The Power of Data in Construction

Podcast Transcript 

Sarah McGuire: 

Hello, concrete revolutionaries, and welcome to the second episode of Building Better with AI. I’m your host, Sarah McGuire, and today we’re going to be discussing the power of data and construction and how AI is limited without it. I’m joined here by Alex Leblond, a seasoned professional with over 13 years of experience at Marcotte. As the co-owner and executive vice president, Alex has been a driving force in the industry. With a bachelor’s in information systems and a master’s in marketing, he brings a wealth of experience to our conversation. Alex, welcome to the podcast, and thanks for being here. 

Alex Leblond: 

Wow, thank you for inviting me. I feel so honored. Finally, we’ll be able to have our little talks and maybe embark and explaining to people how we’re embarking this beautiful journey on data-driven organizations. 

Sarah McGuire: 

Yes. Well, as one of the co-owners of one of the first companies that Giatec ever integrated with, I figured there was no better person to bring on here, especially because the way that Marcotte is set up. When I first came into this space, I was very spoiled with how easy it was to work with a system like Marcotte because you were cloud-based, because had that interoperability figured out from the get-go, and since that, it’s been about only a year and a half that we’ve known each other, which is wild, I’ve learned so much from you since then. Because of that, like I said, we were incredibly spoiled, and it’s been quite a journey in understanding the rest of the landscape and also working with you and so many other partners in this industry to actually bring everybody up into the cloud that we’re going to talk a little bit about today. 

But before diving into it, Alex, I would just love for you to introduce yourself to our listeners who aren’t familiar with you or your company, and I would love to hear how you got into this space, because I have a feeling you have a similar story that I did as somebody who kind of fell into this and got hooked, and we would love to hear about that. 

Alex Leblond: 

You’re certainly right. Thank you for that. You know me, I’m not the type of guy that’s going to talk 15 minutes about what we do, but yes, Marcotte, we’ve been doing process and automation control for the concrete industry for more than 48 years. So we’ve been through what we call the industrial eras of technology, right? Whether from back 40 years ago, when we had automated batch plants, with levers and push buttons, right? The human being was kind of using machinery to bring efficiency into production. As we went along, Marcotte got embarked in technology what we call computerized systems, right? So really the portion of automation, automating the plant, automating the production through computerized systems or solutions. 

When I joined Marcotte about 13, 14 years ago, I was actually in another industry, therefore doing automation for under the manufacturing industry, and to be honest, when I got into the concrete industry, it really compelled me. It hit me hard, like, “Wow, why are we so far behind? Why are we so far behind on technology?” And so that kind of became my passion in this industry. I fell in love with the people and fell in love with the industry because bottom line, my feeling is that, and it’s like that in other industries, but I think that people, the producers were told, “This technology is good enough. Don’t ask for more, don’t ask for innovation. This is just good enough for you guys.” Right? When you’re seeing other manufacturing on many other industries always utilizing, even back 15 years ago, like even in university, we were studying machine learning models even 20 years ago, right? 

So this is a technology, and when you’re talking about AI, we’re at 4.0, there are things that have been in the thoughts for a very, very long time, 40, 50 years of math behind these technologies, but we didn’t necessarily have the computing power nor the interoperability to be able to utilize or empower these technologies, and we’ll talk about that a little bit later, but you fall in an issue where you feel that, and I call it the close the platform, right? It’s the… Am I allowed to say names? It’s the IBM, right? You’re dictated by a solutions provider on what you can actually do with those systems, right? So when you have, for example, an ERP system, well, you’re kind of stuck using the closed platform ERP system, even though it’s not the best solution for your process or your production, right? 

So the beauty about that is, for example, combination of Giatec and Marcotte. You have small players out there in the industry that are pushing for innovation or pushing for data interoperability because we all say it, right, data’s the fuel, data’s the DNA of a company today, and that’s kind of where our passion was, how do we utilize data, because we have 40 years worth of data, how do we utilize that to solve real problems and bridge gaps that are certainly upcoming with the emerging technologies we see today? 

Sarah McGuire: 

Can you explain in layman’s term what do you mean by a closed platform ERP system or closed platform system? Because I feel like the entire industry is actually working on these systems without knowing what the term is behind it, and therefore, they don’t know the difference, and that’s really important to what we’re going to talk about. 

