THE MERIDIAN POINT
How to Spot AI Hype Before It Burns Your Budget
Guest: Alex Brungardt, Director of Business Development, Radus Software
Host: Kumar Dattatreyan
KUMAR: Hi everyone, Kumar Dattatreyan here with The Meridian Point. Today we're joined by Alex Brungardt. I should probably ask him how to say his last name. Hopefully I didn't butcher it. He's the Director of Business Development at Radus Software, the company behind the Metronome delivery and compliance platform. I don't know what that is, but we're going to find out. Here's what makes him worth listening to. Alex sells automation into the most paperwork-buried corner of technology, federal governance and compliance, and his pitch flips what you'd expect it to be. The fix isn't necessarily more AI. It's knowing exactly where a human decision has to stay human. He came up in sales, not engineering, and that outsider's eye is his edge. He's here to talk about automating the front end of compliance and how to spot AI hype before it burns your budget. So without further ado, Alex, welcome.
ALEX: Thank you. It's great to be here. You pronounced the last name perfectly, so...
KUMAR: Okay, good. I appreciate it.
ALEX: Yes, I was actually wondering how you pronounce your last name too, so you clarified that for me as well. Thank you.
KUMAR: Yeah, Kumar Dattatreyan. You know, this is a funny story about that. I was working at a company a long time ago. It had just started, and there was this security guy at the front desk who seemed to know everybody who came in and out of the building. There were about 4,000 of us, and he seemed to know everyone. A few weeks after I started, he said, "Hi, what's your name? Where do you work and what floor?" and so on. Within a few weeks, when I came into the office, he would go "choo choo" and welcome me. I'd go up and think, "Wow, what's he saying choo choo for?" Finally, I asked him a week later, "Why do you do that choo choo thing?" He said, "That's a train. Choo choo."
ALEX: That's funny. Yeah, I love that.
KUMAR: So that was his secret, right? He was able to remember people by associating something about their name with the sound that it makes. That was his secret power. And he was super smart. I think he was studying to be a lawyer while working as a security guard in this building. A year after I started, he was gone, so I assume he got his law degree or whatever.
ALEX: Yeah, it was really cool. That is cool.
KUMAR: It always amazes me when you ask people what they're doing and take the time. It's amazing what people can juggle to achieve their goals.
ALEX: Absolutely. And I always respect that too, when I see somebody burning the candle at both ends to make something happen.
KUMAR: Yeah, it's all about continuous learning, continuous improvement. That could be a topic for one of our podcasts, and we may have to pivot from this one if that's something you want to talk about.
ALEX: I'm totally up for that. I think there's a lot there. They always say that progress is made when the eyes aren't on you. Progress is made when no one's looking, and if you're going to put the work in, it's in the times when no one's watching you.
KUMAR: Yeah, that's a really good thing to keep in mind. Let's keep that in our back pocket and see what you're doing.
ALEX: I think it's relevant to the topic we're talking about today to some extent, because with how fast everything is changing, from AI to governance, risk and compliance factors, what we're trying to do is make it so people don't have to stay up late at night trying to keep up with GRC. They can just build more value, wake up refreshed and ready to go the next day, and focus on delivering business value.
KUMAR: I love that. Automate those pieces.
ALEX: Yeah.
KUMAR: So let's get into it. You'd be the first to say that you never set out to be an engineer, and yet you spend your days selling automation to federal teams buried in compliance work. The conversation we had to prepare these questions was about three months ago, and as fast as things move, these may be stale. Tell me if they are. You spend your days selling automation to federal teams buried in compliance work, right? Is that still what you do?
ALEX: Yes. In addition to other digital transformation efforts, GRC is one of the things we're helping the federal government automate with our low-code, no-code platform.
KUMAR: Okay. And what value do you bring in translating what the customer wants versus what the engineering teams may miss?
ALEX: Well, first and foremost, it starts with the relationship and becoming a trusted advisor to my customers. When you become the trusted advisor, they give you more time, and that time is what you spend asking questions and probing into the business problems. With AI and all of the automation available to us, all the different solutions really depend on starting with why. What is the point of the automation? What is the point of implementing an AI solution here or there? What business value is it going to provide, and how much does it cost?
KUMAR: Yeah, I love that. I think most things should always start with why. Me and my partners developed a product management course, and we created a little picture that illustrates the stages you go through to build a new product. It starts with why. Why are you building it? Then, who are you building it for, and how will you know that you've achieved the outcomes you set out to achieve, for the business, the customer and the people using it? Obviously, you're not building this just for the sake of building it. You're trying to make money. You're trying to make something that people want and use and get value from, which then returns value to the company, or to the shareholders if it's a public company. So the why is really critical to answering that question.
