Jason explains that Anderson Trucking Services is a full service trucking company, from box trucks and flatbeds to specialized heavy haul such as wind turbines, with a large logistics division moving loads around the clock. When Matthew asks whether AI helps match loads with trucks, he says the company started on the back end instead, with the proof of delivery, bills of lading, and paper trail that follow every load, a manual process AI can automate for drivers, driver managers, and logistics brokers. He says they are seeing huge returns, and that many of the other security leaders he talks to use AI throughout application development: reverse engineer an application, have AI build the new version, then have AI check it against the old one. Human stage gates stay in place so nothing moves forward unchecked, and he says a 20 year old legacy ERP system can be traced in minutes instead of days. On the security side, he says the company is dipping its toes in and moving slow and controlled.
Drawing on a pen test from a previous role, Jason describes a black box device plugged into the network after the security program was in place. The SOC identified it within minutes, and a network detection tool let him watch the testers move through the environment in real time. He chose to let the test run so the SOC, the testers, and his own team could each analyze what the testers reached. A honeypot of decoy servers that he had once thought worthless absorbed the attack, and he says attacks went from about 50 an hour to 5,000 an hour. He says typical pen test scripts go after servers rather than endpoints, so decoys that look like older servers draw them in, and an intern built dozens that looked like finance and database servers. On assessments, Jason says to do it yourself: he calls the assessment many companies bring a foot in the door sale built on fear, and he prefers to know his own gaps and ask how a vendor’s product solves them.
Asked how to balance AI’s promise, Jason says vendors mostly pitch basic back office automation such as accounts payable and collections, while the ones separating themselves understand logistics and transportation and the pain points. At the end of the day, he says, only two things matter: increasing revenue or decreasing expenses. He describes a vendor that offered to build a new application in six to eight weeks for a fixed price, with the customer keeping the intellectual property, a model he calls completely different from subscriptions. He has not committed yet but sees about half a dozen candidate projects. He also describes a peer who pulled data from his Microsoft security tools into a self built single pane of glass, and says you can build your own far faster now. His main example is an AI phone assistant: instead of a phone tree, a driver could call in, ask for a next load number, and have the pickup address sent to the driver app, taking workload off driver managers who handle dozens of drivers at once. He says the company’s product is its drivers, a view he carries over from the sports industry, and that the phone assistant is on the road map as one way to improve their experience. Jason closes by pointing listeners to Anderson Trucking’s website and his LinkedIn profile, and Matthew invites him back for a follow up.
Resources mentioned in this episode
This episode is brought to you by CyberLynx.
CyberLynx is a Bethesda managed IT and cybersecurity company. Local techs you know, not a call center. Month-to-month. 24/7 intrusion detection.
We help growing companies with managed IT, help desk, backup and recovery, and a fractional CIO.
Talk to us at https://cyberlynx.com/contact, info@cyberlynx.com, or 301-798-9170.
No Competing Priorities: Securing a County and Putting AI to Work with Eric Matthews - Ep 254
The Summer of Copilot: Education, Context, and Thinking Strategically with Thomas Matthews - Ep 253
Train the Tool: Guardrails, Separation of Duties, and Responsible AI with Roberta Anderson - Ep 252
Guest: Jason Meszaros
CISO and Director of Infrastructure and Cybersecurity
Anderson Trucking Services
Host: Matthew Connor
Matthew Connor: Matthew Connor here, host of the Cyber Business Podcast. Today we're joined by Jason Meszaros, CISO and Director of Infrastructure and Cybersecurity at Anderson Trucking Services. Jason, welcome to the show.
Jason Meszaros: Thank you. It's good to be here.
Matthew Connor: It's good to have you. Before we get too far in, a quick word from our sponsor.
Sponsor Break: CyberLynx. Hackers are using AI to conduct machine-speed attacks. Is your security keeping up? CyberLynx sells industry-leading AI-powered cybersecurity solutions that detect threats in real time, so you know about an attack before the damage is done, not after. Learn more at cyberlynx.com. That's cyberlynx.com.
Matthew Connor: And now back to our show. Jason, for those who aren't familiar, can you tell us about Anderson Trucking Services and your role there as CISO and Director of Infrastructure and Cybersecurity?