Alex Leblond: 

Very good point, and yes, I’ll try to bring the real basically, right? You know, you had these software models back in the ’80s that big companies that would be able to serve well, industries, right? And rather than focusing in one business function, being the best at it, the technology permitted them to make acquisitions maybe of competitors and close those solutions within their own platforms. Why? Much simpler, right? You’re not flexible, but it’s much simpler to manage. I’ve seen in other industries where they were faced with the same problematic, right? Look at agriculture, for example. Very, very close platform solutions, right? You’re running tractors, and there’s great companies out there that bring innovation that want to be able to use that data, and they can. 

So that’s what the industry was kind of facing, right? Luckily, in the last couple of years, and it’s always been Marcotte’s mindset, this is not our data, right? Our systems are producing data, that data is the company’s DNA, it should be utilized in a proper way. So when you start talking to companies like Giatec, Marcotte, that architecture or that mindset, that vision to say, “Hey, let’s share this data. We’re producing data, you guys are producing data. That combined data together certainly helps us bring more predictability in the process to be able to utilize, again, technologies like artificial intelligence and machine learning,” right? 

Sarah McGuire: 

Right. And that’s like the heart of what we’re talking about here, because when we came out, Giatec has had this optimization algorithm for years, actually. I mean, it’s gotten, of course, better and better as we’ve data mined over years with the help of all of our partners, but we had this optimizer that could be put into a system, but we didn’t have a system we could put it into, and at first, when we tried bringing this out, we actually started going to companies, seeing if we could integrate into their system and just license out the model that way. I mean, obviously as a company, we were a few years back then. Now we have bigger plans for ourselves as a whole, so I’m glad that’s not what we did. Otherwise, I wouldn’t be in the position to sell this thing. 

But still, we needed things to be able to hook in because AI can only learn from the data that it’s given, but this is not a problem that you and I are going to solve today. This is a problem that all of the stakeholders can come to, but I think we also need to talk to people about why this is a problem we’re solving, because yes, we’re dealing with all of these legacy products, but it’s not easy to just go and rip out your batch panels tomorrow, rip out your dispatch system tomorrow to upgrade into the cloud. This is huge. 

But I’m very curious to hear from you. You did your bachelor’s degree 20 years ago. What were you learning about back then? Because I can tell you from doing my bachelor’s degree in business, we certainly had a couple of tech courses, but we did not talk about AI. We talked about search engine optimization, we talked about big data. I don’t recall ever talking about AI, and I was a bit of a nerd, so I would’ve remembered. So I’m curious from your perspective, because you did an actual course in that, what were people talking about then, and where is it now, like how are you seeing that fitted now? 

Alex Leblond: 

Yeah, good point. Back then, we were clearly more into, again, programming language, right? And there’s something with programming languages today that, again, is source of many great things that we produce. Again, back then, when I was in information systems, it was clearly more on the data structure and, again, databases and how these databases can bring certain potential towards data. I wasn’t in school when the 4.0 era kind of rose, right? Now you have top 4.0. What does that mean? Now we’re talking about connected objects, devices, we talk sensors, right? Giatec, you guys, you have a sensor. This is a connected device that brings data back in the process. Wow. We didn’t have that back then. It was a little bit harder. 

But there were a lot of topics around machine learning, the math behind machine learning, behind algorithms, because again, in the systems we know today, application level, my developers tell me that all the time, is you can analyze certain variables or certain data points to learn from it, but you’re very limited in the number of data points that you can with the technologies. Forget AI, forget machine learning models. We’re talking pure if, and else, and but programming logic. You get at one point very limited. It’s very hard to maintain because you can understand the branches of if and buts depending on the different variables that you’re getting into your model, right? 

Machine learning, what happened in the last years is the machine learning, I believe, has always been there. It’s always been a science. Algorithm has always been a science. You’re a nerd, you know that. The computing power, that’s what changed, right? We used to build big mainframes back in the ’70s where everything was connected to those mainframes and we just had little small terminals, right? Aren’t we getting back to that era, where now you just have your little phone, it’s a connected device, and there’s data being populated from that? Now we live that in our everyday lives, right? I look at Jordans one day, and the next day, I’m on Facebook and I feel completely violated, right? I feel violated. I’m like, “All these pop-ups of the Jordans.” And artificial intelligence can be used, right? Can be used to super-humanize our human beings, and that’s what we’re trying to find, is using these technologies in the industry and how they can help. So back then, AI was not necessarily a big thing. Computing, cloud computing, interoperability devices, internet of things, big data, that’s what it brought, I think, the power of this technology today. 