ALEX: The why, and then being able to be patient is another thing. I actually studied healthcare and now I'm in IT, so it's just an outside perspective. What that outside perspective gives me is the ability to coach folks on what is actually achievable today. Some of their stakeholders are so busy with their day-to-day activities that a system update or a rebuild gets put off for three or four years because they think it may be costly and consume a bunch of time. With AI and automation now, some of these solutions you can tackle in less than a year. So being able to be patient and illustrate from multiple angles what companies are capable of achieving today, and giving people options, is another important thing. You get stuck in your day-to-day, you get bogged down, and all of a sudden you don't realize how capable some of these enterprise software products are. And in a heavily regulated environment, sometimes your organization is limiting the uses. One customer I was speaking to, a pharmaceutical company, said they're able to use AI, but it's limited to reporting and data and analytics. They can't do any development or testing with it.
KUMAR: Interesting. It is very true that bringing that outside perspective can be really helpful, because if you're in it, you tend not to see what's outside of it. And so you break the cycle a little bit. You're in this bubble, this ecosystem, this echo chamber really, which is probably a better term, and you're not seeing what you're not seeing. You have blind spots. So coming from the outside, of course you have a bias, you have a perspective, you're trying to sell the software that's going to help them, but it's still an outsider's perspective, so you're able to see some of the things that maybe they're not able to see.
ALEX: Yeah. And what's beautiful about working with the federal government, too, is that a lot of times you can separate the software from the services. Some of the engagements you're pursuing, you're simply coming into the existing architecture. There's not a software play.
KUMAR: Right. Interesting. So it's all about helping to serve that goal, which is really nice.
ALEX: It's a lot of fun, especially when you can get access to the stakeholders.
KUMAR: Right. So, Alex, you built your pitch around a specific idea, that the place to put automation in compliance isn't in the doing part of it. It's really more on the front end, the ideation and the planning. Walk me through what that looks like.
ALEX: Yeah. What we tend to see is that governance, risk and compliance is typically an afterthought, something organizations try to force into more of an operational role. What we've done is similar to what's been done in Agile with security, building security into the front end. We've started to build governance, risk and compliance, GRC, into the front end of the planning and development lifecycle. To do so, we created a tool that allows organizations to upload documented components of governance, risk and compliance, whether it's an internal document or a federally mandated or municipal document or order that you have to comply with. In the federal world we have STIGs and the authority to operate process. In the commercial sector, depending on what industry you're in, it could be HIPAA, FISMA, all of these various GRC components. We upload them into a system, and then you begin to ideate and plan your work. When you begin to plan your work, the system pulls together your features and the backlog for it, and pulls governance, risk and compliance features into a backlog for you to accept or reject into your actual permanent backlog.
KUMAR: Yeah, that makes a lot of sense. In a way, these are NFRs, non-functional requirements. Typically, with GRC, security, compliance, those types of things, people don't think about them until the end, when you're closer to deploying something. Then there's a mad rush and a mad scramble: "Oh, what are we missing? Oh no, we have to go back and refactor all these things because we didn't think about how to implement them." So you're shifting it all the way to the left and saying, for this type of service or product, with these types of requirements from a GRC perspective, this tool will give you all the features you would need to implement. Right?
ALEX: Yes. And as changes are made, it helps you automate the updates. Especially for a distributed organization. I mentioned municipal because when you're across multiple countries and multiple states, all of these various boundaries create GRC components that you have to account for. So we try to eliminate the guesswork. We try to eliminate the planning piece so that the organization can really focus on keeping their products compliant and building more functionality.
KUMAR: Yeah. Is it limited to the GRC side, or does the tool, it's called Metronome, right?
ALEX: Metronome can help you ideate in general. The next layer of this is really exciting. It's going to be more of an AI, or orchestration, layer. We're planning for a future where people sit in front of a Metronome dashboard, and they're going to have Claude, and Claude is going to be connected to the systems they use to produce products and applications. Metronome will be the orchestrator, in a way, and it'll be the business health radiator, the information radiator.
KUMAR: Interesting. Very cool. All right, so where does the human stay in the loop here? Some decisions, obviously, you can automate. If you can simulate it, if condition A, then you do this, and so on. AI systems are not deterministic, right? If you put the same conditions in front of an AI, it may pick a different response or a different answer. So how do you guard against that?
