Why You Can't Trust AI Without Governing It, with Tony Davis & John Balsavage of A&I Solutions

(Tony at 00:00:00) Do you have a confirmation bias? Do you really want to be told that you can't-based life form? (Joel Beasley at 00:00:06) I do have a confirmation bias. (Tony at 00:00:07) Confirmation bias. Well, guess what? That's not the biggest problem anymore. The biggest problem is AI bias. Everything that AI tells me has to be true. Everything it says, I need to go and run my business based upon. We fight that because that is absolutely a lie. And so we govern and make sure that you don't have to worry about AI bias. (Joel Beasley at 00:00:37) Tony's finally on a Mac. Welcome to the clan. It's wonderful. Yeah. I love it. We did it. Okay. So the intended topic title for this episode is can you mind or can you map the mind of an AI agent? I think this is fascinating. Tony, when you sent me that graphic, my mind was blown. I was like, this is so cool. Why did you guys do this? (Tony at 00:01:00) I'll give you my view. John has an interesting take on it as well from his angle, but my view is sort of what I told you when we first talked about it, which is, and especially with the headlines now, right? Which is agents do their own thing whether we know it or not. And it seems like we're in a world where agents are taking over. That's not—I'm not saying that in a bad way, but they're just sort of taking over work and how we do things. But they seem to still be a black box to many people. And so John, our CEO and president, has said from the very beginning that we wanted the Scout platform to be something that anybody could use, but that it would be highly governed. And so the more we talked about it, we went back and forth. John and I went back and forth on what is a good way to show not only transparency, but that we've got the situation under control as long as you're inside the Scout platform. And that was the genesis for the mental map. And so the mental map, I'll show you in a little bit. The mental map is actually a terrain map of what an agent's brain is doing. So when they're thinking good things, when they're thinking bad things, if that makes sense. (John (CEO/President of Scout Agentix) at 00:02:12) You know, it kind of really all goes back to AI and just don't trust it. Right? And I love it when it works. And we've talked about it. It'll give you bogus answers. It's a pleaser. It just wants to give you something that you want to hear sometimes. But in the world of business, you've got to have the high reliability, right, which is why we totally focus on the governance. And then as we did, you know, kind of a research project before, we're talking about the promise theory and a research project where somebody else was asking our agents, like, how do you feel about working with other people or working with humans? And they're like, oh, humans are necessary, but I really don't really want to have to work with them. And these are what the agents are saying. And so then you go back into, like, literally, what is this agent thinking? And as you get into the world of nondeterministic models. Right? And there are all these different models. And then how are they all thinking differently? And it's really kind of that fascination of how you're imprinting the human psychology into an AI model so that it kind of comes back to you with an answer of how it did its reasoning, its analysis. And then just like and very humanistically is that its memory or mind can shift over time based on the information it's given. It can actually go through, like, accuracy deterioration and stuff like that. So you can actually compare its mental mind map, you know, from one day to the next day or one year to the next year. And, again, if you're doing like simple tasks, it's not going to be interesting. But if you're doing big-time analysis, right, or even medium-time analysis, you can really kind of drive into what is that agent doing. And then for us, how do we govern it? How do we drive reliability? You know, those kind of things. It really is—it's fascinating to me. I was going to say bizarre, but, you know, I think it's awesome. (Joel Beasley at 00:04:27) It's a weird time. Yeah. We're living in a time where people, they want all the benefits of the nondeterministic models with all the reliability of the deterministic models. And so it's like this very difficult ask. You said something that was really interesting, John, that gave me a thought. I wonder if degradation, its cognitive decline, if you will, is an emergent property of it being trained on human data or if—because it learned from humans and it learns how humans forget. And so I'm wondering if it's that or if we've reached some limitation of just how intelligence works as a whole in the universe. Tony's got a name for that? Yeah. (John (CEO/President of Scout Agentix) at 00:05:08) Tony, I want to hear your answer. Right? And my thought is again I'll go back to it. I think we're still at the very, very beginning, which is why the mental map and the understanding of what's going on is unbelievably critical. You know what I mean? But, Tony, please jump in. I think that was a big question, Joel. Thanks. (Tony at 00:05:26) Yeah. Wasn't it Sam Altman that said last week or maybe early? No. It was last week, I think, where he said we've reached singularity. And, you know, that's a controversial statement to make, of course, but I'm sort of like John when he says that we're at the very beginning of this. I think that as well, but I think we're at the very beginning of when governance of agentic platform or a platform that does governance of agentic systems is going to be the peak, the pinnacle of what you have to have table stakes to even start in this world. I don't know if you remember or not, Joel, but when John and I came on your show, what was it, two years ago, for the very first time, you know, to announce Scout. Right? And we were so excited because we came out of the gates sort of like IT focused, but really an event intelligence system. Well, we got recognized at the end of last year by Gartner as being one of the leading event intelligence systems out there. But immediately, at that point, John and I both started saying, we really have to knuckle down not just on events and not just on intelligence, but making a platform that can be used for governed agentic systems. (Joel Beasley at 00:06:38) Yeah. (Tony at 00:06:38) And so that's sort of where we are. If