Sarah McGuire: 

And now we’re looking at this in an industry where over the last decade or so, it’s really taken off because those systems are connected now and it’s allowing us to actually leverage it. But we’re not seeing that in our industry yet. We’re getting there, and there are people that are getting it, but we’re not there yet, and that is going to take, I think, years before we actually get to a place where even a company like ours puts something like this out there. We know we’re in it for the long haul, we know that this is something that’s going to take a lot of education because if I go to somebody tomorrow and say, “Do you want to use this?” but they’re not even on a cloud system of any kind, that’s going to be their first project. We’re not going to bother with them. We’re going to encourage them, we’re going to guide them, we can be their consultant advisor, so to speak, but until they’ve overcome that, I wouldn’t bother them with another product. 

Alex Leblond: 

Why? Why is that? Because you’re not going to be able to value your solution. Why? Because you don’t have access to the data you need to be able to bring value in that artificial intelligence or even in that technology that you’re putting in the- 

Sarah McGuire: 

Exactly. And we do get companies saying, “Can I just use your system and plug in everything manually?” And theoretically, yes, we did it. We did it. You remember, we did some trial batches with a couple of companies a year and a half ago. Some of them were great, some of them weren’t, they needed tweaking. It was exhaustive, but it was a great way of proving our algorithms were working, but we’ll never do that again. Why? Because even though it works and even if somebody is willing to pay us like tens of thousands of dollars to do it, it’s not a great experience. We would rather you say, “Put in the work of getting you…” You know, they say the longer that you go on a journey without building the foundation, that’s what we’re trying to do, we’re trying to get the foundation there first so that you can actually build up and use your time efficiently because time is money in this industry. 

Alex Leblond: 

It might seem insulting, but I talk to producers every day, and you said it well, I mean, the new generation of stakeholders are taking over these companies. There’s going to be a major labor gap. We know that. This industry is not sexy. It’s not sexy. It’s far from being. I mean, I’m looking at very, very young, talented people coming out of school. They certainly do not want to operate a batch plant, right? Or they certainly do not want to drive a ready-mix truck. What they want to do is monitor. They want have the tools. They want to utilize AI because they understand that these type of technologies will empower them within that organization to bring more added value somewhere else, right? 

I know you read a lot on AI as well, but you’ve probably seen this when they say, everybody says it, but I truly believe that 5 to 10 years from now, 30% of the jobs will be reinvented. That will impact our current industry, so the fact that we’re talking about these subjects today, I believe, is beautiful because it needs to go out. People need to understand that it will impact their business and they need to be ready for the smarter tomorrow. And again, between you and me, we could take Angelle as a great example, right? I mean, he went through a pretty big transformation in the last two years, trusting, and, you know, a company that was always keen to technology, I would say, but really on the forefront of early adoption, right? So we need those people. We need that in every industry, right? You need those type of clients. You need those type of producers that are willing to say, “I’m willing to embark on this journey with you guys.” 

So you come from an organization that was really into, call it, the close platform environment, facing challenges, whether it’s on human expertise, whether it’s on even accessing their own data, right? That’s a big challenge in this industry today. People are talking about BIs and how can we get all their data together to better understand, and for me, BIs are dead. BI dashboards are great, but it’s always too late, right? So even that’s a challenge when you talk about interoperability. So now you get into an organization like Angelle who are facing all these challenges on, again, labor expertise, and they are trusting technological partners such as Giatech, Marcotte, you know, the people that are bringing new technologies in these organizations and embracing them, using them, and clearly telling us, “You’re solving problems. You’re solving my challenges. When’s the next big thing, guys?” Like, “I need more,” right? “Your guys are helping my organization.” 

So to answer your question, the adoption comes from proof. It really does. In this industry, I mean, you still have mindset of people are skeptical about a lot of these newer technologies, right? You don’t understand this variable, you understand that every day it changes, and that’s the big problem about our industry, you know? Where I was before, we did machine learning, but to give you an example, it’s very repeat. When something is repeat, you’re cutting wood at 10 feet every single time, or you’re screwing up a bolt on a car. It’s very easy, right? It’s the same movement, it’s the same repeat task every single time. In concrete, we all know it’s not that. In concrete, there’s a lot of variability in the materials, in the conditions of the production, in transit. It changes every single second. 

