Steve Goudreau is the Director of IT at Ice Industries, a metal stamper and roll former with seven locations across Mexico, Louisiana, Mississippi, and Ohio. Ice Industries works with heavy machine companies including Volvo and Cummins and makes the solar backing rails for First Solar. Steve has spent 20 of his 35 years in IT in manufacturing, and he now has a team of five at an organization of about 700 people. Steve and Matthew dig into bringing OT and IT together, choosing tools as a small team, and putting AI to work in manufacturing. 

Here’s a glimpse of what you’ll learn: 

• How to give controls teams the remote access they want while keeping OT secure and patched
• Why controls guys love repetition and how IT can still offer them something extra
• Why change management and patch management go hand in hand at the OT layer
• How a small IT team can pick tools when everyone has something new to sell
• How awareness training helped an employee question a fake IT help desk call on Teams
• Why trust but verify applies to AI work in design and quality systems
• How a unified AI strategy and prompt skills keep licenses and ideas from multiplying
  
  
  

In this episode…

Sponsor for 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.

Check out previous episodes:

Smart People Know You Can’t Just Let AI Do Everything with Brian Walters - Ep 249

Why Fear of AI Hands the Bad Guys an Edge with Jason Stockinger - Ep 248

Who Gets the Benefit of AI? Enhancing the Core Experience with Neil Goodrich - Ep 247 

Transcript: 

Cyber Business Podcast

Steve Goudreau

Director of IT

Ice Industries

Matthew Connor: Matthew Connor here, host of the Cyber Business Podcast. Today we're joined by Steve Goudreau, Director of IT at Ice Industries. Steve, welcome to the show.

Steve Goudreau: Thanks for having me.

Matthew Connor: Thanks for coming on. Before we get too far in, a quick word from our sponsors.

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. Steve, for those who aren't familiar, can you tell us about Ice Industries and your role there as Director of IT?

Steve Goudreau: Sure. Ice Industries is a metal stamper and roll former. We predominantly work with a lot of heavy machine companies, Volvo, Cummins, and we make all the solar backing rails for First Solar. We have a very diversified manufacturing base, and I'm the head of IT for the organization. I've been with the company just under two years.

Matthew Connor: Nice. It always strikes me how challenging IT is in manufacturing facilities. Not to put down office work, but an office is so much easier when it is literally just a computer and some software. In manufacturing, it's so much more. Can you go through the challenges you have day to day in manufacturing IT versus the office worker?

Steve Goudreau: Yeah, it definitely presents some unique challenges. I've been in manufacturing twenty out of my thirty-five years in IT, and your biggest challenges are dealing with the OT side. All your controls guys want access to their machines remotely. They want their integrators to be able to get in to make changes. How do you allow remote access while making sure it's secure and everything's patched on that side of the network? When I worked at Midway Products out of Monroe, Michigan, right before this, we had a very large project that I led to integrate the OT side into the IT side. That was bringing together the manufacturing side and the IT side in a way they hadn't done before. It was to help sell it, like, you guys have to be on the network. We understand that part, but here are the criteria we want to see if you're on the network. On the IT side we have our security, so what kind of security can you have on the OT side? There are dozens of vendors, and we talked to dozens of vendors. We talked to Dragos and Honeywell and everybody like that to see how they could fit in our very unique environment. The one thing I found out through all these years is that controls guys love repetition. They love having their little network cells that everything's connected to, all using the same IP addresses behind a NAT box. So how do you give them extra functionality? When we were looking at Dragos, it had the ability to look inside, past the PLC and everything, and say, okay, the firmware version on this PLC, or on this robot, or on this actuator, is X number of versions behind, and here are the exploits for it. If you allow somebody remote access, now you're bridging networks. So it was selling it as asset inventory and security at the same time. And then, what other value can IT add at that point? We were looking at, well, how can we help you back up your PLCs? How can we know what recipes you have on your HMIs? How can IT help facilitate this, not necessarily own it?