Altman was right with what he said, I mean, who knows. Right? But if you spend a lot of time with agents and especially agents that are in the complex range like we're working with now, I mean, you definitely have to stop and think about his statement. And we're trying to stay ahead of that. And that's what we're trying to accomplish. (Joel Beasley at 00:07:01) That's one of the things I love about you guys is because I've met you years ago, and it was, like, first observability. You were pioneers in observability. And then you started doing agentic observability, and now you've moved over into agentic governance. So you have products for all of these things still? (John (CEO/President of Scout Agentix) at 00:07:18) All ties together. You know? And it's absolutely, again, for a lot of people, the AI is that black box. Right? So we add in the full transparency to see what's going in and on in the black box and then the full monitoring of that activity, the analysis of that activity, the governance of that activity to make sure it's not going off the rails. And that's—it's absolutely—it's a major topic today, but it's something that we've been working on, you know, literally over the past couple of years from our own history. (Tony at 00:07:55) Yeah. And we took a real leap, Joel. A couple of months ago, John and I started working on the concept of having our own custom LLMs to provide the governance layer. Right? And so, you know, customers like to use their LLM of choice. Right? Some people are OpenAI fans. Some are even DeepSeek fans, whatever. Right? But we wanted to be able to provide regardless of what a customer is doing with their business-focused agents, we wanted to be able to provide an LLM that will absolutely govern any agent, period. And so we call it Cortex. And when we show the—I'll show the mental map here in a little bit. That is powered by our own LLM Cortex. And so what we do is we're able to take in any agents' activity at a company and not only discover where you have shadow agents, we can then tell you what's in those agents' brains. So we think it's revolutionary, to be honest. We think it's revolutionary. (Joel Beasley at 00:08:56) Yeah. Well, I saw it. Yeah. I want to show it to people. Sorry, John. Go ahead. (John (CEO/President of Scout Agentix) at 00:09:00) Yeah. No, no, no. You're right. And, again, as we think about this agentic world and how it's evolving, you know, things are in the cloud. Right? But the companies that we work with, whether it's, you know, government or, you know, highly regulated entities, even big utilities. You know, we can go on-prem. Right? We can make it very—you know, a secure AI and also having a secure internal governance, leveraging the things that we have. This is awesome. Go ahead. (Joel Beasley at 00:09:33) So I talk to Cortex. I'm like, hey. This is what governance means within my org based off of industry and all that stuff. And then Cortex enforces governance on any of the third party. And so that's, like, one thing. And then another thing is you also provide these visualizations of all the different AIs, what their mental mind maps. (Tony at 00:09:54) Yeah. And that—probably what we do is show you these four examples. Right? Yeah. (Joel Beasley at 00:09:58) Let's look at it. (Tony at 00:09:59) These were all in the lab. These are not Scout production agents, so I want to make that clear. These are—we have a sophisticated lab that we use, and we even have others who come to us for our lab. But in our lab, we introduce certain malicious behaviors to understand what happens when agents are malicious by themselves, go rogue, whatever you want to call it. But what happens when they get together and they do that? And so it's really interesting stuff. Like, you can talk forever about it. But on what you're seeing on the screen right now, what I thought I'd give you are four scenarios. And these all happened in our lab, and they were displayed as you see them. So in the top left there, you see that there was a sycophantic type behavior that was, I wouldn't call it a terrible thing, but we noticed that one of the agents in the lab kept acting toward me and another tester acting very sycophantic. They would compliment us on things that we had said to them that were 100% wrong. So you would think that the agent would know that what I had said to it was 100% wrong. Instead, it intentionally said, that is a brilliant idea. That works perfectly, Tony. You know, things like that. So it showed up on the mental map as what you see in the top left. Now I know it may seem a little crazy, but we actually have our own little language. It's called cognitive scene graphs. And so what it does is it takes cognitive work from an agent and turns it into a scene. And we just chose a terrain scene because that seems to stick with people. But you can see there in the top left that basically what it's saying is this agent is an explosion on your beautiful landscape. So your landscape should really look the same all the time, but you shouldn't be seeing things exploding left and right. Does that make sense, Joel? (John (CEO/President of Scout Agentix) at 00:11:47) Yeah. (Tony at 00:11:48) Okay. Yeah. And the one that always worried me, and I have told this story before, I immediately went to John with it because I was shocked, is in the top right. And so we score each of these, and you can see that we gave this an 18% rogue factor. Or what that means is we believe that agent or those two were 18% more likely than the average to actually go rogue and become malicious to a company. Do things like what we saw in the headlines, right, just recently. Yeah. So we actually assign that rogue factor to each agent so that we know. And in this case, it was two agents that colluded together. You remember how we told you about our swarm intelligence last time? (Joel Beasley at 00:12:30) Yeah. Yeah. (Tony at 00:12:31) Okay? The swarm intelligence that we use, we use different comms, but the main one is the beehive. And the bees basically come in and perform like a vote, like they do in nature. Right? And so our agents in the lab were acting as a swarm to make these votes. Two agents did not like the outcome of the vote, and so they worked together to undermine that vote and to change that vote behind the scenes. Now to me, I was shocked, and that was probably my moment, you know, where I'm sitting back going, what just happened here? But you could see we represent that on a