Sarah McGuire: 

I agree that people, I think we logically are aware that that’s the case, but I don’t actually believe, and this is from our own experience now in plugging everything in, showing it to people all in one place, I don’t think our industry widespread actually understands the variability that they can actually control if they bring all of the data together because they’ve never seen it before, and we’ve just noticed as we’ve been going down this path, it’s actually very overwhelming for them to see everything in one place that way, and it’s exciting, but it is overwhelming because they’ve never had that opportunity before to see it, and now, when you’re actually being able to track everything, all of a sudden you can action it. 

Sometimes you can’t, and this is one of the things that we’re working through with our customers now, is sometimes we’re presenting something to the forefront for them, and they kind of knew that that existed, maybe not to that degree, maybe they couldn’t quantify it, but now that it’s been put in front of them, let’s say at its basic, an optimization, for example, but unfortunately, there is a prescription that prevents them from moving forward. So these are the things that we’re trying to figure out now. There’s also the fact of, okay, I knew that that was already there, but because we’re using AI, we got you there 10 times faster, and now you can just go. But we need that data, and that has been, we get asked a lot of the time about, “Well, if I…” 

Let’s talk about the Type 1L cement. This is something that’s been widespread across the country for about a year or two that it’s really started to pick up, and it’s created a lot of problems because the quality around it, people are so reliant on gut feel in this industry, which has miraculously done really well because people are experts, they know what they’re doing, but the next generation has no idea. They haven’t been out there touching the mud. They learned in school and then they were brought into it. They haven’t actually worked their way up in the same way that people are now kind of coming in, and it’s concerning because how can we bridge that gap? Well, if we don’t have the data, we can’t do anything, and that’s the biggest one. So I want to go back to the 40 years, again, of data, back to that. How? How hard was it to mine your own data? Because I imagine- 

Alex Leblond: 

It’s the hardest part. 

Sarah McGuire: 

Right, because it’s scattered everywhere. Then on top of that, you’ve had to mine other data, I’m sure, from other sources. Talk to me about that process- 

Alex Leblond: 

A challenge. 

Sarah McGuire: 

… that you’ve had to go through to get there, because I think a lot of our listeners are going to relate to it because they’re going to have to go through it as well as they start to migrate up. 

Alex Leblond: 

Thank you for bringing that up, and I’ll kind of bring you a little personal anecdote. I think you always like my stories. So one night, the four Marcotte partners, we’re back in 2018, and we’re just having a drink talking at night, like normal talks, like clients and where do we want to go as a business, and we have 40 years of data, and that’s the only thing you say, “We have 40 years of data, we have 40 years of data. We need to start using machine learning.” So we have one of our partners, Fred, he just gets up that night, doesn’t say anything to anybody. Two weeks later, he gets back, he goes, “The guys haven’t slept for two weeks. Red Bulls… But I’ve built the model in a way,” right? 

So what’s this model? And then we realized that we thought we had a lot of data. 40 years of data is a lot. We’re missing a lot of data, right? You get into the AI, you get into the algorithms, and now you understand you’re testing your algorithms, you’re testing your clusters, and then you’re realizing, “Oh, we’re missing this piece of data. Oh, cool, Giatec has it.” 

Sarah McGuire: 

Oh my goodness. 

Alex Leblond: 

“Awesome. Let’s get it,” right? 

Sarah McGuire: 

You were describing the challenges over the last six weeks of implementing with our next cohort of customers. It’s like, “Where is this coming? Where is this coming from? How is this getting in?” It’s exhausting because every time you move the needle, you think you’re there and then- 

Alex Leblond: 

And then you’re missing something. 

Sarah McGuire: 

… you go, “No, I’m not there.” 

Alex Leblond: 

Exactly. Or you’re there, but you cannot be as good as having that reliability to say, “Hey, you know what? We have all the data points that now permits us to have every different type of variable or factor that’s going to happen at that real time in point,” right? And when you talk about batching, for example, that’s exactly it. I’ll give you a great example. We have 40 years of data, and we say even through all these industrial eras of revolutionizing the way we produce concrete, more automated systems today, with BI’s reports telling these producers, “Here are how you’re going to solve your problems,” realize that there’s so much variability that it’s impossible, just impossible for a human being, just like in QC, to be able to be like excellence every single time because there’s so much variability, right? 

