theCUBE + NYSE Wired: Mixture of Experts Series | Markie Wagner, Poetic

Clips
More from theCUBE + NYSE Wired: Mixture of Experts Series

Markie Wagner

Founder & CEO

Poetic

play_circle_outline $50M Series A funding led by Founders Fund, Kleiner Perkins, First Harmonic

play_circle_outline Tokenmaxxing the Bottom Line: 10x Token Efficiency and Economic Value of Lean AI Workflows

play_circle_outline Whispers into Code: Minimal-Token AI Transforms Procedures into Executable Enterprise Workflows with Human-in-the-Loop Refinement

play_circle_outline Achieving 99% Accuracy in Multi-Hour Enterprise Processes: Use Cases for Banks, Insurers, and Telcos

Info

Markie Wagner, Poetic

Markie Wagner

Founder & CEO Poetic

search

John Furrier

>> Palo Alto studio connections, Silicon Valley and Wall Street. I'm John Furrier, here with David Vellante, my co-host. Hello, I'm John Furrier with theCUBE here at our NYSE Wired studios in New York City. Of course, we have our Palo Alto studio connecting Silicon Valley to Wall Street. We've got Markie Wagner here, founder and CEO Poetic, just raised $50 million. Coming from a deep tech background in AI and connecting in to the enterprise area, workflows, hyperscalers, all distributed computing, hybrid cloud with AI accelerating, tons of developer value, tons of enterprise value. Markie, thanks for coming on theCUBE's NYC Wired program.

Markie Wagner

>> Thank you for having me.

John Furrier

>> So let's get into it. Just you got the big fat funding. It's not actually kind of a small series A, you consider all the monster series A's out there, but it's a really focused mission. Talk about the company's mission and then let's talk about the funding and what you're going to do with it.

Markie Wagner

>> Yeah. So a little bit about Poetic. So we're an AI system that can learn and execute extremely complex process at some of the biggest companies in the world with over 99% accuracy and 10 times less tokens than a traditional agent.

John Furrier

>> So talk about the funding that closed yesterday, you guys announced or announced it yesterday?

Markie Wagner

>> Yes, announced yesterday. Yeah.

John Furrier

>> Okay. How'd that go? Was you on a press tour? How many interviews have you done? You must have done a lot of interviews.

Markie Wagner

>> Yeah. Look, I mean, we were on Bloomberg yesterday and I've been lucky to have great conversations with folks at TBPN and things like that. But how it came together, I mean, I think Founders Fund and Kleiner Perkins have been friends of ours for a long time and backers from the earliest days. Founders Fund led the seed round. And so Kleiner and Founders Fund and as well as First Harmonic, which is Ali Rowghani's Fund and other great folks like Genius Ventures and lots of wonderful angels came together for the A.

John Furrier

>> Yeah, that's great validation. It's a great testament to the mission. Talk about the origination story. How'd this all come together?

Markie Wagner

>> Yeah. I mean, background on me. So I actually got my start as a researcher. I've been into AI for quite some time after reading too much sci-fi as a kid. But eventually I found my way to Stanford and I was a researcher and working at Waymo and Google and sort of all these places. But at a certain point, sort of realized that while it's fun to write papers, it's really important to understand what's going on in the real world. If you don't really know what the problems are, what people are doing all day, you're not going to be able to solve them. And so I'd actually dropped out to do the Thiel Fellowship, not even start a company, but just to go and work in some of these really, really old industries and just realized over and over again that most of the work that happens in these places are actually just running on operating procedures. And so people will go and execute these operating procedures all day. These things will be 200 pages and then there's tens of thousands of secret rules. And if you looked around everywhere, banking, insurance, healthcare, it was still people doing these procedures and code hadn't really been able to touch them. And so that was the genesis of deciding to build what would become Poetic.

John Furrier

>> These enterprises, they have these brittle systems that have been around the legacy. I even see some of the cloud stuff as I call it modern legacy. You're starting to see the moves-

Markie Wagner

>> Modern legacy.