Jason Meszaros: Absolutely. Anderson Trucking is a full-service trucking company. We have a number of specialized business units, so we do everything from box trucks to flatbeds. We have specialized divisions where we haul wind turbines, heavy haul. We also have a very large logistics division where we find customers who need stuff moved, and drivers and trucks to move it. So point A to point B, we are moving loads 24/7. That's what we do.
Matthew Connor: Love it. Now, I would imagine this is something where AI really helps on the operational side. I think historically, the logistics of matching loads with trucks is kind of labor intensive on the human side. Has AI been something you've been able to leverage to speed that up?
Jason Meszaros: Not on the front end. It's on the back end that we started with our AI work. When you deliver a load, you're going to have proof of delivery, the bills of lading, all of that documentation and paper trail that goes along with the load. There's a whole manual process that's been in place forever that we can use AI to automate and make much faster and smoother for the drivers, the driver managers internally, and our logistics brokers as well.
Matthew Connor: It's funny, I don't know if you caught the Blackstone president, Jonathan Gray. He was doing his annual talk maybe two weeks ago, and it was really interesting. The really interesting part I found was when he was talking about the ROI for AI. Blackstone has a massive portfolio of companies that basically spans the entire economy. It's just incredible the scale of it and how tied in they are to all of these companies, helping to operate them and grow them, so their access to information is rather unprecedented. What was really fascinating was the ROI they're getting on AI. It was something crazy. I want to say it was something like [unclear] to one. And these aren't small companies we're talking about. These are large companies getting a massive return on investment, which I think is funny, because if you rewind a couple of years, the story we'd been hearing was, "There's no ROI. This is all nonsense. People are just excited about it. It's a big waste." The naysayers were really big about that. And I think what we see now, at least in the last six months from my perspective, is that the power and capabilities of AI are so obvious and so impressive that I think that number makes a lot of sense even for smaller businesses. And I think it goes to what you're doing on the back end, not trying to replace people on the front, customer-facing side. That's got to be a massive return for you guys as well.
Jason Meszaros: Yeah, it is. We're seeing huge return. We're also seeing it in our development. A lot of the other CISOs I talk to leverage AI throughout their application development processes. I could throw out a bunch of examples, but a lot of them are using AI as checkpoints along the way. So not only are you reverse engineering, "We've got to rebuild this application or this interface," you reverse engineer it, you have AI build the new version, and then you can have AI check it against the old one: does it do all the same things the old one did? So you have automated testing built right into it using AI. You always have those stage gates with a developer or a human interaction to make sure you're not pushing things past each stage gate without checking first. But we're seeing huge development savings and time savings. We have a 20-plus-year-old ERP system with legacy code that needs to be rewritten. We can reverse engineer that in minutes, instead of days for a developer to trace back where all the code touches all the other code. It's pretty amazing how quickly it goes.
Matthew Connor: You know, it's funny, because I was just listening to a leading venture capitalist. I'm not going to get the name right, so I'm not even going to try. He basically said it's nonsense that you can use AI now to work with your legacy code and build the new systems. And I'm like, that's the most incorrect thing I've heard in a long time. What are you talking about? In what world? It's so obvious, and it's so good at it. It blew my mind, and this wasn't a dummy. However, they were clearly disconnected from reality on this, and I just can't believe anybody would say that. You're a great example. It works like a charm. It's so hard, especially with the legacy stuff where the developers have long since gone. Anytime a new developer is looking at it, it takes so long to go through and understand everything they were doing. Humans don't document code. Historically, there are very few coders who are great at it. So now you're trying to figure stuff out, and it takes AI seconds to understand all of it. I'm glad you shared that story because it's spot on.
So let me ask you, when it comes to AI and cybersecurity, one of the things I think is really fascinating is that there are some cool products coming out. I'm a big fan of Darktrace. There's a newer one called Dope Security that I thought was super cool. There's Abnormal, and there are some really cool ones out there. But beyond the cool new products, the thing I think is really exciting for the CISOs of the world is being able to have this level of intelligence that can so quickly look at your SSP, your systems, your documentation, your SOPs, your policies and procedures, all of it. Then it goes and accesses your firewall settings, read-only, and sees where everything is and says, "Yeah, you said you're doing this, but this has been misconfigured. It's working, but..." The leveling up in terms of professionalism and thoroughness is amazing. I think the challenge before AI was always that there was so much information and so many things that had to be done. No matter how many people you used, it was hard to get it all done. Now you can leverage AI to see it all, understand it all, check it all, and say, "This is where we're missing. Do you want me to fix these things?" "Yeah, please go ahead and fix those." Or even if you don't, "Thanks for pointing this out. We'll go fix those." That's amazing. The concept for audits, in my view, was always: this is your work, let's check it. Now you can audit like that with near perfection and nearly instantaneous results. To me, that's where we're moving anyway.
