The SiteVisit

AI Agents: Transforming Construction's Future with Roberto Cervantes, CEO at Flowlly

• James Faulkner • Season 6 • Episode 173

Share episode

Send us a text

Roberto Cervantes, CEO and co-founder of Flowlly, takes us on a fascinating journey into the world of AI agents for construction, revealing the crucial distinction between automation and true artificial intelligence. "Automation is a set of steps that are executed with no leeway to deviate," he explains, while AI brings the flexibility to navigate obstacles and find alternative paths to solutions.

The conversation uncovers a painful reality for many construction professionals - project managers trapped behind computers instead of being where they're most needed: on job sites. This observation became Flowlly's catalyst for creating AI solutions that handle the mundane document processing work that consumes valuable time. "My job is to be outside, not here in front of a computer," is the common refrain Roberto heard from construction managers, highlighting the urgent need for technological intervention.

We explore how AI is fundamentally changing our relationship with software. Rather than humans adapting to technology through clicks and manual inputs, Roberto envisions a future where technology adapts to us through natural, conversational interfaces. This shift represents a profound evolution - "Instead of us working for the software, the software is working for us." Yet amid this technological revolution, Roberto emphasizes that human connection remains irreplaceable, especially for building trust and relationships in construction. The greatest challenge isn't implementing AI but teaching people to critically evaluate its outputs rather than blindly accepting them. As Roberto warns, "The biggest risk is people being zombies, accepting everything the AI is putting out."

Ready to explore how AI agents could transform your construction workflow? Listen now to discover how the balance between technological efficiency and human oversight will shape the future of building.

PODCAST INFO:
the Site Visit Website: https://www.sitemaxsystems.com/podcast
the Site Visit on Buzzsprout: https://thesitevisit.buzzsprout.com/269424
the Site Visit on Apple Podcasts: https://podcasts.apple.com/ca/podcast/the-site-visit/id1456494446
the Site Visit on Spotify: https://open.spotify.com/show/5cp4qJE5ExZmO3EwldN1HH

FOLLOW ALONG:
LinkedIn: https://www.linkedin.com/company/thesitevisit
Instagram: https://www.instagram.com/thesitevisit

Speaker 1:

Welcome to the Site. Visit Podcast, leadership and Perspective from Construction with your host, james Faulkner, recorded live from the show floor at BuildX Vancouver 2025. All right, roberto, how are you doing? Good? And you, james, I'm doing very well, thank you. So you are co-founder of Flowly, yes, and you are making AI agents. That's right.

Speaker 2:

Very cool.

Speaker 1:

AI agents for construction, nice. And so, yeah, I wanted to chat with you a little about. I mean, I'm obviously fascinated with where AI is going, so maybe we can talk about some specifics there, yep. So first of all, how do people distinguish between AI and automation?

Speaker 2:

I think automation is more a set of steps that are executed, but there's no leeway for it to deviate. So if one thing breaks, then the automation stops there.

Speaker 1:

I see.

Speaker 2:

AI gives flexibility to that automation right. So even if you find an edge case, it will find its way around and execute it. So it's doing a little bit of thinking, if we can call it that way, but it's just finding its way through. So it's more open to automation. To just tell it this is where we start, this is where we end. You figure out the in-betweens kind of I see, okay Now.

Speaker 1:

So we've seen like big startups, like makecom, you've seen that Yep, where you can basically plug in Zapier does the same thing. So building agents to be able to do a number of different sequences to pull data from one source, crunch that data, then spit it out to another one and maybe do a loop of that. What other industries are you modeling some of this from? Is there other industries that are doing great things with this? Are you modeling some?

Speaker 2:

of this from. Is there other industries that are doing great things with this? I think one of the first ones is marketing, and I think it's where the initial LLMs really shine. They were able to write a lot of marketing copy and then create images and now creating videos. It's an easy way for someone to kind of automate their marketing teams. So I think marketing is one of those industries that's been really good. The other is the whole sales side.

Speaker 1:

Sales automation yes.

Speaker 2:

BDMs, bdrs, sending emails, even calling people the voice. Ais are getting so, so good that in a few years we're not going to be able to distinguish an AI to a human calling us. So those are the two that I think are taking more advantage of the AI agents right now. Other industries are starting to pick up, but those are the two big ones.