John Furrier

>> Yeah, which is the only way to talk about the kind of legacy, but it's modern. But the static workloads, RPA for instance, we've been covering UiPath for years. They're pivoting to more non-deterministic. That static rules, that's how it was done. And now you're starting to see a lot of the AI force, that legacy environment and the modern legacy to say, "Okay, you got to change.:" And their reaction is, "Whoa, we really don't want it." But now with tokens driving value creation, everyone sees it, they're using it, budgets are out of control. I just came back from FinOps for three days live-streaming there. We interviewed 27 practitioners. Look, they're all like, "Look, we got to do more of this. We got to get a handle around this." So understanding what a premium token is, how to manage tokens really is about the function of the work. That directly impacts the workflows, the agentic architecture. What are your thoughts on that? Because everyone's talking tokens. Tokenmaxxing is just the leaderboard. It's like writing a lot of code, but might not do anything. Tighter tokens, probably more the leaderboard we're going to look at soon. "Wow, you use less tokens, you're better." So that's going to be, I think, the new standard. What's your thoughts on this token enterprise architecture integration?

Markie Wagner

>> Yeah. I think we're in the early days. Something that we noticed is at least our vantage point is AI is obviously incredible at writing code and yet even though we're spending money on all these tokens, the impact isn't quite there in the way that we thought that there would be in the enterprise at this point. And the reason why is because AI can only work on what it can touch. And most of the rules that run these businesses are still in people's heads, far beyond the reach of any kind of digital intelligence. And so we see how do you actually make AI useful and use a token in a way that's really valuable? There's two pieces. One is for a task that you want AI to do, can it learn all the rules? Not the ones that are written down, but the 10,000 secret ones that are the things that really actually make that task successful. And then can it follow it every single time? Can it follow those rules and not skip a step? And if the system can learn those rules and then execute them with over 99% accuracy, then you kind of have built the bridge from where we are now and AI existing to being useful and shaped into something that is economically very valuable. And if you miss either one of those, either you don't get the rules or you don't follow them, no one's going to trust it and thus nobody's going to scale it and then the impact's not going to be there. And so we see ourselves as that bridge.

John Furrier

>> And data is key. If you don't have the data, then the accuracy drops, then you got to get the software to kind of rethink that. Agents don't want hallucinations. No one in the enterprise wants hallucinations. They don't want to have really good, bad answers. They don't want that to be the case. What are you guys doing from a value proposition standpoint? What is the actual service product? What are you offering? Because 99% is incredible performance. That's pushing the envelope on the frontier of the performance, certainly in enterprise. Now, is it a data problem? Is it a system problem? How would you scope that and what is the product?

Markie Wagner

>> Yeah. So the way that we see the product's role is really in that learning and execution. So what happens is a customer of ours will have a giant operating procedure. We work with some of the biggest companies on the planet, whether it's top banks, top three insurance companies like an AIG, top telcos, financial services like SoFi and Chime, they will have these operating procedures, which they'll upload into us and we'll generate this sort of AI operating procedure, which is written in step-by-step English. Our system will turn that into code under the hood. So when it runs, there's almost no tokens. Everything that can become code does. It's even cheaper than a fast model or a small model or an open source model, it's code. If it can be code, it should be. And it runs it. And what happens is after it runs, there's experts who will give feedback to say, "You forgot this rule." You forgot, this is the threshold for this dollar amount." And after enough of that loop, you eventually have all those rules written as English, but running as code and you're off to the races. And so really it's that process of getting the true SOP, the true operating procedure out of people's heads and then running it as code. And AI's role is not to run it and figure out what to do as it goes, but to repair it if it needs to. So that's sort of how we see it.

John Furrier

>> So what you're saying is that people are burning tokens to get things into code, that's a problem. You guys solve that problem, right?

Markie Wagner

>> Yeah. So yeah-

John Furrier

>> So you put it into code without tokens. Then you use the tokens for real AI support, repair, reasoning, intelligence. Is that right?

Markie Wagner

>> Yeah, just about. And I think maybe I'll differentiate it from the traditional agent. The way that an agent works today is you give it a goal and it figures out the path there. It's sort of thinking as it goes. If an agent was making a sandwich, it would go to the refrigerator and say, "Huh, should I put lettuce in the sandwich? Maybe not. Should I put a tomato in the sandwich?" And it's really thinking its way through every time. On the other side, with us, it's sort of like it's just following the recipe. So there's not much thinking involved. There's not much AI involved when it runs. Rather than refiguring out the answer, it'll figure it out once and then keep it as that code under the hood. And so to your point, yes, it generates a lot less code because once it has the answer, it just uses it. It's not reinventing the wheel every single time.

John Furrier

>> What do you think about my comment about the token leaderboard shifting to the people who have used less tokens and more output versus the tokenmaxers? Do you think that's going to be the flip? Do you agree with that?