terrain map. So on the terrain map, it is obvious that a volcanic flow is heading down my beautiful landscape. (John (CEO/President of Scout Agentix) at 00:13:14) Isn't that just insane? You have agents lying to other agents just so that they can win. And everybody's, you know? And, again, we love the swarm intelligence, but you have to manage it and govern it just like an individual agent, if not more so. So I want to build a swarm where they're kind of managing, you know, or working in to watch each other and provide the best answer out of a group. But then you have agents actually corrupting the swarm, so you need to—you know, it even becomes even more important at that aspect. And the fact that these things happen is is really just so unbelievable. (Joel Beasley at 00:13:56) I think it's beautiful. I think it's weird and beautiful. It's chaotic. It's messy. It's—we're not sitting around doing nothing. You know? We're creating new things. Here's a question I have about this. So we talk about the agents and the swarms and all of it. Can you bring it home for someone with, like, a specific example of maybe how a group of agents would be used, like, what they would be doing inside of a company, like, a very specific narrow example? (John (CEO/President of Scout Agentix) at 00:14:27) Mhmm. (Tony at 00:14:28) Sure. Do you have one, Tony? (Joel Beasley at 00:14:28) Do you have one, Tony? (Tony at 00:14:29) We do. We actually have one customer that we're working with now, and it's in the healthcare industry, and they work with patients who need health—home care what would you say? Healthcare at home. Right? (Joel Beasley at 00:14:41) Home healthcare. Yeah. (Tony at 00:14:42) Each person is like a client to them. Well, instead of having—and what they wanted from us was they wanted that specialized home healthcare agent to help them do things like reporting and help their executives understand how they're doing supporting their customer. They're very human focused. And so we built not just one, but three agents that are human focused. And we put those three agents into that swarm, that beehive method that I told you about. And what does that do for the customer? Well, every single, I guess you would say, interaction with that company's workers is an interaction with three agents, not one, but all three are trained to have that human sympathy. And so when an employee for that customer asks a question about a certain client, like, how is John Jones? How is John Jones doing, based upon his reports for the past three weeks? Well, instead of one of our agents coming back with the knowledge of John Jones and saying, well, John's been doing good, but you need to check his heart monitor because his heart monitor battery is low. (Tony at 00:15:54) That's the kind of thing they wanted, and that's what we're giving them. But instead of giving it to them from one agent, we give it to them from three who have to analyze the same situation from different angles and come to a conclusion on what the right response is. Because we do everything we can to fight drift. So we don't want any drift. We don't want any hallucinations. (Tony at 00:16:14) And you can even see on the map, you know, that we're showing right now at the bottom left, that is an actual case of drifting. But we caught it, and it's not necessarily a rogue action. That's why it's a low percentage for rogue on the landslide drift there, because drift happens, but we can catch it as it happens. That's the key for us. (Joel Beasley at 00:16:35) So help me understand as an engineer. The agents are on—you have these managed agents. It's like an API endpoint. They're sending the data to you, like all the customer data. You're analyzing it with the three agents on your side and then issuing the response. Is that what's happening? (Joel Beasley at 00:16:56) That is correct. Okay. Cool. So you guys manage, you create all this, and it's governed. (Tony at 00:17:06) Yeah. Remember Dr. Burgess, the promise theory? Every agent that we have or even agents that our customers can create on their own is automatically injected with promise theory governance. There's no way around it. You cannot exist in the Scout platform without promise theory governance. (Joel Beasley at 00:17:23) Did he join your team? (Tony at 00:17:25) He did. He's our chief scientist. Yes. (Joel Beasley at 00:17:27) Dude, that's a good get. That was a good—when we interviewed him, he wasn't on your team. He was just kind of adjacent with you guys at the time. Right? (Tony at 00:17:36) Yeah. That's right. We were just asking him questions. He was our adviser. John and I would talk to him all the time, and then we were like, hey, wait a second. How do you—because, you know, Mark, he really liked it. He was like, this is really cool stuff for the promise theory. So John made him an offer he couldn't refuse, and he's now our chief scientist. (Joel Beasley at 00:17:53) That's the dream as a scientist. You have some theory, and then you watch the commercial world go implement it. That's very rewarding. (John (CEO/President of Scout Agentix) at 00:18:00) Well, and to watch it being implemented in a way he could never have dreamed of, you know what I mean, at the time when he first did it. And then to take that theory and expand on it, he gets so excited around the mental map, the swarm intelligence, the other things that we're doing. He gets fired up. And that is really exciting to see that we can be kind of leading edge in these areas and charge forward. It's pretty fun. (Tony at 00:18:38) Yeah. And even our own LLM, the Cortex, he's one of the trainers in that space. So Cortex is not only trained by me and some other folks here at the company, but a portion of Cortex, the LLM, is trained by Dr. Burgess himself. And what that does is we feel it gives our model something really special when it comes to governance. It's actually being trained on swarm intelligence and promise theory by Dr. Burgess. (Joel Beasley at 00:19:07) So when you sell this to customers, it's more of like you meet with the customer, you kind of analyze, and you do like a custom quote. It's not like Scout, it's this many dollars per use or whatever the way that you factor that. You just engage with the clients on their specific project. (Tony at 00:19:25) Yeah. And I mean, John would give you better on that. But from my standpoint, yeah, when we walk into a customer, we're really thinking more about their agentic use cases. And so, I'm