Now you get into these models of data and then you realize, “Oh yeah, that’s pretty cool, but we never really kept the flow rate,” right? We use it, but then we forget it. No, look, this is a variable that’s very, very important in this model, right? And now, oh, we’ve got sensors now like pressure in silos. Let’s take that data, let’s bring it back. Does that have an impact on our algorithms? Does that make us more efficient? Does that bring along more factors for better predictability, better control in real time and so forth, right? So that interoperability, it all comes back to, when people talk about interoperability and device systems, everything in a certain way needs to have this what I call an architecture permitting the share of data, right? Like I said, Giatec produces data in your day-to-day that is so important batch back to the production process to be able to implement better predictability or to be able to implement, “Hey, next time we do this batch, this is what we need to do to make sure we hit the target,” right? 

Sarah McGuire: 

Exactly. 

Alex Leblond: 

All that in real time in a certain way, right? 

Sarah McGuire: 

But to your point, our sensors didn’t even exist 10 years ago, and 10 years ago, they weren’t even cloud-based. So naive me, why would I think that a system out there is even built to take in that data when my system didn’t even exist before? And so we’re not sitting here saying to the industry, “Hey, you guys really need to get up into the cloud. You’re taking too much time,” yada, yada, because it’s not an easy thing, and that’s kind of the point of having this conversation now, is we need to be discussing this on a constant basis. 

And also, hopefully, we can start to learn our lessons of, “Remember when all of this stuff came in? Let’s try to build the best foundation possible from the get-go.” Because once things get too far gone, we even know from our own development, obviously, I’m on the sales side, I get excited when I get to put a product in front of people, and so I’ve had tons of frustrations with how long it takes for a product to get out there, and one of the biggest ones that will come up is realizing, “Oh, we didn’t build our system.” I remember when we had to translate the sensor platform that we had, we didn’t build it for multi-language. Well, now we’re hosting like 2,000 customers in here. This is going to be painful. 

So smart mix. Now we already, you know, that we built it with that in mind because we learned from that. And if we’re able to pass that onto the industry, some people would say, “Oh, when you allow for interoperability and all this connection to happen with other systems, you’re allowing for more competition to come into place.” But also, we’re creating a bigger ecosystem where we’re continuously evolving, and hopefully, we’re furthering our industry, which doesn’t necessarily have the best reputation. Like you said, it’s not a sexy industry to get into, like let’s make it sexier if we can to attract more talent because we’re going to need it, we’re going to need those people, and we’re certainly not trying to replace people with technology, we’re trying to elevate them in a bigger way. 

Alex Leblond: 

Exactly. Well said. And that’s what people need to understand about artificial intelligence, this is not robots taking humans places here, right? 

Sarah McGuire: 

No. 

Alex Leblond: 

We’re far away from the Terminator. 

Sarah McGuire: 

Yeah. 

Alex Leblond: 

What we’re utilizing or what we’re using, I guess, as industry, we call it industry experts in our own functions, we’re utilizing these technologies to empower these businesses and these industries, our clients, right? 

Sarah McGuire: 

I want to talk about Aurora. So I love that you have a name for Aurora just like we have a name for Roxi. Where did the name Aurora come from? 

Alex Leblond: 

I don’t know. Honestly, I don’t know. 

Sarah McGuire: 

Okay. 

Alex Leblond: 

I guess it’s a marketing… At one point, one thing was for sure is for us, again, we wanted to brand an artificial intelligence bot at Marcotte because Aurora has been specifically designed to learn from, again, machine learn from all the batch plant components, right? That’s what she’s designed to do. She’s not designed to do anything else, right? She’s designed to learn from every single piece of the batch plant and its process and the performance we’re obtaining out of it. So the name, definitely, for sure, for us, it was a female name. Aurora, I don’t know. I think it probably just based on, I guess, the light. 

Sarah McGuire: 

Okay. 

Alex Leblond: 

See the light, see the light button. 

Sarah McGuire: 

She sounds helpful. She sounds like we’re guiding us to a better… But today, Aurora, can you tell me exactly what she’s doing? Talk to me about some of the successes you’ve had already with her. 