I'm curious, are you seeing that and doing that? I get all excited about this stuff, I think it's fantastic. But are you guys doing stuff like that, and is that a good use of it? Or are there some CISOs who say, "Dude, no, AI doesn't get to touch it. We do human stuff here because we want to be the decision makers," which I take exception to. Where are you guys on this?
Jason Meszaros: We are, at ATS, dipping our toes in. We're moving slow and controlled. In my previous role, I worked for a professional sports team. You mentioned Darktrace, and Darktrace was a huge part of a pen test example from that role. We did the standard black box, where the pen tester sends us a box to plug into the network and then we close our eyes and let it do whatever it does. The first time we did that, we didn't have any of the tools in place. You literally plug it in, they do whatever they do in the background, and then they come back a month later and say, "Here's all the stuff that we got into. Here's everything we breached and when we did it."
We did it when I first got there, and we did one four years later, after we had stood up all of our cyber program, one component being Darktrace. We plugged in the black box, and our SOC identified it within minutes and said, "Hey, do you guys have a pen test going?" "Yes, we do." "OK, good. We'll monitor and track everything for you." I pulled up my Darktrace interface, and Darktrace was able to isolate it and say, "Hey, there's your black box right there." At that point I could watch everything it was going to and getting into. I literally could track the entire pen test scenario as they were breaching things. Oh, they got into my SQL servers. Oh, they got into this. At one point, I think it was Monday at about 3:00 in the afternoon (we started on Friday, so they had the weekend to annoy us), they got a global admin account. At that point I was like, I'm just going to go home and drink bourbon for the night. And they took that account and everything went dark. I timestamped it and was taking screenshots and tracking this whole thing. Tuesday morning, first thing, I walk in and everything is blown out, because they went and updated all their scripting and all their attack scenarios, now with a global admin account. They're thinking they're going to get into everything.
What they didn't know was we had a honeypot environment all over our world. So that's what they kept attacking, and I'm watching all of this the whole time in Darktrace. When it was all said and done, we turned off the pen test, we unplugged the black box, and they sent it back. We did our after-action with them, and they came out with this Excel spreadsheet with literally tens of thousands of entries: "This is all the stuff that we breached." I said, "Can you filter out all of these?" I gave them a massive list of IPs and said, "Filter out all of this stuff." They were left with a little list, and they asked, "Well, what was that?" I said, "Those are all honeypots that you guys got stuck in." They said, "Seriously?" I said, "Nice."
The funny part was we'd had the honeypot in place for two years, and I literally had a recurring calendar reminder to go look at it. I'd look at it and think, "This thing is worthless. It doesn't do anything. It just sits here and nobody's doing anything. I've got all these decoys out there and it's doing nothing." Then we ran that pen test and all of those decoys just got hammered. We went from about 50 attacks in an hour to 5,000 in an hour. It was a really fun exercise to go through, and it was really cool to see what a tool like Darktrace will show you during an actual attack or an actual tested scenario like that. It was pretty amazing.
Matthew Connor: In that scenario, when it first plugged in, did you run it where you then shut it down? Where you isolate the black box and say, "Look, this is what would have happened"? Were you running this as a real experiment? We saw it, we identified it, we initially closed it off, boom, you're down and out. OK, now if this weren't in place, we let you go and see what you could find with this layer removed. Is that how the test went? Or was it, "No, you have to allow us to do our thing," so you have to open the door and invite the bad guy in and say, "What could you do in this environment if let in?"
Jason Meszaros: Yeah, we timestamped it and said we identified that it was there within 13 minutes. We could have shut it down and said, "These are the remediation steps we have taken." Instead we said, no, we'll let it go. Our SOC actually said they would also watch it and log everything the testers were doing. So the SOC had their own separate analysis of everything the pen testers got into, we had our own analysis, and then the pen testers had their analysis. At the end we combined all three and asked: where did they actually get to? Did they actually breach anything material? Getting a global admin account was eye opening for us, because it turned out to be an account that we thought had been turned off two years ago. We discovered it then.