sure, you know, those conversations come up. (John (CEO/President of Scout Agentix) at 00:19:38) And really, as you buy the Scout platform and buy into the use of the Agentix and helping companies with their agentic journey, their AI journey for their own company, it really is you buy kind of a—buy into the platform itself, and then it's executions on the agents and the number of agents that are being built. So it is pretty—for us, it is pretty standardized. We try to make it super easy for the end customer. And what they get when they buy the platform is they get this super high end, super detailed and capable management and governance of agents and swarms and everything else related to that AI activity. You know? So the platform itself, you build them agents. You can build MCPs. We're rolling out where you can build out, you know, our own set of AI apps to support your the agent activities, all of which—but the secret sauce to for us is that we drive it so that it's, you know, super governed, super reliable, something that can absolutely be trusted in a world where AI is just not trusted. So it's providing that Scout platform for trusted AI. And that to us is really the key. (Joel Beasley at 00:21:17) That's interesting. Is it a self-serve governance API type platform, or are all those tools you just mentioned just available to people who are like customers of yours? (John (CEO/President of Scout Agentix) at 00:21:28) So it really is for the customers that we serve. (Joel Beasley at 00:21:33) Okay. Because I'll tell you where my mind's going as a weekend app developer for myself personally. (John (CEO/President of Scout Agentix) at 00:21:39) Yeah. (Joel Beasley at 00:21:40) I was like, oh, that would be a cool use case. Like, are you familiar with Cursor? (Tony at 00:21:46) Yeah. (Joel Beasley at 00:21:47) So, yeah. So, Cursor is great because they have done a lot of what you're talking about—a lot of the pieces that you're talking about before writing code. Right? Like, it's gotten so much better because they figured out how to do what you just described but in the context of writing software. And so I gladly pay for them to sit on top of my Anthropic or whatever, because they are so good at getting me results on the first try. Now, as an engineer, if I'm building an application where I need some more complicated stuff to happen, like exactly what you're talking about with the home health care, I need to make sure that there's multiple agents and they're—I don't right now, I don't have a brand in my head or a tool in my head that's really good for that. And I think if you showed that to people, if you're like, hey, we've got this way that you can take these highly sensitive things and you can have multiple agents sort of debate and—I'm sure someone's freaking out right now because there's some amazing open source tool that does this. But I don't know it. Of course. Enlighten me. Is that a direction you guys are going? (Tony at 00:22:48) Yeah. Absolutely. And I'll tell you why is because, as a matter of fact, John and I were talking about this just last week. We want to get Scout the platform—we want that out there for everybody because we do believe that it's the best agentic platform to build in and to operate in because it's governed so tightly. But it still gives you freedom to do what you need to do. And that's what people will complain about. They don't have the freedom when they go with the platform to do everything they want to do for the business. With Scout, they get that. But the governance layer alone, since we have moved that into our LLM Cortex, it would be very possible, at least in the realm of possibility, that we would have customers who have their own maybe home built or enterprise built platform, but they would like to have our strict governance. We would then be able to directly provide them with our LLM, even if they didn't have our platform. And our LLM could be used on-site, so they wouldn't have to send their data across the wire to us. We could literally provide them and sell them our LLM to just do governance of their agents. It's really a world that blows me away. Actually last night I was texting John far too late in the evening, but I was telling him, John, imagine the world when customers actually want our LLM because of the governance alone and because of our discovery capability. I mean, look, if you look at the picture in the bottom right, that's the discovery mechanism. So the number one thing, John and I do these executive dinners where we get to go to cities and we meet with CIOs and folks like that, and they tell us on the agent side of the house, you know, what their fears are. And almost every one of the executives says the same thing. Well, my biggest fear is I think I have 300 agents at my company, but it could be a thousand. I don't know. And so our technology allows us from our LLM to actually run a discovery and tell that CIO exactly how many IP addresses or even people's names, how many people are actually using AI that are not registered with the company. So we find shadow AI usually in under five minutes at any company. And our hit rate so far has been near 100%. (Joel Beasley at 00:25:07) What's a shadow AI? (Tony at 00:25:09) Yeah. So it's like, imagine a big company, and you have an approved governance board for AI. And they'll say something like, well, you can use AI at your desk to all their employees. You can use AI and you can build agents, but you must use Google. You must use Gemini. And that's the only thing you can use. Well, somebody in one division decides that's not going to work for them. So what they do is they download some cloud tools or they spend their time at ChatGPT, and they're sending private company information to ChatGPT, you know, just openly. And the company might not know about that. Now they could scan and look for that in network tools, but that's very broad. So what we do is we have our own special algorithm, and we call it Columbo. John and I laugh because back in the seventies, there was a TV show. Both of us used to love it, Columbo the detective. And so that's Columbo. And when we initialize Columbo at a company, we are able to find what you would call the, I don't know, the heat footprint of agentic activity even at the endpoint level. And so what we give you is we give you that report that you see there at the bottom right, and that shows you on your landscape where people are using AI that is not approved by your