Markie Wagner

>> Yeah. I think it's how much impact can you get with the least cost. I think that's sort of always been what's rewarded is how can you be more efficient with every single bit of token that comes through. And so I do see we get to work with some of the biggest CEOs on the planet and watching them shift from, "How can we use as many tokens as possible?" To, "How do we get as much impact as possible and make each token efficiently used?" Has been quite fascinating.

John Furrier

>> Market, talk about the customers. Take us through some of the early engagements you're doing and the business you're doing now. What is a day in the life? Take us through the engagement and the reaction. Are they skeptical or like, "Whoa, 99%. We have our own knowledge base systems here. We have rules-based databases, we got graphs, we got it covered." I mean, people are really trying different approaches. So when you come in with a new thing that's actually better, they get blown away, they go right to it. Some will... Take us through some of those engagements and what's it been like and what's been the outcome?

Markie Wagner

>> Yeah, I think it was different in the early days. I think people always just only believe results when they see it. There was a quote from Peter at AIG talking about how we saw over 99% accuracy on a multi-hour process, an extremely difficult process. And so in the beginning, I think customers are wondering if it's possible, but after it's possible, you win a lot of trust. And so the kind of customers that we work with, which are some of the biggest banks, telcos, insurance companies, they really do know that what's most important with getting these tasks into production, it's that accuracy. And luckily, a lot of that's measurable, very measurable before you reach production. And so I'd say there's before and you haven't seen it, you don't believe it. And then once you see it, you think the world's your oyster.

John Furrier

>> One thing about the enterprise is that people actually know what success looks like, but they never could do it. And if they did it, they did it in a half-ass way or in a very old-fashioned way, slower way. This is a new era. And you're starting to see how people do pilots are different. I'm seeing people saying, "Screw the IT philosophy of doing a project a certain way." They want to go right for the hardest problem because they know the problem. They just couldn't get to it in the old IT way. And so these domain experts, they know what they're doing. What's your thought on that? Do you take that same approach when you go in? Are they reacting that way? "Okay, we'll throw you the hardest problem. Go."

Markie Wagner

>> Yes, we are. And that's exactly what we do inside of these businesses. I think there's a whole class of problems that have been white whales for these enterprises, things like reconciliation that have been unsolvable with code alone over the past couple of decades. And I think people know where the biggest problems are. They've just never had the tools to solve them. And now that they do, customers quickly want to go and figure it out on what is the process or job that is the heart of the business that if you automated it, it would mean a tremendous amount to cost or revenue or wherever it is. I think in the beginning there was an appetite to try things that are small, but now that we've seen wins and the industry's seen wins, I think there's a lot more appetite to go and work on the most important problems.

John Furrier

>> Yeah, that's definitely . We agree. Talk about the projects you're working on now. What are they like? What are some of the engagements and how would I get involved? I got a lot of complex, I'm an enterprise. I want to bring you in. How do I think about that? What's the conversation like? How do you engage with the customer? You say, "Give me your hardest problem." Do you come in with a team? Is it just a couple clicks?

Markie Wagner

>> Yeah. In the beginning, we'll often work with a leader at the customer to identify, "Hey, what is that one task, that one procedure that if it was turned into software, it would matter tremendously to the business." And so after that, it becomes more of a game of collecting all the documents as they exist today that define that job, again, generating that AI operating procedure and then getting the results in front of experts so that they can almost give that feedback, that data labeling into it to get that English document to the place where all the rules have been written down and that you are at that really high level of performance. And so that's typically what the engagement looks like. I mean, it's quite hands-on. I think even if you have the perfect product, you still have to get people to use it and you can have the best AI ever and there's still a change management component that's really quite real in order to make sure the value's realized. So I think a lot of folks will stop short of the last mile. We think that the most important part really is that last mile, that gap between 80% and 99% and running in production and value realized every single day. And there's a big gap there, but that's what we work on.

John Furrier

>> You know what's great about what you're doing is that every company has this. They write everything down and you actually bring it to life and put intelligence into it. Talk about what happens after. Is it an always on AI intelligent layer using AI to look at it, see what's visible in the network or the software, you're observing it, you're evaluating it? Take us through what happens once it's up and running.

