revolution, taking it to the AI revolution. You guys are in the middle of it. You guys have a vision around how to apply AI to hyper automate environments end to end. Love that term, hyper automate. Automation is a big part of AI, whether it's writing code or doing mundane tasks. Hyper science, your company's in the middle of this. What do you guys do? And take a minute to explain what hyper science is doing and why you think you positioned well for the AI revolution. Yeah, thanks John. Look, if you go back to the big data days, really the best description of that information was everything that didn't fit into databases. And there was a lot of excitement around all that information that we could harness for the benefit of businesses. But it was really in those days around being able to store it, being able to manage it and then do some insights, maybe some insights so all that extra data. I think we've entered the era now where we can actually do software-based decision. We can actually drive benefit and drive automation off of that data. I think the LLMs are a great exposure to this. You know, the open AI revolution that sort of happened beginning of this past year was the best demo that we never created in hyper science, we like to say, because it really introduced this notion that AI is real and it can be a benefit to the enterprise. But the reality is what I see in chat TPP and open AI is that they've really helped teach machines to speak and write human. And hyper science was founded on a different premise. What we've taught our infrastructure to do is to read human-friendly information and to read at the accuracy and the flexibility of humans. We just do it at scale. Now, what business challenge does that solve? Where there's all this information that's trapped inside enterprises where you've taken approval of mortgages and it's gone through and it's got lots of long form contracts and bank statements. All of that stuff doesn't fit in the systems in the back office. So we have all these humans back there, back office workers who are still having to annotate, label and approve. Now we have a paradigm that we can read that information with accuracy and go ahead and start making decisions. So traditionally, the best we've been able to automate back in systems, John, is about 60% with hyper automation, with technologies like hyper science, we're getting 99.5% accuracy and 98% automation rates for the back office. You know, when I think of digitization of the enterprise, in my mind's eye, I think like, almost like these big machines connected to fiber optics, scanners, OCR, optical character recognition those days back then. But you're saying that human-friendly information is now being digitized at scale. That's what I'm kind of hearing. Is that what you're saying? Is that what you guys are doing or is it more of reading it in, ingesting it? What specifically are you guys doing? Well, first of all, define human-friendly information. What is it? And then how do you interact with it? How do you ingest that in? How do companies use that? Yeah, so the easy answer, John, is just like big data. That's everything that doesn't fit into a database. That's kind of what human-friendly information is. It's not structured for machines. It's structured that a human can read it. It's like a statement. It's a bank statement that you get. It's an invoice. You know what those pieces of content are. But machines haven't been able to read that because the layout's so flexible. So the big leap forward is hyper science from the beginning was a born ML company. Our founders were ML engineers. And we really leveraged the notion of computer vision. What we didn't want to do is we realized that optical character recognition, look, they've had 20 years to try to perfect it and it hasn't scaled past 60%. And it's because you have to rigidly define what it needs to look at. So you have to define the quadrant of the page. You have to look for patterns. Computer vision, you train a model and it reads it like a human. And it starts to recognize this is an invoice. And even though the total's in a different area, I know that this should be paid. And so it forms an understanding of the information. And we've gotten to the point where we have core models that understand the basics of human. So we can read handwriting, for instance. We can read if a page has been rotated or skewed. We can know if it's got a smudge or something's been scribbled out. So that comes native in the system. We've trained models on big data sets. But now what you can do is train it from scratch on your business language, what's unique to your business. And so at about five to 20 documents, you can now teach it just kind of like a new employee. Teach it the language of your business and that's how you get high rates of automation. I love that language of your business angle. That's so on point because you think about it, all that stuff is well known. You got contracts, you got sales contracts, you got legal stuff, you got all kinds of things that are going on in the back office. Okay, it could be monotonous. But when you think about converting it fast with scale, okay, with the hyper automation, which I love, then you get into, okay, now you have the data. Then you got the LLMs out there, the open AIs, the Anthropics, the Cohers of the world. So now you got the large language models, the proprietary models. Now you got the open, absolutely frontier models, they call them now. And you have the companies have their own models. As they start getting their models in, we call that the power law, where you have the start to see them curve. Smaller but proprietary, let's say proprietary company-specific data. You guys are in the middle of this. Can you explain how you look at the frontier models and you call them sovereign models? Can you explain that? Because this is where we're seeing the most action when the enterprise is, they realize that they have their own data and that's not mutually exclusive with working with the other models. So models with models are kind of happening. So can you share your vision? Yeah, John, you described it well. This space is evolving very, very quickly. And we're seeing about four emergent vendors now who are really producing these frontier models that are getting into the billions and billions of parameters. And I think they'll quickly pass into trillions of parameters. And while they're going to be amazing technologies for enterprises, they only solve somewhat of a subset of the pie. With every great enabling technology, getting it into the core of the enterprise is tough because you have to meet security requirements. You have to meet transparency requirements, data handling. There's new things that are being introduced that are really important for it to get used at the core of the enterprise. The data that hyper science works with that sits in the back office hasn't been seen by these large language models. And a lot of companies don't want to contribute that IP which has a lot of sensitive information to the frontier models. And so what they want is they want a world that works with their own specialized IP. We call them sovereign models because the term kind of denotes a connotation of borders. You understand the borders of what you're contributing. You understand it with full control. You have full control over everything in that. And that's a core premise of hyper science. Hyper science can run on premise at a customer in an air-gapped environment if we need to. We can also run in your cloud or we can run in public clouds. So we offer all three deployments because we wanted the enterprises to have the confidence that they had full