Ross Sonnabend With Verusen

Intro: Broadcasting live from the Business RadioX studio in Atlanta, Georgia. It’s time for Atlanta Business Radio, brought to you by Kennesaw State University’s Executive MBA program, the accelerated degree program for working professionals looking to advance their career and enhance their leadership skills. And now, here’s your host.

Lee Kantor: Lee Kantor here, another episode of Atlanta Business Radio. And this is going to be a good one. But before we get started, it’s important to recognize our sponsor, CSU’s executive MBA program. Without them, we wouldn’t be sharing these important stories today on the Land of business radio. We have Ross Sonnabend, and he is the chief product officer with Verusen. Welcome.

Ross Sonnabend: Hey, Lee, it’s great to be here. Thanks for having me.

Lee Kantor: Well, I am excited to learn what you’re up to. For folks who aren’t familiar, can you share a little bit about Verusen? How you serving folks?

Ross Sonnabend: Sure. Verusen is a purpose built AI solution that serves the maintenance, repair, and operations MRO space. The MRO space is part of the indirect supply chain that helps keep manufacturing lines up and running, stamping out widgets.

Lee Kantor: And then we’re here to talk about a new AI product that you guys have developed.

Ross Sonnabend: Yeah. So we just released a new bit of functionality around our solution in general that helps explain the decisions our AI solution actually makes.

Lee Kantor: Now, before we get too far into that, do you mind kind of giving us an AI 101 about what you’re seeing today on the landscape of AI, what’s available for enterprise entrepreneurs, organizations, and, um, where maybe there’s an opportunity there for them?

Ross Sonnabend: Yeah, absolutely. So I think first and foremost I is a catch all at this point, right? It means many different things to many different people. And so when I’m talking about AI, I’m talking about a collection of techniques that allow for more automated decision making. Uh, you know, starting with, you know, things like natural language processing, things like machine learning, things like agentic AI, you know, all the way up through generative AI, which is what I think most people think today when they talk about when they think about AI, they are thinking generative AI, which are the ChatGPT, the Claude’s, the Gemini’s of the world. Right. And that’s that’s because those guys are taking a lot of oxygen out of the room. And it’s very cool. It’s very cutting edge. But the big question for those types of technologies are, how can it be used in an enterprise context, whether you’re an entrepreneur or whether you’re an enterprise enterprise software company like we are or an enterprise itself. The big question is, out of all of these different techniques that are available to you, what’s the right use case and when is the right use case to, you know, to use these particular techniques? And so I think, you know, where where we are and what Verusen does is a, you know, purpose built application for this MRO space using purpose built AI.

Ross Sonnabend: So if you think of Gemini or you think of ChatGPT, all of those are generally available models. That means that they’re trained on, you know, publicly available information that’s out there in the world for the purposes of answering general questions. Right. You know, like the other day, I was asking ChatGPT about my low voltage wiring. So that’s great for answering general questions. But when you want to get down to domain specific areas like MRO, for example. There’s not a lot of training that’s happened on that stuff. And so where the opportunity is, is to go deep into domain specific areas like MRO and then build on top of, you know, build solutions that are purpose built for the domain that you’re going after using the techniques and technologies that are available to you.

Lee Kantor: And then so that’s what Verusen has done, right. Like, so you created us your own AI around this specific, uh, industry and the work that you’re doing.

Ross Sonnabend: Right. So like we said, MRO is a very specialized space, right? It’s focused on procurement, folks. It’s focused on maintenance and operation folks. And it is unless you’re in the space, you don’t really know about it, right. And so what Verizon. And so all AI solutions kind of start with the data. What Verizon has done is amassed a certain amount of data about the MRO space. Like for example, we have ingested over 40 million parts that are used in MRO space. That comprises over $12 billion in annual spend. We’re growing that and looking at, you know, expanding that data base, you know, with every customer that we bring on. And so that data forms the basis of all of the AI models that we build or the training that we do that allows us to be able to build what I like, I would call this like a small language model where instead of thinking about it as an LM, you know, we’ve built a model that has knowledge and is specific to to our space.

Lee Kantor: So now how does that help your client? Like how do they leverage this, um, amount of data that you’ve accumulated and put it in a machine that’s going to give it, I guess, more actionable information.