Markie Wagner

>> So once it's up and running, you have a document that is the source of truth for that process in your business, whether it's anti-money laundering or fraud investigations or reconciliation or customer onboarding. What was a bunch of operating procedures is now this AI operating procedure. And so what happens is every day it runs and the experts who know the most about that job will give pieces of feedback into it and that cycle goes on forever and ever and ever. It's always sort of this aligning what AI is doing with what the desired behaviors are. And then what becomes really fun is you can experiment and say, "What if we did this job differently?" What if the period of review was twice as long, what would happen? And what people do is they experiment. They can ask these questions, "What if we change how the job is, test it on all the history and see the difference?" So that is often what becomes the point of focus and you're experimenting to find the best process over time.

John Furrier

>> That's awesome. And I'm hearing some people even going kind of in the rabbit hole by saying, "Hey, let's build a digital twin and let's go figure out, let's do more." Once they get online or AI enabled, more cool stuff can come out of it. So I mean, I can imagine a ton of headroom for you, which brings me to the $50 million question. What are you going to spend it on? You got some fat financing, congratulations. What are your priorities? What are you optimizing for?

Markie Wagner

>> Yeah. I mean, I think it's more of the same of what we've been doing, which is finding the best people on the planet, going to these industries that have existed for an extremely long time, like banking, insurance, and figuring out, "Hey, what does this entire business unit look like after it's been turned into software? What does fraud investigations look like after it's been turned into software?" And figuring that out with our customers. And so continuing to scale, continuing to find the best people on the planet and solving some of the hardest problems that have existed for hundreds of years in these industries that are extremely important.

John Furrier

>> Yeah. And a lot of these industries are regulated. They have the data, they just don't know what to do with it. They tagged it for the right reasons. There's lots sitting around. Great opportunity. We love the fact that you're going to come in and bring intelligence into the enterprise. Final question for you, talk about scale. A lot of these enterprises are operating at scale. What's your thoughts on scale? Do you just ride their scale equation, their AI infrastructure? Do you have something on your side? What's the relationship on the scale side?

Markie Wagner

>> On the scale side, I mean, the only way to create impact at scale is to go and find your own ways to scale. And so we have deep partnerships with a lot of the labs and folks that have a lot of distribution, but ultimately we have to sort of scale to meet the moment as well.

John Furrier

>> Well, maybe to be the brain of the enterprise, Markie, thanks for coming on theCUBE's NYC Wired program. Great. Congratulations and we'll see you around. Thanks for coming on.

Markie Wagner

>> Thank you. All right. See you. Bye.

John Furrier

>> I'm John Furrier with theCUBE. Mixture of experts here at the NYC Wired program and CUBE original, of course. We got our Palo Altos too connecting Silicon Valley and Wall Street. Thanks for watching.

Markie Wagner, Poetic

search

John Furrier

>> Palo Alto studio connections, Silicon Valley and Wall Street. I'm John Furrier, here with David Vellante, my co-host. Hello, I'm John Furrier with theCUBE here at our NYSE Wired studios in New York City. Of course, we have our Palo Alto studio connecting Silicon Valley to Wall Street. We've got Markie Wagner here, founder and CEO Poetic, just raised $50 million. Coming from a deep tech background in AI and connecting in to the enterprise area, workflows, hyperscalers, all distributed computing, hybrid cloud with AI accelerating, tons of developer value, tons of enterprise value. Markie, thanks for coming on theCUBE's NYC Wired program.

Markie Wagner

>> Thank you for having me.

John Furrier

>> So let's get into it. Just you got the big fat funding. It's not actually kind of a small series A, you consider all the monster series A's out there, but it's a really focused mission. Talk about the company's mission and then let's talk about the funding and what you're going to do with it.

Markie Wagner

>> Yeah. So a little bit about Poetic. So we're an AI system that can learn and execute extremely complex process at some of the biggest companies in the world with over 99% accuracy and 10 times less tokens than a traditional agent.

John Furrier

>> So talk about the funding that closed yesterday, you guys announced or announced it yesterday?

Markie Wagner

>> Yes, announced yesterday. Yeah.

John Furrier

>> Okay. How'd that go? Was you on a press tour? How many interviews have you done? You must have done a lot of interviews.

Markie Wagner

>> Yeah. Look, I mean, we were on Bloomberg yesterday and I've been lucky to have great conversations with folks at TBPN and things like that. But how it came together, I mean, I think Founders Fund and Kleiner Perkins have been friends of ours for a long time and backers from the earliest days. Founders Fund led the seed round. And so Kleiner and Founders Fund and as well as First Harmonic, which is Ali Rowghani's Fund and other great folks like Genius Ventures and lots of wonderful angels came together for the A.