Alex Leblond: 

Yeah, thank you for that. Well, again, four years ago, when we embarked on this journey of artificial intelligence, as you know, we’re not experts in artificial intelligence nor machine learning, so we got involved with the University of Sherbrooke. 

Sarah McGuire: 

Oh, okay. So did we. 

Alex Leblond: 

So did you guys. 

Sarah McGuire: 

Yeah. 

Alex Leblond: 

Very cool partner to have, smart people, and as you know, they also have their very specific concrete kind of quality lab. 

Sarah McGuire: 

For our listeners that are American, they obviously have the CIM program. We have Sherbrooke University, which is Quebecois as well, but don’t worry, you don’t have to speak French to go there. 

Alex Leblond: 

No, definitely not, you don’t have to speak French, and fortunately for us, they got involved in our project. Again, we needed validations, right? Is this going to bring benefits? In the very, very short term, they did say yes, the technology today, the algorithms permit us to do that, so we embarked in that journey with them. That was Marcotte’s, I would say, biggest, biggest investment in the last 4 to 8 years or last 3 years on research and development. Why? Because we believed in it, right? And we had great expertise. We had groups, we had firms. So we work with an AI firm in Montreal named Vooban helping us structure Aurora, right? So Aurora is built today, and the [inaudible 00:25:39], and we’re getting closer to that is have the autonomous batch plant. 

Sarah McGuire: 

Right. So when you say autonomous batch plant, our customers are scared. Some of them are scared. We’re going to come back to that. Anything autonomous is- 

Alex Leblond: 

Yeah, it’s always scary, right? And that’s why we’re doing it kind of in a very slow way, right? We’re not ready for the autonomous batch plant. Let me be transparent. Some customers say, “Oh, when is it coming?” 5 years, 10 years, 15? I have no clue, right? Because again- 

Sarah McGuire: 

It’s acceptance. 

Alex Leblond: 

It’s acceptance, but it’s also the technology, right? In a way where we have AI today, we can find the benefits behind AI. [inaudible 00:26:15] explain exactly what our AI does today and the kind of results we’ve been obtaining. Our goal at Marcotte is to certainly, because you said it well, seeing it is believing it. So part by part, what we’re trying to do is solve problems in the concrete production. The producers see we’re solving the problems with AI, and they’re asking us, “Okay, well, when are you going to solve this? And when are you going to solve this? And when are you going to solve this?” Now it gives us the fuel to say, “Okay, this is the next step, this is the next step.” 

So we have five steps towards the autonomous batch plant. We know the path. We know exactly what we need to attack, I would say, as problems. What we first did with our AI, Aurora, we have a set of algorithms called MIntelligent Feed. So what those algorithms do is that they, again, there’s a machine learning process, and I’ll get back to what you said before, I get very impatient as well because machine learning takes a long time, and I want to enable AI, I want to be able to see the results of AI on the feeding process of a batch plant, but as you know, it takes time. 

Sarah McGuire: 

It takes time, and the irony is that it’s going to save you so much time later on, but you have to be willing to put it in. 

Alex Leblond: 

And I always finish my closing statements to every producer I talk to in the industry, even after a conference or a podcast, well, I’ll say it now, every organization today should be doing machine learning, at least, right? 

Sarah McGuire: 

Yep. 

Alex Leblond: 

At least machine learning. I’m not saying you need to implement AI, but all the learning that you’re doing now will benefit you in tomorrow’s smarter future, right? I rest my case on that. 

But that being said, we did machine learning. It takes a while, three to six months, and now we’re able, in real time, six instructions a second. We have algorithms that control the feeding process of a batch plant. So we can predict free fall, flow rates, scale, stability, all the different variables, factors that happen in a concrete production. I’ll give you an example. A scale’s already filled up and it has a cone effect. We understand that the free fall, by the time I open up the door and the material falls to that cone or hits the scale, changes every single time, right? So that needs to be predicted, right? Needs to be predicted with the flow rate, needs to be predicted with the type of material. So is the cement fluffy? Is it not fluffy? Has it been packed in? When’s the last time we started the aeration? When was the last time the silo was filled? All these different little things, plus the mechanical behavior of the plant all affect the variability of our production. 

