TRANSCRIPCIÓN 🔗
speaker-0: Hey everyone, welcome back to Beyond Print, the modern IT changelog. I am Shannon Voice, your host here of our Vasion podcast. And today we have a very in-depth episode with a topic that I think is on everyone's minds from sysadmins all the way up to CIOs. I'd like to introduce my guest, Josh Campos. Now you've been with Vasion about a year. Why don't you give us ⁓ a quick background of your experience here at Vasion and before?
speaker-1: Yeah. So been here at Vation for about a year now. ⁓ it's been really exciting. We've ⁓ delivered quite a few things from an AI perspective and automation. really, really pumped. I mean, I I think some of our customers began to see some of this work, which I'm sure we'll get into a little bit later. ⁓ and more to come. But ⁓ for context, I mean I've I've been in machine learning for over fifteen years. so I've done AI before AI was cool, I guess, before people were calling it A AI. ⁓ so I've been doing it for any one from JP Morgan. I worked for ⁓ Schwab, T D DM Aritrade, and also, you know, been a data scientist, I've been an engineer, ⁓ been a leader, VP, you name it. So now we're here, now we're building some cool stuff. So
speaker-0: Excellent. And and you are one of the AI leaders here at Vasion. We're very excited to have you and and very excited to get into the topic today, which is AI only matters when it solves a problem someone finds valuable. I think it's safe to say that there are a lot of claims out there about integration of AI, using AI, AI powered. And I I think we can we can start with let's have a little bit of a reality check. Right. What is AI genuinely good at today? What versus what's still aspirational?
speaker-1: You're gonna hit me with the big questions right off right off the bat. I mean, AI has been useful already for many different things. So, like I said, I mean, I've I've been at this for a long time and you know, machine learning has been useful in a whole slew of different things. ⁓ you know, from finance, ⁓ you know, being able to make predictions about what people are gonna do. ⁓ all the way to Alexa, everyone remembers Alexa, right? So, like that was not too far from the technologies that we have today that we actually felt somewhat useful. At least you could set a time clock, right? Like a ⁓ you know, if you're trying to bake something. ⁓ and it's evolved quite a bit from that. ⁓ but yeah, to your point, I do think that there's a lot of hype. And I think when the dust settles, if we're able to, I don't know, go five years from now. I think that we'll find that AI has done an incredible, ⁓ incredible job from an engineering perspective. Like, first of all, like being able to help us do like build better code or build better software. ⁓ but the other place where I think it's exciting, ⁓ it's gonna help us a lot in businesses. So that means optimizing all your business workflows, making things easier, run faster, better, more efficient. So that is to me, like when you move all the hype away, you're gonna be left with those two main use cases where you're really gonna get an a significant ROI for the cost of AI. ⁓ I'm sure we're gonna get into this at some point later, right? But like cost is a significant part of where AI is today. Not a lot of people know that AI is actually subsidized. Like for lack of a better word, I mean there is a massive subsidy in AI. So
speaker-0: You know, I I I do look forward to to diving into that a little bit more, but I I wanna bring up and just maybe get your opinion almost on behalf of of somebody who might be listening in, because I think that you've touched on something which is There are a lot of people that are using AI, trying to use AI, whether it's a an individual contributor at a company or all the way up to an executive team who has decided that they want to roll out AI to the entire company. We hear Evasion got to experience that, right? All of us have access to to ⁓ an AI platform that is is meant to help us ⁓ enhance our jobs, which has been fantastic. And it is an incredible tool. But I want to maybe set the stage before we get too deep into some of the specifics here. And and that is, do you feel like perhaps there's some hesitation ⁓ because people haven't been able to get it to do what they want it to do, or maybe they haven't been able to do what somebody else told you it could do? Right. I I just trying to understand this difference between ⁓ hype and results, and and do you think that there's a factor almost leading to possibly some hesitancy for people to dive into AI at this point? Or is it full steam ahead for most people?