Matthew Connor: It's funny, because that ties into so many modern problems. A lot of the problems with OT, as you described, are that it's really hard to secure. You mentioned the firmware, and when you look at the patch management challenge of any organization now, whether all the software's patched, there are a lot of tools to help you do that, but patching in and of itself is almost a full-time job for any size organization. And the really interesting part is that now, in the day of AI, with a zero-day exploit from some kid in Lithuania with a prompt, it's not like patching fifteen or twenty years ago, where if everything was fully patched, you were rock solid, with the exception of maybe a handful of elite hackers in the world who might have a way in. You were pretty solid, for the most part, assuming you were doing proper patching. Today that's not the case. I'm curious about your take, because I know Microsoft is really challenged with the OT side, and they're partnering with companies like Darktrace to use AI to see what's going on on the security side. That makes a lot of sense, because you're using machine learning to see the whole network and get a view of things. And I wonder, as we move through this brave new world of AI, where every day new stuff is coming up and it gets faster and faster, is that the future, where we have to be using AI to see what's going on on the networks in order to defend? With OT, that's kind of the only way, otherwise you've got all these open problems.

Steve Goudreau: Well, on the OT side and the IT side, I can say, hey, look, we have to take this computer down to patch it, and if it doesn't come back up, well, we'll replace the computer. If you brick a PLC or something on there, you're affecting the cash register of the company. So change management and patch management go really hand in hand at that layer. And then there are all the wonderful things AI can do. Especially on the OT side, well, on cybersecurity regardless, but on the OT side, everybody's got a new tool for you. How do you pick, if you're in a smaller organization like we are? How do you find what's best of breed? You can only read so many articles, watch so many podcasts, sit through so many uninvited webinars that just show up in your Outlook, talk to your VAR a thousand times, and try to get the right people involved. One of my favorite phrases I've heard is FOMO, fear of missing out. Like, you don't want to be left behind, because if Ice Industries isn't using AI, our direct competitor is, and they're making their processes or their product better by using it. And on the security side, if you're not using a product that's using machine learning or AI, well, hackers are, every single day. So you have to stay one step ahead, or at least on par with them, or at least understand the challenges you're going to face.

Matthew Connor: Yeah, I think that's the thing. At least on the security side, it is kind of like bringing a knife to a gunfight. The knife was great up until the advent of the gun, and then it's like, well, they've got guns now, so we should probably get guns too. I think that's a fair analogy, that you really do have to keep up with them, at least on par. That's the whole game, otherwise what do you do? You can't just rely on disaster recovery all the time, constantly getting hit, because you'll be knocked out of business eventually. You're just always down, and then, like you said, affecting the cash register of the business. You raise a really good point, how does the small business go about choosing, in a world where there are hundreds of new AI startups every day? There may be plenty that apply to what you want and need, but there's only so much time in the day. How do you do your full-time day job, which for most people consumes their entire day and then some, and then, oh, by the way, you need to hurry up and secure your organization with AI tools so you can be safe? Cool, who needs sleep anyway? It's overrated. I guess that's where things like the Gartner Magic Quadrant help, because you can at least say, okay, it's industry leading. It's that "nobody ever got fired for hiring Cisco" thing. At least it's filtering out a bunch of them. They're doing the job of reviewing a bunch, and it gives you a place to start, so you can limit it down to examining three or five products versus five thousand. Is that the approach you take to whittle through them?