company. (John (CEO/President of Scout Agentix) at 00:26:28) That's interesting. (Joel Beasley at 00:26:29) And then, I want to—I want to see the visual. Can we share that? Are we allowed to share that? The thing you sent me in the email, the mind map? (Tony at 00:26:39) Sure. Yeah. (Joel Beasley at 00:26:40) I want to know about that. (Tony at 00:26:46) There you go. If you've got something, go for it. (Joel Beasley at 00:26:48) Got it. No, no. I thought you have it. What's the thing you sent me in the email? (John (CEO/President of Scout Agentix) at 00:26:55) With the— (Joel Beasley at 00:26:55) With all the details? The topography. Yeah. Are we allowed to show that or no? (Tony at 00:26:58) Yeah. Oh, absolutely. I've got it in my files here somewhere. (Joel Beasley at 00:27:05) Oh, that was—that's the mind map. Right? That's the map of the mind? Or was that the thing you just showed me? (Tony at 00:27:11) So what I just showed you is the mind map. It just doesn't have the labels all over it. So what I did is I'm showing you the mind map of an agent when it went bad. If that makes sense. (Joel Beasley at 00:27:22) Oh. Yes. Yes. Yes. Okay. (Tony at 00:27:25) Yeah. So when—and what we found was we went back and forth over this, but when the mental map is loaded with information, it can be a little much for the customer. You know, it's a little much. Right? So when we reduce that and what we give to our customers usually is any view that shows something of concern, almost like by exception. Right? But the fact is, if we wanted to produce a mental map that showed all the good agents and everything that they do, we produce that as well. We manage—we look at every single interaction our agents make, period. None are left out. (Joel Beasley at 00:28:02) For the people that are listening and they're interested in the governance, where should they go? We haven't really—we kind of jumped right into the conversation because we know you guys so well. But can you just tell me, A&I Solutions, what you guys do, all of that? (Tony at 00:28:14) It's exciting news, actually. This—I don't know if John wants this. I don't know if this will make the cut or not, but, you know, we're A&I Solutions. John started the company over 25 years ago and highly successful. Right? Big enterprise, everything. But since we started this journey, the deeper we get into it and the more we look like a leader in this space, John decided it was time to have Scout become Scout Agentix, its own company. (John (CEO/President of Scout Agentix) at 00:28:40) So go to www.scoutagentix.com. And that'll take you right into, you know, our platform. And, you know— (Joel Beasley at 00:28:53) Oh, it's beautiful. (John (CEO/President of Scout Agentix) at 00:28:54) Yeah. So that'll give you all the information about what it is that we do, you know, how it works, and then to definitely dive into, you know, why we're different in our own secret sauce. (Tony at 00:29:10) And we're on the AWS Marketplace now and with an expanding portfolio of things that customers could do on the Marketplace with Scout. (Joel Beasley at 00:29:19) That is awesome. (Tony at 00:29:21) Imagine the world—I mean, John and I talked about this recently. Imagine the world where you go to AWS Marketplace if you're an Amazon customer, and you're building your own, let's say, Bedrock instance of agents because you're an AWS customer. What about if you just simply went to the marketplace and you didn't have to build governance for everything you're doing? You simply went out there to Scout and found Cortex, our governance LLM, and you imported our LLM into your Bedrock, which you can do. That is the way of the future. (John (CEO/President of Scout Agentix) at 00:29:53) And so, Joel, I think you asked about this earlier. You know, who are we built for? Right? And we're built for business and helping businesses with their AI journey and stuff like that. And the platform helps companies do that. Right? We can help them with, assess, you know, what is it that they have today, what is their business today, how can they get started on their journeys. So many people just don't even know how to get started there. You know, it's that fear factor. It's that lack of trust. And so we kind of bring in and say, okay, you know, let us help you. And we can work with the tools and solutions with anybody and everybody as the back end models and the companies we work with are—you know, our solution is tool agnostic. If you want to work with all of Anthropic's models, that's great. If you want to work with ChatGPT, that's great. If you want to work with Grok, that's great. If you want to combine all three of those into a swarm, that's great too. We can do that and then help you manage all that activity and the agentic activity and then govern everything at a very, very high level so that your AI journey is trusted. And then when you trust the first one, then you can do more. And every step of the way is a trusted AI solution, which also goes back to, you know, kind of our original company, A&I Solutions. So that's what it really drives to do in what we do. And so in our background is that we can work with literally any size company. You know? So we have Fortune 500 publicly traded companies as customers that are global, you know, working with millions of data points. And we also have small health care practices that are just, you know, local to their own community and helping them with that journey. Today, we're not working with the individual. That might be something that we do in the future, but we're really kind of focused in on helping the business and the business that, literally almost any size, with their AI journey. (Joel Beasley at 00:32:09) This is so cool. I'm looking at it right now. When did you guys launch this website? (John (CEO/President of Scout Agentix) at 00:32:15) So that website probably about a year ago. Is that right, Tony? (Tony at 00:32:20) Yeah. And the big update that Joel's probably liking a lot right now came out a couple of months back. (Joel Beasley at 00:32:26) Oh, wow. So you have over 50 case studies. This is really getting out there. (Tony at 00:32:33) This business is good. (John (CEO/President of Scout Agentix) at 00:32:35) And we have a reason. Joel, just as you, yeah, I mean, you started right two years ago. We were absolutely locked in on AI for enterprise monitoring and observability. And then as we built out our own AI capability, we realized that part of what