control over their data and their IP and their customer information. And so that's really the big difference is what we see is the frontier models right now are getting a tremendous amount of traction at the edge of the enterprise where there's not a lot of sensitivity. Chat agents on your website summarizing things from the contact center. But when you get into the core of the enterprise, you need control of your data. You need transparency, you need trustworthiness. And so you need infrastructures that balance that and give you the end-to-end transparency over who trained the model. What was the ground truth data that was trained? So that if you're giving patient advice or something like you not have full traceability over what you provided. You're basically enabling companies to do their own sovereign models basically and leverage that data out of the gate. That's exactly right. And we like to see lots of different models for different types of situations. So one of our largest customers is the Veteran Affairs Organization, a massively complex organization, over 300,000 people, the second largest government agency. And by the way, pretty important to all of us, right? Making sure we get quick care, quick reimbursements to veterans. There's over nine million of them in the US. It's complex. And it used to take months, as much as three months for any submission to work its way through. It's prescription information. It's medical information. It's from all different types of hospitals and doctors. Very complex. They put hyper science at the core. We're reading over a billion documents a year now for the Veteran Affairs. It's every third party upload, every piece of documentation that goes through. We read it at 99.5% accuracy and the automation rates are at 98%. So what used to take three months is now taking three hours. It's stunning. And what's amazing is there were over 14,000 people who were responsible, touching and making this information flow. They can now focus on higher priority tasks. All of that savings now benefits that they can reinvest that in other areas of the VA to make the experience with our veterans better. So it really starts to show that when you get AI into the enterprise, there's real ROI. Like we are truly automating things, not just at the edge, but at the core. It's just, and that's like a public sector example. And you look at these environments that are old and antiquated, outdated. This is a step function, instant upgrade. I mean, because they have old back office stuff. Contracts that you mentioned, human friendly, I don't know, human friendly. It's like a pile of data now, but it's old school stuff, instantly upgraded. Okay, so this is a great example of the impact, one, getting people modernized on a transformation journey check. But the aerial impact is the impact of the workforce. So what does that do for the company? Now that they got instantly upgraded, they now have AI enabled with the data. What happens next? How are you guys taking that to the next level? What are you seeing with those customers? What are they doing next? Because by the way, this is not just public sector. It's all companies get the upgrade, but. But that's what's so interesting about this market is that when we first presented hyper science to the investment community, they compared us to a lot of legacy technology markets. And what I looked at is I took over hyper science is that actually what we're actually enabling and automating a broader chain, because what used to be a 50 to 60% automation rate, the classic thing to do next was to hand it to a BPO. And you try to then handle all the exceptions and all the manual tasks through offshore, lower cost labor. Well, the US government can't do that. So in many ways, it's not surprising that the US government who is really leaned into AI and warfare tactics and logistics is used to buying AI technology. They understand the data requirements. They're actually the leading market right now. They're the leading buyer of AI. And so now they can look at it holistically, not just what technology to replace and get a little incremental benefit. It's, I used to have to use humans for all of these exceptions. And so now we bring that into the enterprise. And I'll give you an example, Oracle. I spoke to Jeff Epstein who was the former CFO who Oracle has much access to as much business, you know, back office software as anyone on the planet. Yet still in the CFO office, there's 6,300 people in that office. So we still have not automated a lot in the back office. And I think a lot of it is because the complexity of this information continues to have to require human intervention. And that's what the AI premise is allowing us to eliminate. And in the few minutes we have left, talk about how you see the innovation in this market for your company and your customers. How should they be thinking about how they go forward architecturally from an infrastructure standpoint, from a software standpoint, as they organize their business, a new operating model is emerging. Do you see a pattern here as an innovator? Is there a certain playbook that you see evolving? What's your vision? What's your advice perfect for the innovators out there? Yeah, great question, John. So I have a saying that I like to say to our head of comms and I say, look, I don't think AI is going to replace workers or humans. But I think the AI workers that use AI will replace the ones who aren't using AI. And I think that extends to businesses that I think that the businesses that are going to use AI are going to replace the ones that don't embrace it. Now the governance, the security, the compliance, it is complex and it is new. But there is a bridge that's evolving with companies like HyperScience that we're now trusted and proven by some of the largest companies and the most highly regulated enterprises in the world. And so my first recommendation is you've got to embrace this. You've got to embrace this at all elements of the business. It is profound. The automation rates, the understanding, I think it's going to accelerate your business and allow you to reinvent it in areas that'll give you competitive advantage. But you've got to work with companies that are trusted and proven. I think we're getting comfortable with the frontier providers. Those are some of the best companies in tech. But there are companies like HyperScience that are also solving, getting enterprise into the core of your business. And I think that that is the real promise is that you have to widen your lens and widen the way you think about technology. It's far broader than just a simple replacement of an existing technique that you have. But I think this, John, is going to be the biggest trend. We've had the biggest transformation of anything that you and I have been working together and seen. It's one of those moments where it's so much fun but so much action. But this risk reward, if you're not on the front, if you're too far forward, you're going to be driftwood. As we say, this wave crashes on you. You're going to be toast. If you wait too long, you're going to miss it. So there's kind of a nice challenge. If you're not on the edge, you've got too much room, right? Exactly. Andrew, great to see you. And thanks for coming on SuperCloud Six, our AI innovators, focused. Thanks and congratulations on the growth you guys have. It's hyper growth for hyper science. Hyper growth, that's what we're seeking, John. Thanks again for having us. Good to see you. I know we see more of you often in the future. So thank you. Thanks for coming on. Okay, that's SuperCloud Six. We'll be right back with more after this short break. Stay with us.