Steve Goudreau: It's funny you mentioned Gartner, because I've probably been accused of being too reliant on looking at that for any number of products. Even when I worked in large organizations, when I was at Midway, I had twenty to twenty-five people under me. Now I've got five, but the scale and scope of what you're dealing with, all the threats still exist. And yeah, you can look at the Gartner. But anecdotally speaking, when I was at Midway, we were looking at bringing on an MDR. The CrowdStrike people were so obnoxious that it made it really hard for us to get through some of their presentations, and at the end we got down between Arctic Wolf and Rapid7, and it just became a numbers game at that point. When you start with nothing, it's hard to say, well, we're going to go right to Rapid7, which covers most of your stuff. Plus you want user awareness and you want vulnerability management, and Arctic Wolf had that entire package we were looking for. And then you still have to maintain a product once it's in. It's like, oh good, we've got something now, who's maintaining all these alerts, and who's updating the Arctic Wolf clients? There were [unclear] clients behind it. And then all the other integrations you have to do with whatever MDR security solution you have. It has to integrate with your Salesforce, Office 365, whatever your HRIS is, Active Directory. There's just a million points of contact you have to deal with. And of course, everything's got an upgrade that affects it in the middle. But I went with Arctic Wolf here from the standpoint of being put between a rock and a hard place right after I started. We had an IT director who'd worked here for twenty-two years, and he left a lot of technical debt behind. Our entire goal with him was that he was supposed to sunset himself and move our ERP from on-prem into the cloud. I was brought in about six weeks before he retired to learn, pick up the slack, and make recommendations. Well, due to extreme negligence and bad network hygiene, we got ransomwared on Groundhog Day of last year, at six-thirty in the morning on a Saturday, after I got back from my first trip down to our plants in Louisiana and Mississippi. I get a six-thirty call from my boss, the president, saying, hey, I think we've been hacked. And I'm like, okay, we probably haven't been hacked. And I get on the phone call, and yeah, we were. They ran through our backups and deleted a lot of backups. We were halfway through our Glacier storage, deleting all that. So I had a decision to make. We needed something, I had experience with this product, and it wasn't the time to look at what else I could possibly do. We had to right the ship as quickly as possible. In an ideal world, and I've talked to my staff about this, if I had to do it all over again and the ransomware hadn't happened, we probably would have done a better use case for everything we had. We might have still ended with Arctic Wolf, we could have ended with Defendify, we could have ended with any number of groups. Probably not CrowdStrike. I hate to say it, I have nothing against CrowdStrike, but I had a buddy, also a director of IT, shout out to Jason Lambert, and they used CrowdStrike. And when CrowdStrike took a crap, yeah, they took a crap too. So I'm not vindicated by not choosing them. If the salespeople I dealt with had been slightly less obnoxious, it would have been a different story. And Darktrace was very aggressive. They wanted to ship us the box as quick as possible, and I was like, well, let's slow down here. I've got to get my feet underneath me. I've got to sell my team, and I've got to sell management. Half of what you do as an IT director is a sales job. Most of it is selling your ideas and things to the business, but you're also selling it to your own staff, to make sure this is a product they like, something they're interested in working on. I found that most people in IT want to work with new technology. Crazy. Go figure.

Matthew Connor: So you were doing this mid-ransom, right? You're getting hit, and you're now bringing in a SOC, SIEM, whatever you want to call it, a security solution. Wow. Okay, well, that plus a new backup provider and the new EDR, because they tore right through the EDR.

Steve Goudreau: We had the old Malwarebytes for the enterprise.

Matthew Connor: Yeah, right.

Steve Goudreau: But listen, we had a couple of Server 2003 machines in production, twenty-two-year-old servers.

Matthew Connor: Yeah.

Steve Goudreau: And one of them, when we shut everything off, did not come back up. That PERC controller on it said, I don't have any container information. So, things that had been left up for four or five years.

Matthew Connor: Yeah.

Steve Goudreau: And plus, right after that, we were at end of contract with our SD-WAN and telephony provider, and then we had to finish the migration of our on-prem ERP system to the cloud variant. So last year was a very, very busy year.

Matthew Connor: Yeah, so, trial by fire as soon as you get in, huh?

Steve Goudreau: Well, the good news is the organization stood behind IT and said, tell us what you want and you've got it. And that's what we did. We turned everything around. We also replaced every single endpoint and server in the entire organization, replaced most of the switches, migrated from one SD-WAN provider to another, everything. And we instituted a real user awareness training through Arctic Wolf. It's funny, it's clever, people seem to at least enjoy it. And then you start the hard work of saying, hey, I know you guys got used to just getting text messages for this, but now you're going to have to use the authenticator app. And yeah, I know you like using Google, and we hope you're using Microsoft. And yeah, I know you like using this. We're going to do this. The bigger battle is always, as you probably know, working with all that shadow IT stuff that IT doesn't have a hand in. Like our HRIS system. It's not SSO or anything like that, so HR just kind of runs that. They can terminate our user there, but it doesn't talk back to Entra or anything to terminate anything, or vice versa. So there's still all of that. We've done all the foundation, it's solid, we've been working on it, and I've got an outstanding staff. It's just a mild amount of PTSD from how last year kind of went down.