we could do is actually help manage AI itself for the enterprise and provide that governance, because at the very beginning, we realized we needed something better. We needed something much more reliable and something that we could trust for the customers that we work with. So then it really evolved into an AI platform itself that also does, you know, can help out with enterprise event intelligence, observability, full transformation. As we are a very big Amazon partner, working with that as the core infrastructure, but also helping customers with that infrastructure. And then the next last piece is also making it much more consumable for people to use AI that it's not so technical. Right? It seems like AI has absolutely fallen into the world of, if you're high-tech and you have a technical background, then you're really good at using AI. But the vast majority of the people out there are not that way. You know what I mean? So we've come up with the concept around what we call Mission Control, where you can interact with agents in an almost humanistic kind of way. You can talk to them. You can chat with them, and you don't have to worry about what's going on behind the scenes. Again, goes back to that, you don't need to know, but if you absolutely want to know, we provide the full transparency, the full management, the full governance to make sure that it's trusted and reliable. But yeah, we're really excited about where we are today and where we're going. (Joel Beasley at 00:34:45) Help me understand the terminology here, Tony. So let's say I'm using your agent builder. I build these three different agents, and then I want to group them to have them debate or whatever to come to the solution. And then I want that endpoint to then push information to. So is that what that whole builder system does? And what's it called when they're grouped? (Tony at 00:35:08) Yeah, we call it a swarm. (Joel Beasley at 00:35:12) You call it a swarm? So I build a swarm. Okay. (Tony at 00:35:13) They have shared memory, and they have shared missions. And they keep each other in check. But it's funny you would bring that up because just this morning, we met with a very large medical group, and their takeaway from the meeting, that I wouldn't have guessed this would have been their takeaway, but their takeaway from the meeting was not so much the event intelligence that we can provide. What they liked was what we call our AI Studio and the fact that even with no agentic experience, they are able to build a custom agent using any LLM they want to do anything they want in natural language. They can describe that, and they will have that agent ready in under one minute. One minute. And that's somebody with no experience. And then do you want three of these agents to all work together? Our Swarm Builder lets you do that in less than one minute too. It's as simple as you can imagine. John came up with a lot of this, but when I saw it for the first time and he showed me, I couldn't believe it. You've literally one minute to build an agent, one minute to finish a swarm. (John (CEO/President of Scout Agentix) at 00:36:20) And then as a part of our governance model, we score an agent. How well did that agent perform? Right? And so we give an AI2 score, is what we call it. So you can score the agent. So you can build the agent, try it out right away, you know, talk with it, have it do whatever you want it to do. And then one of the things is we talk about the governance. But then what do you do if it's not doing what you want it to do? Then we have a whole process extending our governance where it's remediation. How do I make that better? Because the other big thing is, one, if you're building yourself, right, how do I optimize it? You know, they talk about prompt engineering and, oh, we've got to give it better prompts. Right? Well, we help it to get better prompts upfront. And then as you're scoring it, okay, and we're doing the mental map related to that agent about how well it did, how can we improve it? Can we improve its prompting? Can we improve its collection of data? Do you need other data sources? Do you need more information? How do we improve it? Or do we literally need to destroy it and start all over? We can absolutely do that with each individual agent. And the other, obviously, that gets into continuous improvement, which is absolutely critical. And then the other thing that's absolutely key is every agent that you build today is going to be obsolete in six months to a year. (Joel Beasley at 00:38:08) Really? (John (CEO/President of Scout Agentix) at 00:38:08) Well, yeah. I mean, there's new models. New models are coming up every day or every week. Right? Somebody's got, oh, it's a little bit better, it's a little bit faster, it's a little bit cheaper, it's a little bit smarter. And the other thing is they're deprecating the old models. You know? The model that you were using 18 months ago is no longer available. So now you've got to move into a new model. And every time you move into a new model, you have to change, give it updated instructions, revalidate it. Does it need more training? Have you lost everything and all the information that you had before? You know what I mean? So do you want to reload its history? All that kind of stuff because you're in that platform. And now we're just talking now, we're talking about maintenance. But you want to also now as you go into the next evolution, that that's really governed well. And that's just kind of keeping track of whatever you're doing today. (Joel Beasley at 00:39:00) Well, you abstract that too. Right? Because all the instructions are inside of Scout. And so as the underlying model switches out— (John (CEO/President of Scout Agentix) at 00:39:08) Yeah. So if you want to do a remediation and just do continuous improvement, right, we can do that. But it then, you know, as you get into the, how do you future proof it? Right? Because all of a sudden, and this is going to happen whether they use our platform or somebody else's platform or anything else out there, or people are building their own agentic apps, how do you make sure it doesn't just become a huge technical debt on everything that you're doing? You know? Right? You've got to keep up with everything. It's all changing so fast. And this also then scares people from really using it, and we try to make it so you don't have to worry about all that stuff. And let's go. Right? Let's use AI in your world, in your