Ross Sonnabend: Yeah, it’s a great question. So in our space, along with, you know, almost any other space I’ve ever worked in, you know, data is a problem. In the first question they have to ask yourself is, what do I have to do to prepare my data to be used by some of these systems? What Verizon has done is we’ve kind of eliminated that question by using AI. So we take data as it stands in your legacy systems and ingest that data. When we ingest that data, we use advanced technologies using Llms and NLP to make sense of that data and fit it to our data model. Once it’s in our data model, then we actually run machine learning models to make recommendations on these inventory stocking policies. Right. And that’s something that historically humans have had to do, right? They had to look at maintenance records. They’ve had to look at the expected life of every little part that goes into a machine and make a gut feeling, you know, recommendation that says, I think we need to have, you know, five of these on hand and ten of these on hand. What we’ve done is we’ve taken that knowledge, built our machine learning models to make these recommendations, and then we use generative AI through this new capability that we just launched to explain the decisions that you’re making. So the benefit and upshot of all of that is a as a customer of Verizon, you can get to value in under 90 days, which is which is really good.

Ross Sonnabend: Second, you don’t have to tell us everything about your data. The system understands the data as it’s been given to us. We map it to our model and we’re able to make recommendations very, very quickly. And thirdly, you know, lots of companies use AI or ML machine learning to make recommendations or help make decisions, but what they don’t do is help explain why they made those decisions. And so there’s this criticism of AI that it’s a black box technology. In other words, if you were to go and say like, why did it make this decision? A lot of companies will say like, well, it’s just the AI algorithm making the decision for you based on the inputs that we’ve trained on it. You know, what we’ve tried to do is go that extra mile to be able to say, if you know nothing about AI, but you know a lot about MRO, we want to be able to give you the data that you need to validate the decision that was being that was made. And we do that in a plain English generative AI, uh, set of statements that we generate for every recommendation that we make.

Lee Kantor: So this AI agent is not a kind of a search box for your clients to use. This is just an engine for you to help your clients.

Ross Sonnabend: So I think that’s a really good distinction. It’s a really good point. When we talk about agentic AI, we’re not necessarily talking about chat bots. You know, when I talk about agentic AI, what I mean are task driven, purpose built, like for for our purposes, let’s call them little applications that do one thing really, really well. So this agent that we’ve built that we call our explainability agent all it does, its entire purpose in life is to look at the outputs that our machine learning models output and explain them using plain English understanding.

Lee Kantor: Right. But again, this isn’t like your clients aren’t going to a portal and typing in. Explain this to me. You are using this to give reports to your clients that explained things to them.

Ross Sonnabend: Right? It’s less about reports. And like in our in our user experience, every part that we have, we make a recommendation on. So every recommendation also has an explanation on the screen. Yeah.

Lee Kantor: And that explanation is a new development right. Like that’s the new thing.

Ross Sonnabend: Yeah. I mean it’s all relatively new, but that the new new thing is that we’re using generative AI to, to develop those explanations so that if you don’t know anything about AI, but you want to understand why our system made the recommendation that it did, it tells you, you know, 4 or 5 plain English sentences right there on the screen.

Lee Kantor: So, um, giving this new information to your clients, is that like, how is that helping them make better decisions or helping them, you know, make another dollar.

Ross Sonnabend: So the benefit of our system in general is usually working capital savings, cost savings, or not buying something that they would have otherwise bought. Spend avoidance. Right. So what we’re doing, what explainability does in general is it builds a layer of transparency, right. Because we’re not afraid of explaining why we made the decisions that our system made and trust. Right. So that now you have a skilled operator on the other end saying, okay, why did Verusen make this decision? We tell that operator why the decision was made. And then they get to either agree or disagree with it, make sure that they, you know, make sure that they agree with it. And then they, you know, they go and are able to execute that. That bit gives someone who doesn’t necessarily trust AI the ability to feel good about the decisions that it’s making. Number one. Number two is in our space. At least, there’s a growing problem of skilled workers aging out and not enough skilled workers coming back into the same roles. And so our system, being able to explain the decisions that it’s making, can actually help people who maybe have less experience in making some of these decisions, gain more trust and be able to make the decisions that they would otherwise not be able to make.

Lee Kantor: And does this happen faster than it did previously?