John Furrier

>> Yeah, that's great validation. It's a great testament to the mission. Talk about the origination story. How'd this all come together?

Markie Wagner

>> Yeah. I mean, background on me. So I actually got my start as a researcher. I've been into AI for quite some time after reading too much sci-fi as a kid. But eventually I found my way to Stanford and I was a researcher and working at Waymo and Google and sort of all these places. But at a certain point, sort of realized that while it's fun to write papers, it's really important to understand what's going on in the real world. If you don't really know what the problems are, what people are doing all day, you're not going to be able to solve them. And so I'd actually dropped out to do the Thiel Fellowship, not even start a company, but just to go and work in some of these really, really old industries and just realized over and over again that most of the work that happens in these places are actually just running on operating procedures. And so people will go and execute these operating procedures all day. These things will be 200 pages and then there's tens of thousands of secret rules. And if you looked around everywhere, banking, insurance, healthcare, it was still people doing these procedures and code hadn't really been able to touch them. And so that was the genesis of deciding to build what would become Poetic.

John Furrier

>> These enterprises, they have these brittle systems that have been around the legacy. I even see some of the cloud stuff as I call it modern legacy. You're starting to see the moves-

Markie Wagner

>> Modern legacy.

John Furrier

>> Yeah, which is the only way to talk about the kind of legacy, but it's modern. But the static workloads, RPA for instance, we've been covering UiPath for years. They're pivoting to more non-deterministic. That static rules, that's how it was done. And now you're starting to see a lot of the AI force, that legacy environment and the modern legacy to say, "Okay, you got to change.:" And their reaction is, "Whoa, we really don't want it." But now with tokens driving value creation, everyone sees it, they're using it, budgets are out of control. I just came back from FinOps for three days live-streaming there. We interviewed 27 practitioners. Look, they're all like, "Look, we got to do more of this. We got to get a handle around this." So understanding what a premium token is, how to manage tokens really is about the function of the work. That directly impacts the workflows, the agentic architecture. What are your thoughts on that? Because everyone's talking tokens. Tokenmaxxing is just the leaderboard. It's like writing a lot of code, but might not do anything. Tighter tokens, probably more the leaderboard we're going to look at soon. "Wow, you use less tokens, you're better." So that's going to be, I think, the new standard. What's your thoughts on this token enterprise architecture integration?

Markie Wagner

>> Yeah. I think we're in the early days. Something that we noticed is at least our vantage point is AI is obviously incredible at writing code and yet even though we're spending money on all these tokens, the impact isn't quite there in the way that we thought that there would be in the enterprise at this point. And the reason why is because AI can only work on what it can touch. And most of the rules that run these businesses are still in people's heads, far beyond the reach of any kind of digital intelligence. And so we see how do you actually make AI useful and use a token in a way that's really valuable? There's two pieces. One is for a task that you want AI to do, can it learn all the rules? Not the ones that are written down, but the 10,000 secret ones that are the things that really actually make that task successful. And then can it follow it every single time? Can it follow those rules and not skip a step? And if the system can learn those rules and then execute them with over 99% accuracy, then you kind of have built the bridge from where we are now and AI existing to being useful and shaped into something that is economically very valuable. And if you miss either one of those, either you don't get the rules or you don't follow them, no one's going to trust it and thus nobody's going to scale it and then the impact's not going to be there. And so we see ourselves as that bridge.

John Furrier

>> And data is key. If you don't have the data, then the accuracy drops, then you got to get the software to kind of rethink that. Agents don't want hallucinations. No one in the enterprise wants hallucinations. They don't want to have really good, bad answers. They don't want that to be the case. What are you guys doing from a value proposition standpoint? What is the actual service product? What are you offering? Because 99% is incredible performance. That's pushing the envelope on the frontier of the performance, certainly in enterprise. Now, is it a data problem? Is it a system problem? How would you scope that and what is the product?