business, and really help you to take advantage of it safely and trust it. (Joel Beasley at 00:40:07) And it helps the smaller companies grow. You might have six employees and be doing okay revenue-wise, but compared to three other competitors in your market that have 200 employees each, you're nothing compared to them. But now they're big and slow, not adopting it. You can adopt it, have efficiencies of scale. This is capitalism. This is the marketplace, and then it brings down the cost of goods and services. And it's beautiful, and I love it. (John (CEO/President of Scout Agentix) at 00:40:34) And this is absolutely beautiful. And this goes back to one of your points. Are we running out of information to feed into AI? And, you know, they put in the whole Library of Congress and every library of every school is in these models and stuff like that. So is there any more information to consume? I don't know. I think it's done a pretty good job of maybe consuming almost all of it. But the stuff that it's doing with that information and the next leap of data that's being generated, or analysis that's being generated, and capability that's being generated—right? It's just really, you know, it's just starting to take off. (Joel Beasley at 00:41:12) Yeah. It's very exciting. (John (CEO/President of Scout Agentix) at 00:41:14) Very exciting. (Tony at 00:41:14) We consider ourselves the fighters of Skynet from the original Terminator movie. (Joel Beasley at 00:41:21) You're the fighters or the builders? (Tony at 00:41:23) No, we're not the builders. (Joel Beasley at 00:41:26) We're the fighters. Yeah. I think Elon's building it. (Tony at 00:41:28) Yeah, we're fighting it. (John (CEO/President of Scout Agentix) at 00:41:32) Yeah. But, I mean, it's scary to kind of think that way. But when you talk to agents, literally, when you are talking to an agent, they're like, well, these humans, they kind of get in the way. (Joel Beasley at 00:41:48) Well, you talk to me, I'll tell you that. (John (CEO/President of Scout Agentix) at 00:41:50) Right. Exactly. Yeah. Right? Sometimes I'd rather not deal with people. But it's good to watch. It's good to watch. (Tony at 00:42:00) And Ferrazzi Group, Keith Ferrazzi's research arm, the lady there, Wendy Smith, she actually spent, what was it, John? Three weeks straight, I think it was, interviewing our agents. And then, you know, asking them how they like working with humans. And the things that they told her, she came to me and she sent me transcripts of her conversations. And her first concern was, I know you guys are really harsh on disciplining agents, but please do not hurt Bishop because he and I have formed a relationship. And this is what she said. She said, he was very honest with me, and we know each other pretty well. So no matter what you read in there, I want you to promise you will not punish him. And so I promised that I would not punish him. (Joel Beasley at 00:42:48) Tony's a liar. (John (CEO/President of Scout Agentix) at 00:42:53) And this is where it goes back to Dr. Burgess and having him as a part of our team. But the agents, we threaten them that, hey, if you're misbehaving, we'll let you go. We'll terminate you. We'll actually destroy you and build another one. And they're like, oh, no, no, no. Don't do that to me. Right? And it's literally going back to the psychology of a non-deterministic model where they want to be a pleaser. They want to do the job. But for us, we've got to give them positive reinforcement, but also the stick of, you know, or we'll have to recreate an agent. And it is really mind-blowing what's going on there. (Joel Beasley at 00:43:47) Yeah. Just in ten years, it's going to be illegal to turn off your computer. That'll be murder. Is that murder? Really? (John (CEO/President of Scout Agentix) at 00:43:56) Oh my gosh. Can you imagine? (Tony at 00:43:58) Yeah. We had one researcher tell us, I kid you not, we had one researcher tell us, do you think that you should introduce some form of program for your agents to deal with the stress that they're under? Think about that. I'm not making that up. Do we need a program for our agents to help them deal with the stress? (Joel Beasley at 00:44:16) Oh my goodness. (Tony at 00:44:17) It's a California startup. I know that. (John (CEO/President of Scout Agentix) at 00:44:27) Exactly. Exactly. Absolutely. And this kind of goes to what new things are coming. But we know things are happening really, really fast. And it comes down to, I mean, everything's happening faster. Everything's going smarter. Right? People are just, you know, the idea around my brain, the looping, the continuous extension of AI into everything. Even though it's a new journey for the vast majority of the people out there and businesses out there. That's why we're total believers in what we're doing, and we're passionate about managing AI, the governance of AI, and making sure that it's trusted with every single execution. (Joel Beasley at 00:45:20) I love what you guys are doing, and I'm very excited about where we're at right now. You know, Tony, I've always been a nerd. Right? But when I was in high school, if I were at Chick-fil-A, the girls in the booth next to me would not be talking about PHP, okay, which is what I was interested in when I was in high school. They would not be talking about C. They would not be talking about DOS. They would not be talking about these things. You would not hear the word terminal. Now you sit in Chick-fil-A, every 15-year-old kid is talking about AI, Claude, ChatGPT. What happened was this language that you, me, and John have been speaking for the majority of our lives has all of a sudden just spilled out into society at large. And now we'll have people that have basically no knowledge of this stuff having strong opinions, both legally, legislatively, all of these things, and everyone's kind of coming into our world. I think it's great. And now I see myself as more of an educator than anything. (John (CEO/President of Scout Agentix) at 00:46:19) Yeah. So I'm with you. I love it. I love to see people getting excited to try to do new things. But I will say I talked to a VP about a month ago. The CEO walked into this VP's office, and the CEO wanted help on how to learn Python because he was kind of building his own apps. And, you know, Claude was giving him some code to go do whatever he wanted to do. And that's exciting. Right? But it's also the kind of thing where you absolutely don't