Matthew Connor: That is an awesome story. Fantastic. Can you walk us through how you set up your honeypot and what you guys use? And two, they really didn't get anything outside of your honeypot? The rest of your production servers or your office environment, they missed that completely because they were so focused on it? What happened there?
Jason Meszaros: Their scripting goes after, and this is pretty typical, it will go out and look at the environment and say, "Are there Linux boxes out there that we can attack?" And they'll look for certain versions. "Are there Windows servers we can hit?" They skip the whole endpoint side of it, because the servers are where your crown jewels are anyway. Why would I go after somebody's endpoint? So they target the servers. All of our servers were hardened. At that point I think we were on the latest version of Windows Server, 2016 or 2020 maybe, whatever it was. And we had a bunch of 2008 and 2012 servers sitting out there. Those were all the honeypots. They just went, "Oh, look, there are dozens of these we can just go exploit." So they start going, and they just keep hammering them until they think they get through something, and they get stuck there.
The best part was, one year I had an intern come in who said, "I really want to learn security, I want to get into this." I had set up maybe five decoys in the environment, and I had him just go nuts in the honeypot. He literally created dozens and dozens of them. Once you get it going, they're easy to replicate and rebuild. So he had a field day coming up with all these different ones: this one will look like a finance server, this one will look like a database server. It was pretty cool.
Matthew Connor: And most of that was done by an intern. Awesome. Now, with the global admin account and the lessons learned, I can't help but think that as the environment grows, it's one thing when there are 20 people, but when you're talking thousands, it's an entirely different world. I think this is one of those places where AI really helps, to be able to quickly audit. Sure, you could do it yourself. You could write the script, but that's tough unless writing PowerShell scripts for Exchange is what you do all day, every day. Now it takes five seconds to create the prompt, and boom, you've got the script, it runs, and you have the audit. One of the things I really get excited about is having that level of ability natively in an organization without having to go buy yet another product. "Let's do an audit on our 365." You can build that yourself, and now you've got your full audit. It's just amazing how spectacular this stuff is. So for you on the security side, as a CISO, is that stuff you're playing with? Or do you still like going out and getting the product, or having a third party come in and run their assessment? Where should people be these days?
Jason Meszaros: Do it yourself, in my opinion. The assessment most companies come in and do is their foot in the door for sales, so they can say, "Here's the stuff we found wrong," and do the fear-based selling. I can go to them and say, "I already know that I have gaps A, B, and C. How do we solve that? I know your product solves A, B, and C, so let's talk about that." The reality is the big consulting companies are still going to say, "Well, we need to do our own assessment and confirm that A, B, and C are gaps," which at this point you're going to use an AI tool to do anyway. So you can trust me, or we can let you do it.
On this topic, I'm in a number of CISO groups. It's all the good guys trying to get together to prevent the bad guys. The Cyber Risk Alliance, the Gartner groups, and the Inspire Network are some that I belong to. There's a group called Collective Defense that is literally CISOs from all over the world sharing information, so when somebody's getting attacked, they let everybody else know and we know what's going on. We got hit in early July with a pretty standard phishing attack, and everybody else was saying, "Yep, we're getting hit with the same thing right now." It was ShinyHunters behind it. One of the guys in that group, someone I've known for a long time, literally used AI to build his own internal platform for tracking all his vulnerabilities and his compliance. He's got dashboards that show him all of his access reviews. He's just used it to build everything, and it's pretty slick. I'm like, this is the stuff that vendors should be [unclear].
Matthew Connor: Yeah. It's part of the reason why they're saying SaaS is dead or dying. It's interesting, because the moat, as they call it, the thing that protects the incumbent, was that it was hard to build that software and build that platform. It is not anymore. Now you can build this very elegant, hyper-personalized solution. Fantastic. The interesting thing about how that is going to affect these vendors and software manufacturers is that it's no longer about "I can give you a place to track your client relationships." Those are a dime a dozen, and we can build our own in a day or two and tie it directly into our system. I think the future survival of these SaaS providers is providing far more value than that. It's the data itself and how it connects. Not just your platform, because you can build your own platform. For those companies that are ambitious and want to experiment and build that internally, I don't think the SaaS providers provide any sort of competition there. The key point you made is that you're customizing it. You're building it for you. You're not building a generic solution where 80% of it works for 80% of the customers. This is a solution that's 100% for me, built around my world, my data, my environment, and my security.