Ross Sonnabend: So it is for every recommendation that we make. The explanation is there instantaneously. So if you were to pull up a record in our system, you would the first thing that you’re going to see is the explanation that the system has given it. So other before we implemented this, you know, what you would do is you’d go and you’d look at the screen. It would have these different metrics and KPIs on there saying, here’s what your old policy was, here’s what your new policy is, here’s the service level it’s expecting. And then you, as a skilled operator, would have to put all those different data points together and say like, do I agree with this or disagree with this? What the explainability does is it removes the need for someone to spend the time connecting those dots. We’re connecting the dots for them.

Lee Kantor: And then what is this thing rolled out right now or what stage are you at in its development?

Ross Sonnabend: Oh yeah. So this is this is our first generative AI solution, and it’s the first agent that we’ve built. It is generally available today with our software. Um, but the really cool and exciting thing is, is that this is a foundational capability for other types of AI agents that we are currently in development on.

Lee Kantor: Oh, so this is the first of many?

Ross Sonnabend: Absolutely. Like, we are committed to Agentic AI, which again is more focused on building task oriented AI applications that do one thing really, really well. So like for example, one of the next agents that we are, you know, working on is around accepting recommendations, right? So today in our system, you have if you have 250,000, uh, materials that you are keeping, an inventory will make a recommendation on all 250,000. And we expect that someone is going in and reviewing that and making a decision. That’s not that’s that’s good. And it’s important. It’s kind of the state of the state. Uh, you know, three years ago Today, we think we can use AI to help accelerate those decisions and acceptances with human guardrails on there so that so that humans are not taken out of the decision loop. But instead of focusing your time on accepting or rejecting a recommendation, we really want you focused on achieving the business goals that you’re trying to achieve. And in this case, those business goals are identifying where you already have materials and, you know, in your company instead of having to buy them. Identifying obsolete materials, identifying where you have, um, off contract buys that can be made on contract buys. These are the kinds of things we want our customers focused on. Not not, you know, having to go through 250,000 recommendations and make a an agree or disagree decision.

Lee Kantor: Now, how do you decide which, um, kind of specific thing to focus on next. Like how are you prioritizing this? Are you getting input from your customers, or is this something that internally you’re doing on your own? Like how do you decide you know, which is the next? You know what comes next on the roadmap?

Ross Sonnabend: It’s a great it’s a great question. Right. And it’s always, you know, this is the the biggest burden of of being a product manager in general is how do you make the decision on what gets prioritized when you have all of these different competing priorities. Right. So we’ve got customer feedback. We’ve got our own set of views. We’ve got prospects feedback. We’ve got our sales team’s feedback. You get feedback from, you know, ten, 15 different vectors. Um, ultimately it comes down for us to two things. One is what about our system? Can we use AI to improve and kind of ten-x, right? How can we make the experience ten times better than it is today? And that example of just making a a, A giving an explanation. You know, today when we didn’t do that before, that’s like a ten x type of improvement. Um, so that’s one that’s one branch of the decision tree. The other branch of the decision tree is, you know, where are customers struggling, right? Where do customers want help? Uh, because that’s the lowest hanging fruit. We see challenge here. Let me help fix that challenge. And that’s where we, you know, we got to this acceptance agent idea. It’s customers don’t want to have to go through and accept 250,000 recommendations. They want the value that you get after you review those recommendations. So how can we use AI to help them get to that value more quickly, you know? And then thirdly, you know, um, sales. Right. What are gaps in the market or gaps in our product that we can use AI I to to shore up. So those are kind of the three main vectors that we look at when we think about how we productize the so product.

Lee Kantor: So if somebody wants to learn more, have more substantive conversation with you or somebody on the team, what’s the website? What’s the best way to connect?

Ross Sonnabend: Uh, the best way is uh verusen.com. So it’s www.verusen.com or verusen.ai. And, you know, we have lots of materials available there, demos available. And then if you want to get in contact with someone, there’s a button right in the middle of the screen to say contact us.

Lee Kantor: And Verusen is spelled v e r u s e n.

Ross Sonnabend: Yes.

Lee Kantor: Well, Ross, thank you so much for sharing your story today. You’re doing important work and we appreciate you.

Ross Sonnabend: Thanks, Lee. I appreciate you having me.

Lee Kantor: All right. This is Lee Kantor. We’ll see you all next time on Atlanta Business Radio.