speaker-1: I think you get all different versions and everybody is in I I I believe that AI is is a journey, right? And most people are at different stages of that journey. ⁓ there are some people that actually, you know, ⁓ got into AI, like using AI in their businesses and they did not see what they expected. ⁓ naturally maybe some early adoption kind of stuff where like there's a little bit of disappointment. But models have been getting better over time. Things are getting significantly better. Now you have tooling that's involved with some of this technology that allows you to do even more. So things are certainly evolving. ⁓ so it just depends where you landed, right? But I do believe that this is a transformational technology. ⁓ but how you implement it is everything. So knowing what to do and when to use it is just as important as what to use. So it's r it's really critical that as as a business owner, as an you know, as an enterprise to understand and have a strategy about how you roll this out, ⁓ you know, where the actual, you know, ROI cases are, like where you actually get, you know, where it's like at the very least break even. Cause there's a lot of people implementing AI today where ⁓ that's not actually it's ⁓ it's not profitable, right? So I I think a lot of you know, we see a lot of productivity tools and use cases
Load full transcript
that are not as, you know, token efficient. That all that means is that you would otherwise spend more money using AI than you would a a human being. So
speaker-0: So let's get into what you mentioned earlier. ⁓ speaking of of organizations making decisions about AI, cost is a huge factor. You mentioned that a lot of AI is subs subsidized right now. ⁓ talk to me about that. What is what does that mean to me if I'm listening in on this? What do you mean AI is subsidized? How does that affect me and my business?
speaker-1: Yeah. there's a very, very deep rabbit hole to go down, but we'll keep it very s very high level. ⁓ but at the end of the day, ⁓ in order to for AI to to run, right? Like the the way that AI works today and the way that we understand it is that we ask a question of AI. That question is gonna be routed to one of these very large data centers that we've heard a lot about. ⁓ and in this data center, there's gonna be a model that is sitting at the in the data center that Then looks at your question and answers. There's obviously a little bit more, you know, depth to this, including like what other information makes it with the question that you're asked, but ultimately that's what's happening. The response is then sent back to you. Anytime that this happens, it costs money. Well, why does it cost money? Well, because the data center takes electricity. So that is the number one cost driver outside of like the hardware itself, like all the Chips and all the stuff that we hear from NVIDIA that they're doing an amazing job building. So there's a hardware cost and there's an energy cost. And the two put together is basically, you know, boiling down to every single question that you ask, every single word that you ask. There's a cost attached to it. So imagine a world where you want to ask a question about like how do how do I make my business better? Like it might be a financial question, right? Like, how do I improve? you know, sales in this particular, you know, region. If you've provide that information along with all your historical financial data, that is going to be expensive, right? That might cost, you know, $20, $30 for that question. Well, what if you keep asking it questions? What if you keep trying to solve for things? Well, now you've generated a $500 conversation. And the question that we all ask ourselves is like, Was that worth it? Sometimes it is. Sometimes it's not, right? It depends on like what you got out of it. But ultimately what's happening is that these ⁓ hyperscalers that we all know and love, right? Like from AWS, Azure, Google Cloud, et cetera, ⁓ there's a cost associated with the technology that's sitting in their data centers. There's also a cost associated with the hardware that they've had to buy to make this happen. All that cost, if they actually charged you specifically for how much each one of these words and these questions cost, it would be significantly more money than what we're getting charged today. And, you know, as business owners, I'm sure we can like reason over the many, many different times when we actually gave people discounts to, you know, take our services and to engage in business and to build relationships. Like this gets done all the time. The same is true with AI. So you're seeing a lot of the Frontier Labs, which are the companies that build these models, along with the hyperscalers that are housing the models and you know, answering your questions and sending back the information. They're basically charging you less for what it truly should cost. And that's that's kind of the the one of the things that is like the most worrisome about where we are today, is that we're not giving a lot of thought to the use cases in general. I'm just talking in generalities. The industry isn't given a lot of thought to these use cases. And so we might end up with use cases that might not actually be ROI positive because we're subsidized today.
speaker-0: One of the areas where Vasion has has found such a stride in how we are using AI and integrated AI is this marriage between ⁓ deterministic and probabilistic workflows, right? So going back to what you're saying, there's obviously a cost involved with AI. I think everybody knows this, although maybe they don't understand the full cost. But how are we using it to justify the cost? And I do want to to just speak specifically a little bit here about VASIN, because this is where we work and this is what we understand. But talk to me then and talk to this audience about when you're using AI in the right way and justifying the cost, you can bring these things together, your your business processes and in our case, document transformation, digital transformation. ⁓ intelligent print automation and we can use AI to enhance that journey and the cost is justified.