Jason Meszaros: Yep.
Matthew Connor: And I think the pushback to that idea is, "Are you crazy? You can't vibe code an enterprise app and have enterprise-level security." Yeah, you can. You sure can. In fact, everybody is doing it, all of the manufacturers, and it is better than when you do it by hand. Sure, there are still some phenomenal programmers who are just at the top of their game. Let's say it's 5% of developers. They're awesome, ultra impressive. You probably don't have them. Most people don't. Most developers are lucky if they even come close to the kind of insight and skill that the leading AI platforms offer. The way it thinks about security now, and the way you can direct it to think about security, is night and day compared to a year or two ago. I just think that's the way to do it.
Jason Meszaros: Yeah. And you said there's that 5% at the top. Now you go to that next layer. Is that next layer 20% or 30%? Those guys are going to be as good as that 5% with AI layered in, and suddenly you've got 30% of the market, and the rest of them are going to be left behind.
Matthew Connor: Yeah, and that's a good thing, because the reality is the bulk of them are struggling. Before AI, if you ever coded anything before Google came along, how hard was that? You had the manual on syntax and that was it. You had to piece everything together yourself. It was tough going. Then Google comes out and you can suddenly search, "What's this error?" and get a lot of help. AI is just taking that to the next level. Even if it's just a tool that assists the decent developer, it really does take them to the next level. I think we will see, not a thinning of the herd, but most developers will either become really good at understanding and implementing the strategy of the business, being more of an architect than an actual coder, understanding what the needs are and directing it. That's really what we need much more of.
Jason Meszaros: Which is unfortunate, because my daughter, who is just graduating in computer science from the University of Maryland, loves coding. She's very good, very smart, and for her, it's annoying to have to use AI to code, because she loves coding. But unfortunately that's not the future. The future isn't sitting there doing the coding yourself. It's being able to understand the code that was written, understand what is needed, and direct it. I think that's going to be a real challenge for a lot of people, to move out of coding and into strategy.
Matthew Connor: Well, they're all going to have to go back and take some business classes so they understand the business side of the world.
Jason Meszaros: 100%. Because at the CISO level, that's your job. It's communicating the business needs from the C-suite down to the rest of IT and security. It's translating from tech speak into business speak.
Matthew Connor: How do the two mesh? It reminds me, did you ever see Office Space? "I'm a people person."
Jason Meszaros: Yeah, exactly.
Matthew Connor: I think that is the wave of the future. It's interesting, because so many people grew up in that millennial stage where there was a dependence on technology and less on people, and I think people thought it was going to go in that direction. But I think AI is making people skills much more important than they've ever been. They were always important, but for a short period in human history, I think we had this expectation that it was all going to be about your tech skills, that they would be your key to success. I think that was very short lived. It lasted 10 or 20 years, and I think that day is over.
Great. So when it comes to AI, and I love that you guys use it on the back end, I think this is where it gets really interesting for every business: how do we fully take advantage of this opportunity to have this level of intelligence and automation, and do that in a way that makes people's jobs better and not just getting rid of people, being like, "We don't need you"? How do you go about managing that balance?
Jason Meszaros: Great question. I'll answer it this way. I get hit up every single day with vendors saying, "We've got AI. We can change the world with it. We can do all these amazing things." Most of them come in and, if you hear their pitch, they pitch the basic back office. "We can automate your accounts payable, we can automate receiving your invoices, we can automate collecting your money for you." It's your basic finance, back office, and HR type of scenarios. It's not anything earth shattering. All they're doing is automating normal things. The ones that are separating themselves are the ones that come in and say, "We understand your business, we understand logistics and transportation, and we know all the pain points." Which, again, goes to our previous discussion: this is all about how you add business value and understand the business. Those are the vendors that come in and say, "We understand logistics, we can help you with these pain points, and we've already automated these for somebody else, or we know how to automate these." We're starting to get to the point where people have done it for somebody and now they're replicating it, and replicating is really easy with AI as well. But there are still a lot of people who just come in and say, "We can do the basics," and they're not going anywhere, because everybody's looking for somebody who understands their business and how they can leverage it to make themselves more money. At the end of the day, that's what it's all about: how do I increase my revenue or decrease my expenses? Those are the only two things that matter. And how do I do that with AI?