Matthew Connor: Oh yeah, no, for sure. At least it wasn't... wait, were you already awake when you got the call, or did you wake up to the call?

Steve Goudreau: I woke up to the call.

Matthew Connor: Okay. So there's still that. I'm sure there are still times where, if your phone rings in the morning too early, it's going to trigger some stress responses, for sure.

Steve Goudreau: Yeah, I got in Friday night at like seven-thirty and got to bed at midnight, and then at six-thirty in the morning I'm getting a call from the president. But again, the organization circled the wagons. I had formed an IT steering committee of our CFO, our director of financial ops, and the president, to say, okay, this is what you guys should know every day, and this is how you can steer where we're going as an organization. So we'd already set these things up. When we jumped on the call, it was like, okay, ERP is down, how do we recreate these processes manually so we can still capture the data? It was all hands on. The business kept rolling. It was just a matter of bringing all the data back and rebuilding everything kind of from scratch. And as you start peeling back that onion, we're trying to bring Server 2025 domain controllers into a 2008 R2 environment, and they had started the upgrade process to 2016 and just stopped. So there was a whole bunch of stuff in the schema that we had to remove. There were a lot of two and three o'clock in the mornings righting the ship. But the good news is that when you have the tools in place, as I was leaving work today, our Arctic Wolf gave us a call saying, hey, there's a one-on-one phone call going on right now with a threat actor with one of your users, because we have Teams as our phone system. Somebody had set up a tenant and set up a user called "IT help desk," and it called somebody on Teams, and it just said "IT help desk," and then they said, oh, this is Liam from IT, I need you to download the software, you're having an issue, we need to get on your computer right now. But because of the user awareness training, they were like, whoa, urgency never comes like this, you're not getting random calls, and I'm in the same building and there's no person named Liam. We're an organization of about seven hundred people. If we were seven thousand people, they might not know everybody in IT by name, but they also would know that this is totally outside the practices, that you can't just randomly see somebody calling somebody in accounting saying, hey, I've got to get on your computer real quick.

Matthew Connor: Well, and wait, so Arctic Wolf contacted you and said that call was ongoing. How did they know it was in progress? That's fast.

Steve Goudreau: Because, well, they have hooked into our Office 365 tenants. However they identified the threat actor, they saw the VoIP call and said, hey, probably from the caller ID.

Matthew Connor: Yeah, that's smart. You see that it's coming up as IT when it's clearly outside. Smart. That's beautiful. But there you go, that is a wonderful testimonial for Arctic Wolf. That's fantastic. And the interesting part is that would work so well for so many modern businesses, where everybody's all over the place working remotely. Whatever person in accounting is at home, they've never seen, they don't know all the IT people, and maybe they've got an MSP they're working with and they don't know all the people there. So that's really well done. I like it. That is a great call out for Arctic Wolf on that one. And you guys have been ransomware-free ever since. I mean, sure, you ripped everything out and modernized everything, so that certainly helps, but ever since, it's been smooth sailing, right?