Markie Wagner

>> Yeah. So the way that we see the product's role is really in that learning and execution. So what happens is a customer of ours will have a giant operating procedure. We work with some of the biggest companies on the planet, whether it's top banks, top three insurance companies like an AIG, top telcos, financial services like SoFi and Chime, they will have these operating procedures, which they'll upload into us and we'll generate this sort of AI operating procedure, which is written in step-by-step English. Our system will turn that into code under the hood. So when it runs, there's almost no tokens. Everything that can become code does. It's even cheaper than a fast model or a small model or an open source model, it's code. If it can be code, it should be. And it runs it. And what happens is after it runs, there's experts who will give feedback to say, "You forgot this rule." You forgot, this is the threshold for this dollar amount." And after enough of that loop, you eventually have all those rules written as English, but running as code and you're off to the races. And so really it's that process of getting the true SOP, the true operating procedure out of people's heads and then running it as code. And AI's role is not to run it and figure out what to do as it goes, but to repair it if it needs to. So that's sort of how we see it.

John Furrier

>> So what you're saying is that people are burning tokens to get things into code, that's a problem. You guys solve that problem, right?

Markie Wagner

>> Yeah. So yeah-

John Furrier

>> So you put it into code without tokens. Then you use the tokens for real AI support, repair, reasoning, intelligence. Is that right?

Markie Wagner

>> Yeah, just about. And I think maybe I'll differentiate it from the traditional agent. The way that an agent works today is you give it a goal and it figures out the path there. It's sort of thinking as it goes. If an agent was making a sandwich, it would go to the refrigerator and say, "Huh, should I put lettuce in the sandwich? Maybe not. Should I put a tomato in the sandwich?" And it's really thinking its way through every time. On the other side, with us, it's sort of like it's just following the recipe. So there's not much thinking involved. There's not much AI involved when it runs. Rather than refiguring out the answer, it'll figure it out once and then keep it as that code under the hood. And so to your point, yes, it generates a lot less code because once it has the answer, it just uses it. It's not reinventing the wheel every single time.

John Furrier

>> What do you think about my comment about the token leaderboard shifting to the people who have used less tokens and more output versus the tokenmaxers? Do you think that's going to be the flip? Do you agree with that?

Markie Wagner

>> Yeah. I think it's how much impact can you get with the least cost. I think that's sort of always been what's rewarded is how can you be more efficient with every single bit of token that comes through. And so I do see we get to work with some of the biggest CEOs on the planet and watching them shift from, "How can we use as many tokens as possible?" To, "How do we get as much impact as possible and make each token efficiently used?" Has been quite fascinating.

John Furrier

>> Market, talk about the customers. Take us through some of the early engagements you're doing and the business you're doing now. What is a day in the life? Take us through the engagement and the reaction. Are they skeptical or like, "Whoa, 99%. We have our own knowledge base systems here. We have rules-based databases, we got graphs, we got it covered." I mean, people are really trying different approaches. So when you come in with a new thing that's actually better, they get blown away, they go right to it. Some will... Take us through some of those engagements and what's it been like and what's been the outcome?

Markie Wagner

>> Yeah, I think it was different in the early days. I think people always just only believe results when they see it. There was a quote from Peter at AIG talking about how we saw over 99% accuracy on a multi-hour process, an extremely difficult process. And so in the beginning, I think customers are wondering if it's possible, but after it's possible, you win a lot of trust. And so the kind of customers that we work with, which are some of the biggest banks, telcos, insurance companies, they really do know that what's most important with getting these tasks into production, it's that accuracy. And luckily, a lot of that's measurable, very measurable before you reach production. And so I'd say there's before and you haven't seen it, you don't believe it. And then once you see it, you think the world's your oyster.

John Furrier

>> One thing about the enterprise is that people actually know what success looks like, but they never could do it. And if they did it, they did it in a half-ass way or in a very old-fashioned way, slower way. This is a new era. And you're starting to see how people do pilots are different. I'm seeing people saying, "Screw the IT philosophy of doing a project a certain way." They want to go right for the hardest problem because they know the problem. They just couldn't get to it in the old IT way. And so these domain experts, they know what they're doing. What's your thought on that? Do you take that same approach when you go in? Are they reacting that way? "Okay, we'll throw you the hardest problem. Go."

Markie Wagner

>> Yes, we are. And that's exactly what we do inside of these businesses. I think there's a whole class of problems that have been white whales for these enterprises, things like reconciliation that have been unsolvable with code alone over the past couple of decades. And I think people know where the biggest problems are. They've just never had the tools to solve them. And now that they do, customers quickly want to go and figure it out on what is the process or job that is the heart of the business that if you automated it, it would mean a tremendous amount to cost or revenue or wherever it is. I think in the beginning there was an appetite to try things that are small, but now that we've seen wins and the industry's seen wins, I think there's a lot more appetite to go and work on the most important problems.