Matthew Connor: Well, and when it comes to those things, this goes back to our earlier discussion of the hyper-personalized, build-it-yourself model. Has there been anything a vendor has brought you where you thought, "That's awesome, and I wouldn't even try to build that internally. Let's do that"?
Jason Meszaros: I haven't seen that yet. We did have a vendor come in a couple of weeks ago, and they used a different model. Rather than coming in and saying, "We can rebuild stuff, or we can build this automation for you," they said, "We'll build you a new application or a platform. We'll do it in six to eight weeks, it's a fixed price, and you keep the IP. It's yours when we walk away." I think they're doing it for exactly what I just mentioned. They need to do it for somebody so they can then replicate it. If they've done it, they know they've proved it out and they can go replicate it. Their model is that they go in and build something, charge a flat fee, and then add it to their arsenal of other things they can build. When they walked through the demo of their platform, they'd already built a whole lot of stuff for a whole lot of companies. It's a different model from "we're going to get you on a recurring subscription fee." They give you a flat fee, build you what you need, and walk away, and you keep the intellectual property. Completely different model. Everybody's subscription based today because it's the easy way to have recurring revenue.
Matthew Connor: And who doesn't love recurring revenue, especially if you're a venture capitalist or private equity? That's what's been driving the industry for a couple of decades now, right?
Jason Meszaros: Without a doubt.
Matthew Connor: I love that model. Was that something you guys bid on? Were you like, "Yeah, I like that, let's go ahead and do it"? Was it reasonable? Could you build it yourselves, but there was enough extra value in this that you wouldn't even consider a SaaS solution or doing it yourself, and this makes perfect sense?
Jason Meszaros: We haven't pulled the trigger on it yet, but we've identified probably half a dozen projects that we could use that for. We're in our budget process right now, and one of the projects we scoped out is, let's try these guys out and see what they can do for us.
Matthew Connor: I understand if you don't want to drop a name right now, but it would be interesting to do a follow-up episode afterwards and see how it goes, because I think this is the future of software. I think the SaaS model is on its way out, with the exception of a few. As a general rule, I think it is going to change drastically because of this level of capability and how democratized it is. I think security is one of those areas where I'm not sure there's a benefit in building your own. Sure, build the tools yourself and leverage your own internal AI, but building another Darktrace like you mentioned, why would you do that? The risk you run is that the only throat to choke is yours when something goes wrong. Instead you can say, "We have this best-of-breed, industry-leading product. We've implemented it." That makes sense, because now there's a third party that is specialized in that as an industry leader. You can take that to the board. I think it's really hard to take it to the board and say, "We're going to build our own." You could, but you're laying it all on the line, and for what? To save a few dollars? I don't think the cost is there. But I'm curious if you agree: I think on the security side, the SaaS model will continue to work for the EDRs, the SOC, the SIEM, the Darktraces, all of these. It makes sense, right?
Jason Meszaros: So my buddy that built his own platform just started pulling the info and data from all of those. He's a Microsoft shop, so he's got all of his stuff in Sentinel, and he's pulling the dashboards in from there. He's got his EDR data that he's pulling in. He's pulling all of that into the framework, the platform that he built, to consolidate everything into the proverbial single pane of glass. Everybody loves that. But you can build your own single pane yourself, without having somebody else try to do it, and you get there way faster now.
Matthew Connor: Totally agree. And OK, I've never said this before, and not tooting our own horns or anything, but it fits so beautifully into this conversation: at CyberLynx we did the same thing internally for all of our managed clients. We now bring in all of the access data and all of the alerts, from Microsoft to audit everything for the 365 tenant, and from SentinelOne, tying right into there. The cool part is that you get all these alerts and this data, which historically you would send to a SOC to go through and monitor. Guess what's really great? Your own AI SOC. You've tied everything together, you've basically created it, and now you get to build that out even further and keep tweaking. That is so fantastic. And the speed is unprecedented. It is as fast as you can get. Within seconds you've got your AI analysis of the data and the alert exactly how you want it, and it's just spectacular. You talk about the single pane of glass. Wow. I feel bad for the SOC. It's the same deal, right? It's right on the cusp of, "Well, I like it as a backup SOC." Internally you build your own single pane of glass where you can now have your own alerting, but the problem is you're getting these alerts so fast and so specialized that you look at your SOC and say, "Guys, seriously, what took you so long? Yeah, we knew about that. We knew about this, and we already took action."