Speaker 1:

Yeah, so obviously we're looking at, um, how to save time. This is there's uh in construction. We talk about a number of things I mean, robotics gets into this too uh, which is dull, dirty and dangerous, and this is the dull part. Yeah, this is the stuff you're like. Seriously, I gotta go and crunch all this crap. It takes me hours a I don't want to do this. So initially, we had this transformation from paper to digital. That was like the first thing, and then the next one. Then is that, what do we do with that digital Right? So this is like the and I think that this is where I have an interest, obviously, with SiteMax and what we can be doing. Thank you In terms of the opportunity there is in order to have people doing more on construction, that is, more critical work rather than the dull work. So this is kind of your mantra is like how do you, how do you get people to be actually working in construction rather than uh on mindless stuff?

Speaker 2:

yeah, and I think when, when we started, uh, we were in toronto and we went out to talk with project managers and construction sites and they all said the same thing, which was I'm sitting here in front of a computer all day reading all these documents to try to do something, but I should be outside. My job is to be outside, not here in front of a computer, but for contracting, whatever, I need to be reviewing all these documents. And that was kind of like the light bulb moment for where it was like, well, the ais are really good at just reading the documents. They don't complain. They can read thousands of pages in seconds. So can we do something to help there? So so, yeah, the idea is how can we make people engage in more thinking activities?

Speaker 2:

and there's just normal copy pastpasting activities.

Speaker 1:

Right, okay, yeah. So it's basically alleviating the bottleneck of volume versus consumption of pertinent data. Yeah, okay, Lord's heart. So where do you see this all going 10 years from now? What's your pillow dream?

Speaker 2:

that you think about, when you think of. Where all this?

Speaker 1:

is going.

Speaker 2:

I think what AI and its agents and the way things are evolving is changing the way we interact with software. Before, like you mentioned, when we moved from paper to digital, it was still a lot of clicking on buttons. Right, that was the best we could do at that point with the technology, and that was fine and we all got used to that. But now these AIs are allowing us to work with technology in a way that doesn't require clicking on buttons, doesn't require changing how we do things. So, instead of us working for the software, the software is working for us, if that makes sense. So, for example, you can have the bot that joins meetings right and writes the meeting minutes For a human. The only thing they have to do is just click on the accept the bot to the meeting right, the same button they're clicking to accept all the other people who are joining the meeting and from there the AI will take over and write the meeting minutes. It could extract a lot of information, do graphs and whatever you want with the meeting data, right?

Speaker 2:

So I think it's changing how we work with technology and so that's going to completely change how we feel about software, right? So the new generations are going gonna grow up used to talking to an ai right, having having an ai as a friend and things like that. So they're gonna expect different things when they come to to to work right, instead of looking at things in a computer just asking, asking about to to find some information and tell me or create a podcast with that information, right. So I think that's where we are, is where we'll be seeing less user interfaces like we use, like we know now, and more interacting like if we were interacting with another human with technology. Oh, interesting.

Speaker 1:

Okay, yeah, I mean I think you and I see the same thing. That's good, yeah, no, I see exactly the same thing. So as you look around a show like this, you know here at BuildX, like, what are your thoughts? When you see everybody selling there, does this seem like an old school method to you in general, people trying to sell their stuff, communicating with each other and because, like when let's just say you have, instead of coming to the show, let's say I have an AI agent, that's like just get me all the information from all the booths, and maybe we need virtual booths because maybe that's the reason. But some things are physical. There's door companies here. You actually need to feel what it looks like. There's physical things and the physical world is difficult. But do you think we're going to get there in terms of I don't need to come to something like this for some things and I can just have an agent go and get stuff for me? Is it really about the repetitive and redundancy? That's just a waste of time.

Speaker 2:

That's a good question. I think some parts of this process, yes, but I think what we learned from COVID is we are social creatures, so we very quickly got bored of being inside right.

Speaker 2:

We crave going outside and interacting with people. So definitely there are things about this process that me sometimes, as an introvert, gave and I would love an AI to take over. But at the same time, seeing someone gives me certain trust that I maybe don't have if I'm just interacting with a website. So I think that that part that's so very ingrained in humans, the trust you won't be able to to take over with, AI, right, okay, what is so?

Speaker 1:

let me ask you this. So, in terms of, let's say that a lot of the, what are the risks of AI in construction in general on the job site? Is there like a what like the AI? And construction in general on the job site Is there like a what like the? There is a safety element between, and there's obviously going to be safety pluses, as in, benefits, as in can predict more things, et cetera, but there is a consciousness of human beings interacting with other human beings that a human needs to fact check. If you will, you need to actually open your eyes and go okay. So do you see the dovetailing of cameras and AI replacing that person's having to fact check that thing?