Steve Goudreau: Yeah, for the most part. When we were looking at EDRs, I had experience with SentinelOne before, so I'm like, look, I'm not going to mess around, we're just going to go right with SentinelOne. Of course, you have to sit through the upsell speech. I'm like, look, I just want the EDR, I don't need everything else. And my VAR, I've had the same account rep, shout out to Madeline Hara at CDW. She was my account rep when I was at Midway. I just told her, this is what we're going through, this is what I want, I don't want a sales pitch. I mean, you might get a slide or two, but I don't need to be convinced to an upsell, not at this time. And our local MSP that we worked with, we have two really good ones, one in Sonnet, my guy Klein, I said, look, this is what I'm looking for, this is the number of licenses I want, I don't need to jump on a call, this is exactly what I want. And they just took care of it. And those kinds of relationships, because these couple of companies have worked with Ice for a long period of time, they knew exactly, hey, if I say this is what we want, I'm not going to sit through a death by PowerPoint to try to upsell me. I've sat through dozens of them since then, from DocuSign [unclear]. I'm like, look, I just need five licenses for this. Okay, well, let's jump on a call real quick. And I'm like, oh, that's a lot of slides.

Matthew Connor: Yeah. To your point, though, having the right partners is really important. Nobody is an island, nobody succeeds alone. Nobody knows everything, nobody's got access to everything. You need to have the right partners, and the ones who won't kill you with a thousand slides just to get five licenses. That's a funny story. I'm curious, you mentioned earlier implementing AI to keep up with the competition. But I think, interestingly, in manufacturing, when it comes to physical goods and services, we're in this really interesting transition period, where the manufacturing process is highly automated, with all the machines that take care of most of the work. And correct me if I'm wrong, but it seems like one of those places where there's not a ton of room for completely revamping the system and making it fully AI, because most of the work is highly automated anyway, and there is a manual labor portion to it. Is this one of those things where, at a metal shop, AI is not going to really fully transform the place until it's all robots in there or something?

Steve Goudreau: Well, I would say some of the tertiary hands that go into it, like your quality systems, things that are looking for defects, AI is going to have a big impact on that. Your quoting, your CAD/CAM designs, things of that nature, where you're still leveraging your own internal creativity, but some of the monotonous work, like instead of having to draw three holes here at 2.2 mils apart, I could just say, hey, do this for me, and then verify the output. We had a couple of meetings today on AI, ironically enough, at work. One of the things is the old Ronald Reagan saying, trust, but verify. It's nice that you can put all your information in there, and then you're like, oh, this looks good. And I'm like, but did you verify the numbers looked right? Did you verify that? Oh, it called the person Mark instead of Matthew. Oh wow, right. Because AI can hallucinate. And we're in the middle of our own AI journey at Ice right now. Because we don't know what we don't know, we're working with a company called Vision Ventures, who has experts. Basically they've interviewed dozens of people in the company, asking, where are your pain points? And they delivered to us a wordy but very long book and a spreadsheet of, here are your high-ROI ones, here are the ones that'll take six to eight months to do. They're not pointing out tools necessarily to do them. They're just saying, here's where AI in any capacity can do A, B, and C, and which are the ones that have the biggest ROI. Your sales, your quoting. Because every product that we touch, that you touch every day, has something AI built into it. I mean, you have a note taker. We have Read AI as our note taker. We have Teams Premium as a note taker. Adobe has it built in. Office 365, whether you want it or not, keeps popping up with Copilot all the time. And then people start asking, well, can I get a Copilot license, because they see Copilot. I'm like, well, you have Copilot Basic with our premium SKU. A full Copilot license is expensive, and until we determine as an organization, are we going with Claude, are we going with ChatGPT, are we going with Copilot, what's going to be our preferred or recommended agent of choice, or maybe you say we went Claude Fable for all the Cowork stuff. Maybe for everything that's already tied into Copilot, like Read AI uses Copilot for transcription, maybe we're going to go that route for this. Does it make sense to grow it organically, or do we just hyper-focus on certain tools? These are the questions we're pondering and asking and looking at all the documentation. I'm sure your inbox is much like mine, and especially LinkedIn gets blown up with, hey, if I can get fifteen minutes of your time. I'm like, well, if I had fifteen minutes, when I'm still trying to figure out exactly what the organization is prioritizing. Look, here's everything you have a possibility on. Everything's possible with AI, right? Here's a possibility on everything that has a really good ROI. Is this what we focus on, or is the company looking more long term, like, we want to have our preventive maintenance schedules predicted by AI, based upon the bearing counts and spins and actuators and thermal temperature readings? That's going to have a bigger impact but lower ROI, but less downtime, happier employees, happier customers. Or is it a multi-pronged approach? That's kind of where we're in the middle of right now, because everybody's got an idea, everybody's got a tool, but we're not Cinderella. Not every shoe fits.