ALEX: Well, one, the final way is that every feature or story that you accept into your backlog has to be manually accepted. But before that, we are big on AI. Currently, AI is automation, and there's going to be some kind of manual logic programmed into this tool that determines how decisions are made. Typically, we're not leaving the decisions to the machine in the first place. What we don't want is for the solution to create random artifacts. What it's doing is pulling existing pieces of work and combining them with the GRC components that you have to account for, so that you can build the GRC.
KUMAR: So if I'm understanding you, the tool is automating the parts of the work that don't necessarily require a decision. The decision still...
ALEX: It's research. We're automating the research. We're automating the documentation and the updates, all of the research and R&D portion of it. But the human has to be involved. And we're serving federal organizations and federal agencies, so we have to be careful about how much we depend on AI outside of these. A lot of this stuff is available through just a regular rules engine, right? You can build all the logic to do whatever research is needed. Of course, AI precisely amps up that capability, because it can take large chunks of information and data and quickly scan it and make sense of it, and so on. Where our value is added is when you integrate with the existing systems architecture. When you integrate with the existing system architecture and the existing tooling, you're able to identify gaps in your governance, risk and compliance that you actually need to fix. So it's proactive, but it's also an audit system.
KUMAR: Interesting. Okay.
ALEX: Where Metronome is really powerful is when you use it as a tool to plan and track your work across the organization. The governance, risk and compliance piece of that is one of our more advanced AI applications within the Metronome platform. But the platform itself was built as a low-code, no-code platform to provide a command center across your existing and future systems.
KUMAR: Okay. So the name is indicative of what it is. A metronome keeps time, right? If you're a musician, you use it to keep time while you're playing your music. So it allows an organization to keep time and keep delivering value, I guess, by providing an overarching view of all the services and applications that are out there. And I guess the AI layer is what will help your clients make better decisions, right?
ALEX: Yes, from a client's perspective, depending. And you can also build in... There's a chat feature. You can ideate with the chat feature. You can also write and send reports directly with the chat feature, which is very helpful.
KUMAR: So what's your test for when a decision has to stay with a person and when it can be automated? Or is there one?
ALEX: I would say it's case by case, of course, but I think there should always be a human watching over it. I'm not at the point where I trust, especially when coaching my customers on unleashing a complete automation solution. I drive a Tesla and I have Full Self-Driving, and it's gotten so good that I trust it to make decisions for me all the time. It's driving for me.
KUMAR: You probably trust it more than yourself at this point.
ALEX: Yeah.
KUMAR: Forget how to drive.
ALEX: No, I have another car for that. It's a manual shift, old school, gas-burning. But when I get in the Tesla, I get lazy. I let the car make all the decisions.
KUMAR: So what's to say it can't do some of the things that we think are reserved for a human? At what point do you get to enough of a margin of error where you say, "Okay, this thing is functioning exactly like I would as a human, and we can take one more decision off the human's plate"? I'm not advocating for that. I'm just wondering what you think.
ALEX: No, I think it's test first. I think that's a cultural thing, right? When it comes down to it, you have to have a culture of testing. That, to me, is where it led when I started to think about the question a little more: if you're testing appropriately, then you should be able to trust the system. I was taught early in my career, by a DevOps coach, that a bug is a test not yet written.
KUMAR: Yeah.
ALEX: So there you go. If you're able to effectively test the system and you're getting 97 percent accuracy, or whatever your threshold is for making that decision as an organization, then go ahead and automate those decisions. But at the same time, I'm big on observability. I'm big on checks and balances and creating systems that produce near real-time data, so that if the system makes a bad decision, you're able to identify it more quickly than not, right?
KUMAR: Yeah. In our prep call, we talked about this, the case for building agents with a test-first mindset. That's the phrase you used, training them on the testing itself, so the need for a human to check every output keeps dropping. For someone who wants to try this without betting the business on it, where should they point their agent first? How do they know it's actually trustworthy?
ALEX: I don't think program managers should be writing their own reports anymore. And if they are, they probably don't know what capabilities they have, even within their email, like the Microsoft or Google suite. That's where I would start. If you're getting started with AI, start with the tools you have. Look into the settings, look into the new features they're releasing, and determine if they fit within your AI policy. If they do, accept those settings as opportunities to test. Initially, R&D is a great place to start, implementing some AI into new applications where you aren't putting existing applications at risk and you're not asking your customers to go through a huge learning cycle. Maybe you're making a new, more optimized version available to a different audience.