But going back to that other question of when you build your own: I like it as a supplement right now. You've got your own single pane of glass, you've got your own alerting, you're ingesting all of this super inexpensively and super fast. But at some point, when do you feel so comfortable that you say, "I think we can let the SOC go"?
Jason Meszaros: Yeah, don't tell them that.
Matthew Connor: I think I just did. I've been telling them, though. You've got to be doing the same thing, because if you're not using AI and revamping the entire SOC to be an AI-first SOC, where it's adjusting fast, responding fast, and doing a much better job, what are you even doing? You're not going to be in business long. That said, I like the personal approach of a SOC that calls and says, "Hey, cool." That's great, because we're much the same way with the human aspect of business. It's really important to have people interacting with people. I don't want to sit down at a restaurant and have a robot waiter. It's not exciting to me. Maybe one time it's, "Oh, that's very cool, the robot could serve food," but I'm certainly not going back all the time. There's something weird about it, and there's something very nice about interacting with humans. But I think, just like every other organization, it has to be AI-first on the back end with that human touch. Otherwise, what are we even doing in this October of 2026? I don't know, is it just me?
Jason Meszaros: It is not. So if we take this back to transportation and look at how we would leverage AI, our legacy phone system has the prompts: press one for HR, press two for finance, press three for whatever. Now you can have drivers call in and say, "This is Joe, and I'm looking for my driver manager, Jeff," and AI will say, "Jeff's not available, but can I help you?" He says, "Well, I'm just looking for my next load number." "Oh, your next load number is 12340." "Where do I pick that up?" "Here's the address. I'll send it to your driver app so you have that." It does that kind of simple info lookup. Our driver managers are dealing with dozens of drivers all at once. They can't answer the phone every single time to give Joe his new load number. So how do you speed that up? How do you make Joe's life much faster and better, and take a huge chunk of workload away from Jeff?
Matthew Connor: 100%.
Jason Meszaros: And what's crazy is how easy and fast that is to build. That was one where I was blown away by how powerful and easy it is to set up now, with the API calls and everything. It was built with simple prompts, and it's just ridiculous.
Matthew Connor: And you don't even have to have, I think it was Jeff. When he calls in, they can immediately look up Jeff's number and say, "Hey, Jeff, are you looking for Joe? He's tied up right now. What can we do for you?" Super cool, and boom, knock it out. It's amazing, and it's so fast and easy to set up and so user friendly. I think those are really great examples of leveraging technology not just because we can and it's super cool. The phone tree was always really annoying. It was always so bad. It slowed you down and didn't help you get where you wanted to go. It wasn't providing you with any information, unless it was operating hours, which you could have gotten faster with a quick Google search. But now, wow. Whether it's an internal or external customer, the customer experience is just there. How do you beat that?
Jason Meszaros: For us, everything is about the driver, because our drivers are the ones who make us money. If their wheels aren't turning, we're not making money. If they don't have loads, if they're not moving, we're not making money. We take care of our drivers. That's one of our big things. I equate it to my previous life in the sports industry. We had nine players on the field. They were the product. Here, our product is our drivers. If they're not able to do what they do, it's not going to help us. In fact, we have huge initiatives within the company around how we enhance that and how we make their experience better. This is one of them. AI is going to make a huge difference for us.
Matthew Connor: That's great. Have you already rolled out some of the phone stuff, or is that on the road map?
Jason Meszaros: It's on the road map right now.
Matthew Connor: It's so exciting. I can't wait to hear how that one goes, because that's such a great example. I'm not the hugest fan of Amazon in some regards, but in other regards, what they've proven about being hyper-focused on the customer experience and making it as smooth and effortless as possible is the right focus. That should be everybody's driving principle moving forward. It's going to be really cool to hear how that goes for you. I'd love to have you back on the show once you've gotten that down, to go over what you did and how it played out. It's shocking how easy and fast it is to set up and to tweak, thanks to AI, unlike anything in the past, when anything involving voice recognition was painful. I could go on and on. I'm looking forward to hearing how that goes for you guys.
Jason, this has been an absolute blast. I can't thank you enough for coming on. Before we go, can you tell everybody where they can find out more about you and more about Anderson Trucking?
Jason Meszaros: Anderson Trucking is at atsinc.com. Myself, I can be found on LinkedIn, so pretty easy to find.
Matthew Connor: I love it. Jason, thanks again, and until next time.
Jason Meszaros: All right. Thanks, Matt.