speaker-1: Yes. That's right. It's it's really funny how that go you go from ROI negative to ROI positive when you pick the right use case. So to the to the point of like, well, what does it look like the world today when we look at business processes? There's so many of them and ever every organization has its own set of business processes. And when we say, What are the al alternatives here? I can either, you know, continue to do what I do. That means there's humans involved, there's automation involved, there's emails involved, there's documents involved, or, you know, what some of these other companies are suggesting, ⁓ just here's AI, just give it all to AI. Well, so now we're dealing with extremes, right? So ne neither is right. The goal is to find a balance. And I think that's where AI really shines. Specifically, I think what, you know, one of the things that's super exciting about what Vation can do now with intelligent print automation is that you Can live in a world where a deterministic workflow can also be paired up with a probabilistic, you know, solution, which would be like an agent. So you would have this world where
part of your workflow that exists today might be a human actually sending an an email to ask someone a question, or it might be like, ⁓ send it to this person so they can analyze the spreadsheet. Well, now you can build the exact same workflow invasion, but also, you know, invoke an agent that can solve the same task. So you have this kind of, as you as you mentioned, this this marriage between the two that we believe is going to be very successful. ⁓ not a lot of companies are thinking about it that way. I feel like that's what makes Vasion innovative, that we do meet what our you know, we meet our customers where they are, specifically. And, you know, if you're not a fully digital or you're halfway through the digital, or maybe you're like 99.99% digital, we will meet you where you are. So that includes using documents as like the trigger of what's gonna happen next, or it could just be like you just have a paper and that's where we start. So funny that Vation can actually trigger an agent from a piece of paper. That is like the world that we live in today. So
speaker-0: You know, it it's remarkable because we've had some some other guests on the show. I recently met with one of our customers ⁓ and and he was talking about this very thing, about how they found Vasion 10 years ago. ⁓ so before the the full capabilities of AI had really been integrated into anything ⁓ on a large scale. ⁓ but having been with us for 10 years now, ⁓ and and now being at the point where he he is fully into the intelligent print automation space and and using our ⁓ our tools and he said that it his it has allowed their company to save, I believe it was 10,000 man hours a year and close to a million dollars. And and and it's remarkable to think of how when you are trying to figure out how to use a new tool, there's a lot that you can come up with. Right. But like you're saying, as you really start to hone in and you get to that, to that last mile, let's call it, right before the finish line, if you will. But there can be a lot of very tricky things to navigate and to figure out. But it doesn't need to scare somebody from starting. And I love that you said, just bring us the piece of paper. Start right there. Right. And ⁓ before before we started ⁓ recording, you were telling me an example, and I'd like to call back to that. In the healthcare industry, ⁓ well, I'll just let you take it. Give give us that example again of of how this works, like a real life example.
speaker-1: Yeah, honestly, I I I would have to say it was one of the most fun moments that I had in my career. And I I've I've done some pretty interesting things before, but this particular example, the person that we were demoing this agent to was so excited. and you could just see how much it meant to them because you could see the human impact that it would make, especially on a healthcare organization that is trying to take care of people. So for them to say, we can do so much more, we can help that many more patients when this happened. But specifically the use case, if people are familiar with the healthcare industry, when you render a service, you're usually having to bill for that service. And the billing process, it's not ideal. It's very convoluted. ⁓ there's ⁓ some standards, but usually what ends up happening is when, you know, hospitals build insurance companies, there is a back and forth of sorts. usually there is ⁓ you know, triangulation between what services were s rendered versus how much the insurance company's willing to pay versus all these different things. So it actually it's incredibly complex. There's no amount of code that you could ever write to automate this. So For most hospitals, they basically just restort to, well, I just gotta throw money at the problem. So they they just have to go out there and hire a bunch of people and they they end up with these large teams. Now, if you're a smaller hospital, like that's where you really struggle. And so what ⁓ smaller hospitals or even meets medium sized hospitals end up having to just say, Okay, we're gonna do our best. We're gonna go out there and bill as much as we can. If there are any errors that we can't clarify, that usually means that the hospital writes it off. So this is a straight loss to the hospital. Basically means money just goes away. In fact, we actually saw some of these business processes where where it would get to that point, it would it was called poof. Like that the money just goes poof. So that's it. Wow. So it's a it's an incredible problem. That is just not the one particular customer that we were helping. But it's a prevalent issue across the healthcare industry. ⁓ so what we did, again, back to what we said about Vation, like we meet you where you are, we're also solving that last mile problem. So, okay, we built the entire platform for agents. So what? Well, so that we can actually go build these last mile problems. ⁓ and we built a ⁓ healthcare agent that is able to take in all these claims. And then solve them before there's ever an issue. And so the AI is really good at understanding, ⁓ you're trying to build for milliliters instead of like units or instead of this. So we were able to build that and demo it to one of our customers. They brought us papers, like their actual claim papers. ⁓ and we're like, okay, well, let's take that. And so we scan them, we send them to the agent, and the agent found all kinds of things. And in the demo itself, like people were like, ⁓ my gosh, like this, you could just see like the the in their faces that this actually does work. And all of a sudden, if you can actually get what it's owed to the hospital, because this is truly owed to the hospital, they render the service, right? Then that means they can take care of more people. They can hire more specialists, they can, you know, so it it's a it's a very, very, very impactful situation. And what's exciting is that this is just the beginning for Vation,