Matthew Connor: Yeah, no. And the interesting part is that now with AI, you can be your own cobbler and make your own shoes. Unlike years past, if you want to create your own SOC product where you get alerted anytime there's, let's say, impossible travel, or a million different things, it takes you all of a few hours to put it all together and have a pretty decent product that you can now run internally. So the possibilities are endless, but the time is quite finite. And so, to your point, where do you even start? I think everybody's on that crazy journey right now of, what direction do we want to take? Best bang for our buck? Which products are we going to go with? How are we going to get there? And I think historically organizations had a lot of time to make these decisions. It was a much slower-moving vessel than it is now, where everything is moving so rapidly that, as you said before, it's also a matter of keeping up with the competition. Because if they are moving twice as fast and they are twice as efficient and twice as profitable, then they're going to be less expensive, and you're going to eventually have a hard time competing. So I think it's a real challenge that everybody faces right now. As IT leaders, it's dealing with that stress daily and helping to guide the organization down the right path. That really becomes the challenge.

Steve Goudreau: Yeah. And people I've worked with in the past who are currently leading IT departments, we talk frequently about, well, what are you doing? And I'm like, oh, that sounds really cool, I wish we could do that. Or maybe you're the person doing the cool thing that they wish they could do. I think everybody likes the AI note takers. They're super cool. You get tasks, transcriptions, really cool. Then you start asking people, well, who reads these? You got the recap at the end of the meeting. Did you read those, or were you just looking for your piece? I'm like, yeah, it's doing cool stuff, are we taking advantage of it? That's the funny part. And the only way to really fully maximize that sort of thing is with AI. For instance, with all of my AI note takers, I have eight Claude agents that go through and check my email, and I know they don't get to write anything, there's no writing email, but they get to read all of my email, read all of the notes, and tell me, take a look at my tasks, what did I miss? At the end of the day, did I even read all my email, the hundreds of emails that came in? It's a problem that I have. I will promise something in a meeting, be like, yo, we'll get to talk and we can do that, absolutely. And then you go to the next meeting and I've completely forgotten what I said, what I promised to do in that meeting, and by the end of the day, or a week later, it's like, oh, I totally forgot I said that. That one, I think, has been the greatest savior, the note taker in conjunction with, I use Claude, it doesn't really matter, in conjunction with Claude keeping me honest and keeping me tracking on that stuff. It's like, just check my email at the end of the day, did I reply to everything? I don't even have to ask it, it's set to check. And what do I have on my list, instead of having to manually take care of a task list, because I'm not going to write it down. I've got all the capability in the world, all the tools to be able to write it down, I could even do pencil and paper. I don't write it down. For the life of me. And my wife is like, what is wrong with you? She loves her list, and you check it off, and you get a little dopamine hit every time you cross it off. I'm like, I don't know, it seems like a pain to write it down and then cross it off. I don't get excited about that. But now when Claude does it for me, that's a whole other story. Now I'm getting my dopamine hit. Now I'm like, yeah, that's great, my day went well. Oh, I did forget about that. So I think that's the only way to do it, a long personal story on it.

Matthew Connor: I think if you only use part of the AI tool, you make a great point, because for the longest time I think everybody was running around with that, and I think most people still are. It's like, look, it's so cool, it took all the meeting minutes and it's got all my tasks and everything. Yeah, but you didn't have time to read it. You're not reading it, you're not tracking on it, so what good was it? Until you funnel it into an AI agent that can then actually do something you've directed it to do. Otherwise you just build up this encyclopedia of all my meeting minutes, and who cares? Nobody's going to read those. You don't even read them. So yeah, I'm right there with you.