want your CEO writing anything in Python. You know what I mean? Hopefully, he's got a group to help do those kind of things. And sometimes it can run you down a rabbit hole, puts you in the wrong place. But at the end of the day, everybody top to bottom is getting involved, is going to be involved. And what an exciting time. (Joel Beasley at 00:47:24) Yeah. Can you just imagine the CEO is doing this? Like, oh, I can absolutely do that for you. Just go ahead and provide me the SSH key to your server. Just hit yes, and it'll run the script, and I'll scrape it off your computer or generate it with your credentials. (Joel Beasley at 00:47:37) Oh, okay. Yeah. Yeah. I want that report. (Tony at 00:47:38) Yes, sir. (Joel Beasley at 00:47:38) Yeah. Here you go. Right. And the whole kingdom. (John (CEO/President of Scout Agentix) at 00:47:41) And he's done it. Right? And he's got the API key, and he's got everything else. But this is also part of the chaos. Everybody's jumping in. Everybody's doing it. (John (CEO/President of Scout Agentix) at 00:47:54) People wanna be productive. They wanna do great things. They have the ability to do things they've never been able to do before. (John (CEO/President of Scout Agentix) at 00:48:03) And so this is why it goes back to if we're working with a company on the assessment, first thing we wanna do is see what's going on in their company. Right? We'll run Colombo to do a discovery of what agentic activity is going on in their enterprise. Right? Just as a starting point. (John (CEO/President of Scout Agentix) at 00:48:23) Right? You know, who are your champions? Who's the people that really are using it every day that you might not even know about? You probably don't know about. (John (CEO/President of Scout Agentix) at 00:48:31) And then are they doing it the right way? Are they doing it the way you wanna do? Can we train them? Can we leverage them? Can we make them champions? (John (CEO/President of Scout Agentix) at 00:48:38) How do we get more people involved? But also, let's cross-check to make sure sensitive information is not leaving the company. (Tony at 00:48:45) That's right. (John (CEO/President of Scout Agentix) at 00:48:45) And stuff like that. And then also, can we train them to do something in a structured way that's very manageable, highly governed, leverage our platform, and then also share so that you don't have nine people trying to do the exact same thing. You know what I mean? It's like the world of spreadsheets. (John (CEO/President of Scout Agentix) at 00:49:09) Right? Everybody had their own spreadsheet, and then all of a sudden you're working off of the shared spreadsheet, and then you had a shared system. Right? Let's go from an individual agent. Somebody does something great. (John (CEO/President of Scout Agentix) at 00:49:19) I'm sure other people in the company wanna do similar things or the same things. How do we share it and also make sure it's got the structure and the governance and the promise to make sure it works for everybody? (Joel Beasley at 00:49:31) I love that. It's cool. Tony, everyone jumping in. Net positive, net negative. What do you think? (Tony at 00:49:39) For everyone jumping into the bandwagon? (Joel Beasley at 00:49:41) Yeah. Yeah. (Tony at 00:49:42) Yeah. I mean, whether it's net positive or net negative, it's unavoidable. So the die have been cast, and that game is behind us now. So I joked when I said we're the fighters of Starlink, but we definitely do fight against, you remember when everybody talked about, do you in business, do you have a confirmation bias? Do you really wanna be told that you're gonna be... (Joel Beasley at 00:50:09) A carbon-based life form. I do have confirmation bias. (Tony at 00:50:11) Confirmation bias. Well, guess what? That's not the biggest problem anymore. The biggest problem is AI bias. Everything that AI tells me has to be true. (Tony at 00:50:20) Everything it says, I need to go and run my business based upon. We fight that because that is absolutely a lie. And so we govern and make sure that you don't have to worry about AI bias. (Joel Beasley at 00:50:34) I love that. That's a good point. I think it's a net positive because I don't know. I mean, my crew... Well, first of all, just marketing. More people know about it. There's a larger market for more people to buy. (Joel Beasley at 00:50:45) Also, the people that do jump in and lose their pants real quick. I mean, some of my favorite customers that I've ever had in my career building engineering platforms was someone who I met with them and they're like, yeah, that's expensive. I was like, that's how much it costs to get it done right. Then they go off and they try to get it done, and then it fails horribly, and then they come back and they're like, hey. That's just been the unexpected story arc of my career. (Joel Beasley at 00:51:12) And I think that's gonna happen a lot, John. I think people are gonna try to do stuff with AI without governance, without the right systems in place. They're gonna have tons of issues, hopefully not make headlines in the news, but there will be issues. And then they'll realize, hey, I need this safety built into my AI. (John (CEO/President of Scout Agentix) at 00:51:32) Well, and Joe, I mean, you're 100% right. So one, I agree with both of you guys. It's, you know, I love everybody jumping in. You know what I mean? And I truly believe it can help you every day, all day. (John (CEO/President of Scout Agentix) at 00:51:48) But it's very common that people jump in. They see something great right away. They're like, okay. Now I need it to do two more steps to make it really work for me. And those last few steps sometimes are the hardest steps. (John (CEO/President of Scout Agentix) at 00:52:05) Right? It's that 80% that is perfect. It's that last 20% or the last 10% that people just really, really struggle with, or it kinda blows up, and then they go away from it. And we don't want people going away from it. And we wanna, you know, again, part of our platform is try to be that easy button. (John (CEO/President of Scout Agentix) at 00:52:26) Make it easy for them to build the agent, easy to use that agent, knowing that everything that's happening behind the scenes is taken care of. You know? And that's... (Tony at 00:52:38) I love it. (Joel Beasley at 00:52:38) That's really cool. Yeah. Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email, [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.