right? Like so we can't see w I mean, w we're too excited about what we're gonna see next on this. So but
speaker-0: It it really is incredible to hear these stories and and being on on the social marketing side of the business, I've been able to to interact with customers not only through interviews, but through the reviews that they leave on on the aggregates like G2 and Trust Radius, through our in-person events, ⁓ whether it's Fasion on the Road or at one of the conferences. And we get to hear this over and over and over. And and I was at a conference earlier this year where we were speaking with a hospital administrator. And this one of one of the many aspects of of the Vasion platform, right? ⁓ of this, of this, ⁓ this journey to to ⁓ modernize, consolidate, and automate. And they were explaining one aspect of that. And he literally turned and he looked at, you know, my colleague that was there in the booth with me, and he said, You can do that? It and it was something that had it had never occurred to h to him. And this is not a criticism to him, but there are so many who don't fully understand, like we're talking about, that last mile problem might not even be on your radar because you're still trying to figure out how to navigate the first mile. And I do think that that is where we come in as an organization in this space, and we are able to help you get started and then continue that journey of digital transformation with AI solutions that actually solve problems.
speaker-1: That's right. Yeah. I mean, you even circling back all the way to the beginning, which was one of the key issues today, is like if you wanted to adopt AI, you're just given tools. And then what? The what happens? Like, do I have to learn all of this? Do I have to, you know, become a an expert? And and that's that's the challenge. That that's why we didn't approach it like that. We approached it by, okay, yeah, so we're gonna build a platform so that This is scalable so that we can provide this to any customer over any industry. But we also said business problems is the reason why we're here. Solving that is what actually makes the world go round. So we think that this technology can absolutely do that if used properly. So it is
speaker-0: It is a a wonderful thing to have such an expert on the show. I really appreciate you taking time to meet with us. I have one final question, again, almost on behalf of somebody who may be listening in. What would be the advice that you would give someone, you know, an IT director, CIO, systems administrator, anyone who is maybe at the beginning of this digital transformation journey, regardless of of industry, what would you say is the best way to begin? And how can Vasion help them start?
speaker-1: It's great question. I would say let's start with what not to do. Usually that and I'll and also carry into what what I I believe is the right way. But when we all got started on this using some of these new technologies, the very first thing that everybody did was just like just just use it, just throw it out there. And ⁓ n not even though there was a lot of thought given to the fact that, hey, this is revolutionary, we should do something about it. ⁓ not a lot of thought was given to what specific business problem to solve. And that I think is the most important part, right? Because that allows you to understand, okay, if I understand the business problem, then I can also derive what the potential RI for solving that business problem would be. And so right off the bat, you set yourself up up for success so that you don't get like the AI disappointment that I call. Because that that that really, and I feel like there's big amount like there's a quite a quite a bit of this right now where you get some buyers and then you also get some sellers on this technology where they're like it doesn't work. And some other people are like, this is gonna change the world. Again, the right answer is somewhere in the middle. ⁓ but making sure that you pick the right use case where you can figure out if I solve this, it's going to add this much to my bottom line, that's the best approach to to to do with AI. I think the very next thing would be, okay, so if I were to solve this, what kind of technology? And then now you begin to ask the question is do I build it or do I buy it? And so a lot of considerations there. ⁓ every single hyperscaler, like, you know, ⁓ they're they're getting just they're so excited to sell you something. So you have to take that into consideration because whether or not, like if you buy it or you build it, you might end up having to need. you know, some of some of those individuals to come help you on that journey. but as far as like what Vation can do, I think Vation is excellent at meeting customers where they are. And w again, we're all looking to help everyone on this, you know, dis digital journey. And one of the things that is really awesome about the automation and AI platform that we build is that you don't have to be fully digital, right? Like that's that's the key. Most of most of the solutions that are out there, they almost ⁓ they don't even require, they just assume that you're already digital. So it makes it quite challenging. Right. So for us, you're able to come, you know, with what you have and we're able to figure out solutions on okay, is it a is it full determinism? Is it a combination of, or maybe this is something that we can fully delegate to AI, whichever the case, we have a solution for that.
speaker-0: Thank you so much, sir. It's been great. And ⁓ hopefully we'll get to chat more down the road.
speaker-1: Sounds good?