Steve Goudreau: It becomes really challenging, because you implement that for you, but then it's like, well, that would be great for the boss, for a whole bunch of people. When you're talking office workers, it's a godsend, especially if people have meetings. And then rolling that out on top of all the other things you had to do, and then maintaining that and making sure it's working. So it's an interesting time we live in.

Matthew Connor: It is. And then convincing the organization to pay for it, because that adds up, right? And I guess that's the real question, it ultimately comes down to ROI. What's funny is that initially there was a lot of, I'm going to say heat, about this in the first few years of AI. Since ChatGPT came out, there was a big push, everybody's like, oh, that's really cool, let's get on it. And then they started checking out, like, it's not really a good assistant. It doesn't write my emails. Well, it does, but it kind of gets in the way. It hallucinates. It makes me look bad whenever it sends some link that doesn't even exist, and now I look like an idiot. So I think a lot of people got a bad impression of it early on, and we're not used to the speed of development. If Jimmy's hopeless today, six months from now we assume Jimmy's still hopeless. We're not going to keep checking every six months, like, is Jimmy smart now? No, Jimmy's still hopeless. But the AI is advancing so rapidly that six months ago AI was not half as powerful as it is today, and a year ago, even less. Now it's so much more advanced, it's got these great guardrails, it hallucinates so much less. It's still not perfect, but now we're at this really interesting inflection point where you can rapidly do stuff and implement. Like the note taker, that's a fifteen-minute project, and boom, it's up and going with a couple of prompts. And awesome, now it's running and you're golden. So getting back to ROI, I think there was a lot of heat about ROI, and it's like, yeah, you're going to spend a hundred bucks per employee for this, and are they really going to use it? If you do that over seven hundred employees, that's some solid money every month. The company's making real money, but that's still real money you're shelling out every month. And are you getting the ROI on that? I think it was really challenging until now. And I would be surprised if, with a little bit of direction, people couldn't quickly build agents and solutions and leverage, I'm just going to keep saying Claude, leverage Claude to be so much more productive than they could have been six months or a year ago. And I think the challenge is everybody's so busy that people aren't going back and revisiting and being like, oh, should we do it now? You're going to check every six months, is now a good time? I think it's a really interesting thing because of the psychology of it, people don't think like that. They're still like, no, Jimmy's still hopeless. And I think that's a problem for adoption.

Steve Goudreau: Where I've lucked out, where I'm at right now, is we've got some power users, like our director of financial operations, Bennett, a super sharp guy who loves playing around with new stuff. He led, before I started and still is doing it, our MBS implementation, and he's gotten into Claude. That's the meeting we had today. It was going over Claude. We did an enterprise license, we set everything up today. We brought in our director of business development and sales guy, and we went through what Cowork is, when to use this, when to use that. It's basically, you guys are smart, here's a new tool, here's how I'm using it, here's how I'm going to share information back and forth, now you use it. If you have a problem, ask me, but if you find something, then you tell me. It's that kind of collaboration. We're in the middle of an ERP enhancement project, and we have one of our head shipping guys at our Cincinnati location, Jamie Cortez, who knew everything about our ERP system. Just amazing. It's like, well, I didn't know that, because I typically don't interface with him. And he sat in this meeting, and somebody's like, well, I would like to know how to do that, and he's like, well, here's how you do it, and they're like, oh, well, that's cool. So now you've got this homegrown subject matter expert who's kind of thirsty to learn new things. And now, because of this, you can say, hey, let's ask Jamie, what does Jamie know about this? It's a little unorthodox, but it's kind of what you want in an organization. Our CEO, the owner, has been playing around a lot with Claude, and he's like, just the stuff that it can do. He knows what it can do for his personal side. He goes, it's got to do something for the business side. And we're like, well, we're looking at it. I guess we want to make sure that we're not putting the brakes on everything, but making sure all the security and everything that we can control is controlled in a manner that's safe and effective and cost effective for the organization. That can be its own project in itself. Everybody wants the latest new tool, and before you know it, you're looking at corporate credit cards going, why is this person paying for an OpenAI license? This one's got a Copilot. This guy's paying for their own Claude license. It's like, wow, we need to wrangle everything in and come up with a unified strategy. And if somebody has expertise in prompt engineering, let's see what they're doing, because prompts will work mostly universally across all the platforms. But getting people to understand just how powerful prompt engineering is, when you're asking a question or saying, explain this like you're a five-year-old or something like that, it's really powerful stuff. And I always hit people when I talk about prompt engineering. It's like doing Boolean searches in Google. Most people still just go, hey, can cats eat waffles? And you'll get, yes, cats can eat waffles, and then you'll get some. But if you know when to put plus, equals, this in parentheses and everything else like that, it totally changes the way you can search. But most people are like, well, I didn't know you could do that. I'm like, well, do that on steroids. And that's your ChatGPT, your Claude, your Gemini, you name it.