KUMAR: Right. So a lot of your work is in the GRC space, understanding what the requirements are and pushing the requirements process as far left as possible: what are the stories and features that need to be in the backlog for a team to make sure that whatever they're building is compliant? Looking out a few years, if the boring, document-heavy work of compliance is the first thing AI takes over, because you're doing that, right? You're chewing up all the requirements from a GRC compliance perspective and saying, "This is what you need for this project or this product, because it's in one locality or it's across many municipalities. So these are the things you need to do." Then what does a governance and risk team look like in, say, three years? Is it more people, fewer people, different people, different skills? What would you say?
ALEX: I would imagine a similar size of team, just building different skills. What you'll find, again, is that organizations will start to do work they never wanted to do before. With AI and automation, you're going to have more time for the tasks you put off. You're going to have less technical debt.
KUMAR: That's a good point.
ALEX: What we find is that there's a lot of overlapping GRC. Your organization is given all of these federal requirements to account for, but then it builds its own policies that are sometimes even more stringent than what the government provides. What we'll be able to do is that auditing piece, something organizations didn't want to take on before because their leaders were prioritizing an ERP implementation, going to the cloud and reducing cloud costs, and so on. Auditing GRC and implementing more lean Agile were four, five, six down the list. Now you're going to be able to do all of those things.
KUMAR: Yeah. It makes sense. And I wrote an article that dropped today. Of course, this is a recording, so this will probably air sometime in the future, and it will have already dropped by the time you watch this, all you listeners. It's about the disruption AI is causing in the job market. There are all kinds of statistics out there about how many jobs AI has chewed through already, and they're all over the place. Some are pretty high, some are pretty low, and some say there's really no change. And of course, there are lots of articles and data out there that say that so far the ROI for AI is nothing or negligible. But my older son works as a product manager in a commodities trading firm, and they're organized as teams of two, a product person and an engineering person, with a bunch of agents. They were bigger teams. They had whole development teams, and they probably still have all their developers, they're just organized this way at present. Through the agents, they're creating a bunch of new code and new products and enhancing existing products with just two people and a slew of agents, chewing up some tokens in the process. By his estimation, they're probably spending five grand a month in token costs, which is a lot cheaper than a team of developers. What's your take on where the industry is going? Are we really moving from the two-pizza team to a two-slice team at some point in the future?
ALEX: I'm an optimist. In some organizations that will occur. Like you said, you'll see some organizations with less headcount. And you'll see organizations that do what your son's organization is doing, flatten it out and add more agents. They probably have, what, five times the number of teams now, because they flattened it out and gave every pair of PM and developer a bunch of agents to work with. I think that's a great model. I prefer that model because everybody keeps their job and they actually get more production out of it. But I also think we're going to see a lot of job growth, and a lot of opportunity for American innovation. This country is known to be one of the most technically innovative countries of all. That's what we're built on. Optimistically, I hope all of these people are given more time to think creatively and build their own products. It's going to stimulate the economy that way. I think that takes a lot of time, for people to train and...
KUMAR: For sure. It's going to be an interesting next few years. It's been really interesting the last few years, especially since the pandemic.
ALEX: Yeah. Oh, and I think kids should focus on learning some technical stuff, but also on English language arts. What's real text? What is computer-generated text? How do I talk to people versus how do I talk to a machine? I was talking to some folks, small business owners who worked with the National Geospatial-Intelligence Agency, the NGA, and they were saying how... Oh, I lost my train of thought there.
KUMAR: You were talking about the ability to read text and interact with people rather than machines. That's going to be a skill young people need to keep, right? Because we're increasingly communicating with machines and less and less with people. Or maybe we're still communicating with people, but through a machine interface, like Instagram or Facebook or whatever we're using, and we're sort of...
ALEX: They're saying that gone is the day of writing a proposal. You're going to create a prototype. And if you're sending proposals and not prototypes at this point, you're missing out.
KUMAR: Yeah, that's really interesting. It is changing a lot. The world is definitely changing. All right, so we're getting toward the end of our time box. A couple of fun questions. If you weren't in software, what would you be doing?
ALEX: If I weren't in software, I'd probably be somewhere in health administration, managing a hospital or doing something at the crossover between people and technology. I really enjoy learning big systems, and I enjoy the healthcare system quite a bit.
KUMAR: That's very cool. In our prep call, you said, "Gone are the days of the lawyer. You just feed the contract clauses into an AI tool, Gemini, Claude, whatever, and you're good. You get the advice you need." So what's the last thing AI talked you out of paying for?
ALEX: Taught me what?
KUMAR: Talked you out of paying for something that you would have paid for before, and now you just use AI to do it. Of course, you're still kind of paying for it if you have a more expensive plan or whatever.