John Furrier

>> Yeah, that's definitely . We agree. Talk about the projects you're working on now. What are they like? What are some of the engagements and how would I get involved? I got a lot of complex, I'm an enterprise. I want to bring you in. How do I think about that? What's the conversation like? How do you engage with the customer? You say, "Give me your hardest problem." Do you come in with a team? Is it just a couple clicks?

Markie Wagner

>> Yeah. In the beginning, we'll often work with a leader at the customer to identify, "Hey, what is that one task, that one procedure that if it was turned into software, it would matter tremendously to the business." And so after that, it becomes more of a game of collecting all the documents as they exist today that define that job, again, generating that AI operating procedure and then getting the results in front of experts so that they can almost give that feedback, that data labeling into it to get that English document to the place where all the rules have been written down and that you are at that really high level of performance. And so that's typically what the engagement looks like. I mean, it's quite hands-on. I think even if you have the perfect product, you still have to get people to use it and you can have the best AI ever and there's still a change management component that's really quite real in order to make sure the value's realized. So I think a lot of folks will stop short of the last mile. We think that the most important part really is that last mile, that gap between 80% and 99% and running in production and value realized every single day. And there's a big gap there, but that's what we work on.

John Furrier

>> You know what's great about what you're doing is that every company has this. They write everything down and you actually bring it to life and put intelligence into it. Talk about what happens after. Is it an always on AI intelligent layer using AI to look at it, see what's visible in the network or the software, you're observing it, you're evaluating it? Take us through what happens once it's up and running.

Markie Wagner

>> So once it's up and running, you have a document that is the source of truth for that process in your business, whether it's anti-money laundering or fraud investigations or reconciliation or customer onboarding. What was a bunch of operating procedures is now this AI operating procedure. And so what happens is every day it runs and the experts who know the most about that job will give pieces of feedback into it and that cycle goes on forever and ever and ever. It's always sort of this aligning what AI is doing with what the desired behaviors are. And then what becomes really fun is you can experiment and say, "What if we did this job differently?" What if the period of review was twice as long, what would happen? And what people do is they experiment. They can ask these questions, "What if we change how the job is, test it on all the history and see the difference?" So that is often what becomes the point of focus and you're experimenting to find the best process over time.

John Furrier

>> That's awesome. And I'm hearing some people even going kind of in the rabbit hole by saying, "Hey, let's build a digital twin and let's go figure out, let's do more." Once they get online or AI enabled, more cool stuff can come out of it. So I mean, I can imagine a ton of headroom for you, which brings me to the $50 million question. What are you going to spend it on? You got some fat financing, congratulations. What are your priorities? What are you optimizing for?

Markie Wagner

>> Yeah. I mean, I think it's more of the same of what we've been doing, which is finding the best people on the planet, going to these industries that have existed for an extremely long time, like banking, insurance, and figuring out, "Hey, what does this entire business unit look like after it's been turned into software? What does fraud investigations look like after it's been turned into software?" And figuring that out with our customers. And so continuing to scale, continuing to find the best people on the planet and solving some of the hardest problems that have existed for hundreds of years in these industries that are extremely important.

John Furrier

>> Yeah. And a lot of these industries are regulated. They have the data, they just don't know what to do with it. They tagged it for the right reasons. There's lots sitting around. Great opportunity. We love the fact that you're going to come in and bring intelligence into the enterprise. Final question for you, talk about scale. A lot of these enterprises are operating at scale. What's your thoughts on scale? Do you just ride their scale equation, their AI infrastructure? Do you have something on your side? What's the relationship on the scale side?

Markie Wagner

>> On the scale side, I mean, the only way to create impact at scale is to go and find your own ways to scale. And so we have deep partnerships with a lot of the labs and folks that have a lot of distribution, but ultimately we have to sort of scale to meet the moment as well.

John Furrier

>> Well, maybe to be the brain of the enterprise, Markie, thanks for coming on theCUBE's NYC Wired program. Great. Congratulations and we'll see you around. Thanks for coming on.

Markie Wagner

>> Thank you. All right. See you. Bye.

John Furrier

>> I'm John Furrier with theCUBE. Mixture of experts here at the NYC Wired program and CUBE original, of course. We got our Palo Altos too connecting Silicon Valley and Wall Street. Thanks for watching.