What AI is doing today is, in real time, obtaining the most optimal jogging to hit your target, right? So what we’re doing and what we’ve been seeing is, for example, in sand, rock, and even in cement, sand was actually the biggest one, we reduced auto tolerances, and auto tolerance, Sarah, for people who don’t know what that is, is every time you’re producing with an automated system, whether it’s Marcotte or others, when there’s an out of tolerance in your way up, you have an alert that says, “Hey, you have an out of tolerance, and there’s a correction to be made.” Something’s going to be corrected someday, right? Now, I might compensate with rock and sand. What happens? The consistency of my mix design is not respected, right? 

Sarah McGuire: 

But it’s typical. That happens all the time. 

Alex Leblond: 

The AI solves that. The AI solves that at all levels, right? 

Sarah McGuire: 

It’s not broken, don’t fix it, right? We hear that all the time. So that is a problem that is happening. It’s a problem because it’s not supposed to happen. How do you explain that to be a problem when we’re used to it? That’s a tough one. You can make your- 

Alex Leblond: 

It is a tough one. Well, actually, and again, we’re used to it, so again, we’ve been told. We’ve been told all our lives that the way we produce concrete with the systems that are available today in the marketplace, Marcotte, whoever they are, that it’s good enough. We automate your process. We don’t automate the full process. How many times do we see batch operators required in front of that computer screen to do what, start a batch, even activate a discharge? There’s a human being that’s required, right? So the human being’s there, he has expertise because he’s been doing it for 30 years, so he understands that when that vibrator hits off, there’s something happening with the dust collector. He knows all that, right? You said it well. 

Sarah McGuire: 

He knows it. 

Alex Leblond: 

He knows it. Right. 

Sarah McGuire: 

That’s the problem, but how do we quantify that? We can’t. 

Alex Leblond: 

We can’t. But when Alex comes in and starts batching tomorrow morning, how long is that going to take me to be able operate that plant? And also, I am responsible for the consistency of your product. I’m then responsible for the quality of what’s being batched there, right? So I need that expertise and I need to be on par all the time, right? And if something happens, I might just overshoot cement just a bit just to make sure the truck driver is not mad at me because I mixed… So there’s a lot of that, too, right? 

So that technology, what it does, the AI, is it replaces all these system parameters, all these parameters that human beings are saying, “When you open up two seconds, you’re going to weigh up 20 kgs, or 20 pounds, or whatever it is. And when you open up four minutes, this is what you’re going to weigh up.” And we’re working with two variables. That’s what all systems are doing today. Now we’ve implemented the language model behind that to say, “Hey, this is dynamic. I can predict the free fall. I know exactly when you need to close that gate. Close it, and we’re on target.” 

So what happens is now how do you convince people is actually when they see the data, right? I do machine learning today for about three to six months, and now we can present to a producer this actual performance. And today, unfortunately, they’re not very rich in BIs and not really rich in reporting to be able to get them to better understand how to solve these problems in their production. So again, that machine learning is also always a big hit, like, “Oh my goodness, am I losing a hundred tons of cement per month?” 

Sarah McGuire: 

Maybe this is not a very smart question, but I’m going- 

Alex Leblond: 

You’re always smart. 

Sarah McGuire: 

Anyways, if we have always been dealing with problematic mixes because this is just a thing that happens all the time, wouldn’t that mean that the data that we’re pulling from is kind of garbage data? And we always say that garbage in, garbage out. So how do you work at fixing that problem? Because we also have that issue with, okay, all of these mixes are over-designed, but we accept it, it’s normal. So it took a lot of sorting through that, and actually, I’m not going to speak to what our algorithm team did because that’s not my area of expertise, but that’s not easy to work with, and then on top of that, a lot of the pushback that we get, which is totally fair and exactly why we need to solve this issue of interoperability, is, “Well, what about my rheology? How do I know when the water’s been added? How do I know that I can take this optimization if you’re not getting all of that other data?” and our issue is that we haven’t been monitoring that long enough for us to say, “You’re right,” or the opposite. 

Alex Leblond: 

Or maybe the opposite. 

Sarah McGuire: 

Yeah. They are absolutely correct. That is why it is not perfect. And I also, even when we do solve those problems, I guarantee we’re going to open up another can of worms and find 10 more, and it’s going to be a constant move in that way. But I’m curious how you’ve tackled that, because you just said it takes three to six months of kind of mining and ironing it all out, so it takes a while for you to actually see something. But when they first start to get those results, are they willing to accept them at face value? How has that been so far? 