ALEX: Sometimes AI talks me into paying for stuff. I'm sitting here thinking, should I do this myself, even at home? You can translate this into the cost of automation, right? Should I do it myself or should I pay somebody? A lot of times, if you feed those problems into AI, it'll help you justify one way or another. The other day we were looking at buying a new couch or something like that, and it gives you all of these solutions, whether you should refurbish or get a new one. But yeah, I use it. I have it talk me in and out of purchases.
KUMAR: So it's helping you make decisions every day, all day.
ALEX: My son is into trading cards, and we use it. Now they have applications for that too, but why do you need an application if you have Gemini? You pull up Gemini, scan the card, and it tells you. You can justify to your kids whether or not it's a good deal, whether or not to buy. I'm even teaching my son how to use it that way.
KUMAR: That's great. Teaching him about economics to some extent through trading cards and things of that nature.
ALEX: I think it's very common. These kids are into these collectors' markets, and they're learning very raw economics. It's supply and demand at its finest.
KUMAR: It's the lemonade stand of the... [Transcription gap: the source text repeats a single word here and part of the conversation is missing.]
ALEX: ...do these various procedures, vaccines, you just ask Gemini. And it gives you... They're going to give a newborn a bunch of vaccines, and you think, "What are the odds of them being exposed to these things?" That's what Gemini will ask you: "What are the odds of them being exposed to these things?" Then you think, "Oh, maybe I want to do that one. Maybe I don't need to do that one yet." So the logic is really helpful.
KUMAR: Just be careful, because these models are all based on what they know from the internet, so they take in all the viewpoints and feed back what they hear. You have a pediatrician for a reason.
ALEX: Yeah, that's why you have to... I love those. They have the training. I love those, they have actually [unclear] to say, like, "I didn't get my MD from Google."
KUMAR: Right, exactly. Exactly. All right, one last one. The most overhyped thing in AI right now is what?
ALEX: The most overhyped thing in AI right now is calling it artificial intelligence already. When we get to the next level, AGI, artificial general intelligence, that's when we'll really go from automation to artificial intelligence. I think calling generative AI artificial intelligence is overhyped in general. But what's underhyped, I think, is the race to achieve AGI as a nation, as a national interest. I don't know. I'm a little scared about that. Once it achieves that general intelligence, what's to say it's not going to replicate and keep improving?
KUMAR: I'm scared of it too. My theory is, though, if we don't, who will?
ALEX: Yeah. So, you know, I don't know that we should. Any... humanity... I don't know if anybody should.
KUMAR: Yeah, it's one of those things where if we could all agree not to, maybe I'd have to get on board, you know? But I don't think that's going to happen.
ALEX: Yeah, it's a little scary.
KUMAR: All right. Well, that's a slightly ominous tone. But if it happens, it happens, and humans will adapt and evolve along with it. To me, right now, the AI we have is more about augmenting our knowledge and intelligence. What I fear most, I guess, is that we're relying a little too much on it to make decisions. I do it all the time, and I have to tell myself, "What decision would I have made without the information I got from Claude or Gemini or whatever?" That is a fear of mine, that we start to rely on it more and we stop using our brains. We stop thinking. In fact, I read something somewhere that our brain mass has declined over the last hundred or two hundred years, and it will continue to do so, because we keep building tools to augment the thinking we have to do. Those tools are getting more and more sophisticated, so our brains don't have to work as hard. And what's the cost to humanity? Who knows? Maybe it's okay.
ALEX: I'd be fine. We only use 10 percent of our brains anyway, so maybe it's all right. I'd be fine without a lot of this stuff. To be quite honest, I embrace the technology, but I'd be fine without a lot of these technologies. I would adapt. I was born before all this, before the internet and all that, and text messaging came out the year I was born, so I grew up googling stuff. I've had a smartphone since I was... You know, I had the first Samsung Galaxy that came out. I remember. So I've googled things my whole life. But what we'll see soon, and you don't see it anymore really, is that databases, being able to do a Boolean search, doing real hard-nosed research, is a dying art.
KUMAR: Interesting. Yeah, I guess we'll see. Well, Alex, thank you so much for appearing on the show. This was a fascinating conversation. We didn't get to what we started with, which was more around the disruption and the learning people need to do to stay on top of things. When people aren't watching, what are you learning? What are you teaching yourself? How are you staying relevant in today's AI-dominated world? Maybe that's another episode for us to think about.
ALEX: Absolutely. I'd love to.
KUMAR: All right. I appreciate your time. Thank you.
ALEX: Of course.
KUMAR: Yeah. Thanks for watching, everyone. Bye.