Matthew Connor: I think the really cool thing is, since you guys are just now implementing Claude, and you mentioned your ERP, this is where I think you're going to have a lot of fun. Prior to Claude, connecting to your ERP to get data and do anything, forget about it. You needed a full-on developer, somebody with serious skills, and there was no way you were going to pay that kind of money to have that in-house. And then directing those very expensive resources to build a particular product or run some project became a really big thing, so maybe they put out a couple of things a year. Oh cool, we could build this dashboard. Great, maybe. And you're spending a lot of money for that. Now, holy crap, just wait. When you connect Claude to your ERP, the CEO can just go ahead and talk, boom, what do you got for me, and get everything they want from the ERP. Because the ERP, like Salesforce, you name it, any ERP is now just housing the data and a user interface to put it in. Now the real power is with Claude, where when you connect it, you just talk to it. You want to build a new dashboard that you see every day? Boom, done. You want to pull together stuff from your ERP and Google and maybe the SharePoint site where you've got all these documents? Now you get some real power. That's going to be a really fun one for you. It just blows my mind how fast a non-technical person can now connect Claude to their company's ERP and just hit the ground running.

Steve Goudreau: Yes. Scary.

Matthew Connor: Yeah. Because, oh hey, Jane in accounting just destroyed the whole... but for the most part, as long as you say, hey, careful, and you help them set it up and it's read-only, great. They can read these things, perfect, go to town. Now they can access all the information, the ERP, and they've got all the internet and all the company's documents. Where do you want to go? What do you want to know? That's where I think things get really, really exciting, when you're able to pull all that information together and create the kind of outcomes that you want, reports, dashboards, anything like that. It's mind-blowing, it's the coolest thing ever. It's just wild. I go on and on about this stuff, but I look forward to hearing how that goes for you guys. I know you're just getting into it, but I think this is where the fun is. This is where it gets really interesting for small businesses implementing these things, to advance the organization, because it's not even baby steps, just one step at a time. Before you know it, you're so far down this path and you're an entirely different organization next year. And yet everybody's job becomes so much easier and better, because you're not grappling with the ERP. That's what I think is really fun, it allows you to interact with the data and the information in a natural way. Just talk to it. Versus trying to finagle the user interface of the ERP, and oh, we don't have that report, and oh, to build that report is really hard. Now it's just, hey, give me this report, I want this in it, or you see this report, go ahead, I need to add these three things to it. You've got access to the ERP, you've got access to the report. Okay, here you go. Boom. Done. Freaking wild. Just wild stuff. Anyway, I go on and on, but I think that stuff is awesome. Steve, I'm going to go on and on about this stuff if you let me, but it's been such a pleasure having you on. I want to give you the last word before we go. Can you tell everybody where they can find out more about you and Ice Industries?

Steve Goudreau: For Ice Industries, you can go to our web page, iceindustries.com. We just launched on a new platform. We're showcasing all of our latest technologies for roll forming and all the business lines we have. We're geographically dispersed between Mexico, Louisiana, Mississippi, and Ohio. We have seven locations, and we're growing. We have strong leadership. So go to iceindustries.com and check us out. And if you'd like to see a meme every now and then, you can find me on LinkedIn.

Matthew Connor: I love it. Steve, thanks again for coming on. Until next time.

Steve Goudreau: All right, thanks.