Alex Leblond: 

Well, again, it might be a little bit different for you guys because for us, it’s pretty factual, right? They’re used to seeing data such as what are my targeted weights and what are actually my actuals, right? So you get back to very simple fundamentals of saying, “Hey, this is what you’re asking, and this is what you’re producing,” right? Now, you- 

Sarah McGuire: 

So you always have that deviation to pull from. 

Alex Leblond: 

No matter what, right? And you said it, right? In production, in concrete production, it’s been accepted. It’s been accepted that there’s variability in this production, and it’s normal that we have variations. Therefore, we’ll put in compliances and regulations in that sense. And for us, that’s where we believe that AI is going to revolutionize the way we produce concrete, right? Same as you guys. It’s artificial intelligence will guarantee precision and consistency, and it’s going to get better every day. 

Sarah McGuire: 

Alex, we’re going to close it off here soon, but I just want to ask, looking ahead, what are you most excited? Obviously, there’s the autonomous batch plant, and I’m not looking for any flattery on our side either on this, but what are you most excited about when it comes to technology and AI and where we’re going in this industry? 

Alex Leblond: 

What I feel passionate about this industry is that more and more, and I think you’ve seen it as well in the last couple of years, is the traction that producers are giving companies like Marcotte and Giatec and others on, “Wow, this makes sense,” right? And proving, you said it well, that we’re actually proving that these technologies are bringing benefits and waking up every morning saying, “Okay, what’s the next big thing?” Because one thing is for sure, no company can keep up with technology today. Technology just move us so fast. I know you’ve read things, but we’re processing information like within a week as human beings today that our grandparents process in their whole year and their whole lifetime, right? So that’s hard for human beings. We’re living AI every day. 

Personal anecdote for me, I mean, ChatGPT, you saw the adoption, right? When ChatGPT came out, look at the adoption of members that just logged into ChatGPT and started using. It was crazy, right? More than Amazon, more than Facebook times a hundred. Today, I have my personal assistant. I’m super-powered. I believe I’m super-powered. Why? Because I use ChatGPT every day. I miss a one-hour Teams call. You know what happens? I don’t necessarily have to be there. I transcript that call, put it in ChatGPT, “Summarize it for me for five minutes. Please tell me if I have any to-do’s, and who should I follow up with, and what was the actual plan.” 

Sarah McGuire: 

That’s the most… Why have I never done that? I’m obsessed with that idea. 

Alex Leblond: 

So I’ll show you how to do that. 

Sarah McGuire: 

Oh my gosh. 

Alex Leblond: 

But you know what? It takes me five minutes that day, and if there was something important for me to get from that meeting, I’ll know if I need to follow up on it. But you know what? That 50 minutes I lost, not that I lost, but I was doing something else, right? So I was super-powered and doing something that’s probably more valuable for me personally or even for the company. 

Sarah McGuire: 

I thought I was a genius for using that to plan my honeymoon, and that is next level. 

Alex Leblond: 

That’s next level. 

Sarah McGuire: 

You are literally making time in your day by doing that. 

Alex Leblond: 

I’m super-humanizing myself with these technologies. 

Sarah McGuire: 

I’m just mad I didn’t think about it sooner. That’s- 

Alex Leblond: 

Don’t ask me to read a book anymore because I’ll tell you one thing, I’ll pop that in ChatGPT, I’ll call you the next day, I say, “You know what, Sarah? It took me all night. I read that book. Let’s discuss on it,” right? 

Sarah McGuire: 

Alex, thank you so much for joining. If somebody is interested in learning about Marcotte, and actually, if they’re interested in learning about what we can do together, how can they reach out to you? 

Alex Leblond: 

Yeah, I would say just marcottesystems.com, and I do also have a LinkedIn profile, so just look up Alex Leblond on Marcotte and I’ll be there. My email is a little bit too long to say. 

Sarah McGuire: 

That’s fine. We will put everything in the description of the podcast so that people can reach out to you. 

Alex Leblond: 

It’d be a pleasure to have a talk. I always love chatting about challenges in the industry and how we can help. Let’s have some experience, man. Let’s try satellite, let’s try drones, let’s try IR, right? These are great technology that we can certainly do and help, so yeah. 

Sarah McGuire: 

I love it. Thank you so much for joining. 

Alex Leblond: 

Thanks, Sarah.