Speaker 2:

I hope not, and I think that's the biggest risk People getting complacent and trusting the output of the AI forever.

Speaker 1:

Isn't that inherent that they're going to do that though?

Speaker 2:

I think, yes, they will, and there probably has to be some control systems to prevent that from happening, because we don't want that happening. Maybe in five years, the AIs are so good that, yes, we don't need to supervise the output anymore, but right now it's not there and, by definition, an AI is just predicting the next token.

Speaker 2:

So, as much as we try to control what that prediction is, we never know what that prediction will be. So that means we always need to check the output, because it's going to be a random output, so there's always still a need for a human to supervise. So the biggest risk, I think, is that people are just being zombies, accepting everything that the AI is putting out Right and everything that the AI is putting out.

Speaker 1:

Right. So one of the things that we are experiencing today and this is where I think this is a great conversation- Maybe we need to have some booze or something.

Speaker 1:

So we're in an economy right now where people are in a consumption economy with their phone. So you'll see people spending. They'll come home from work, they'll spend four or five hours on their phone and they're consuming. They're not making anything, they're just consuming and the level of consciousness of the consumption is very low and it's required that they find something that tweaks their interest, because when you have so much volume, only some things get through for you to stop and watch that again or whatever it is.

Speaker 1:

So do you think that this is going to these controls that you're talking to have these controls in place to mitigate the flow? Like if you, let's just say, old school, you had to go and look through 50 documents and you'd be like, okay, and then you'd do an analysis between them and you'd go what are the trends here? Put something into a spreadsheet. You do an analysis between them, you go what are the trends here? Put something into a spreadsheet. There is an interactive backwards and forwards there for the consciousness of data. If you just give that to me, I might just rely on it, and this is kind of what you're getting at right on.

Speaker 1:

Do you think that there will become a point of complacency that needs to be mitigated, and I'm just trying to figure out how that. I mean the opportunities in construction are all things that are I mean, everything's mission critical, but some things are less crucial. They're all crucial, don't get me wrong. Everything's important in construction. But there are some things crucial. They're all crucial, don't get me wrong. Everything's important in construction. But there are some things where you're like yeah well, if we didn't get that right?

Speaker 1:

you know it wasn't a huge miss, whereas, like safety or something, it's something that gets killed. So there's it seems like there's opportunity for these controls to get put in place on things that are not so crucial. Right that these? Controls to get put in place on things that are not so crucial. Right, that are just a benefit to me.

Speaker 1:

So maybe there are things that, well, maybe we're in there. Wouldn't it be great if business? Wouldn't it be great if I could get this information? I can't get it today, rather than I have it today. It takes too long. What about the? I can't get it? Yeah, because the efficiency side is going to have risks, whereas if I never had it before, it's a net benefit regardless of what happens.

Speaker 2:

But do you? Right now we don't have it because we are not thinking about those things, because we don't have the time to think about those things, we're spending all the time doing the other stuff.

Speaker 1:

No, I just think it was too arduous and maybe it was coming out of multiple systems and APIs weren't connecting. Zapier wouldn't do it. You know, what I mean. We couldn't get it because there wasn't a method. So, that's what I'm saying. I think that there's some significant opportunities there.

Speaker 2:

Yeah, and that, I think, would help us take everything to the next level. Right, because now you're seeing things that you were not seeing before. So what does that mean and what do you do with that data? So definitely the AIs could be doing that, but I still think people the first thing they're going to go for is the automation of the things they're doing today, and I think they're part of the thing that we need to figure out is how to teach people to do critical thinking. How can they understand the output by asking questions to the AI to find hey, this number that you're putting here, why is it there? What is it coming from? So, trying to understand those things, and how do people learn? Today?

Speaker 2:

They first have to go through that process of being the ones filling in the forms, filling in the spreadsheets kind of understanding where the data is in the PDFs and all those things. But are they really thinking? Not right? They're just copy pasting and they're just walking there. So you start learning once you give that to someone and that person turns back and starts screaming at you because you did something wrong. Right now you're all okay, I messed it up here. So how do we help young people become that person that screams at the AI to know that something is wrong Because, yeah, the data inputting is no longer going to be done by a human. So what's the next thing that humans do today after someone filled in the spreadsheet?

Speaker 1:

Yeah, it's interesting when you say screaming at the AI, because that can be kind of interesting. It's kind of like, you know, when there's a company that you have a service with and they only have an email address for you to contact them and they're not getting back to you. Yes, it feels like that. Sometimes you get someone on the phone and you're like what the hell? Like well, I can't get this thing to do something. So, yeah, so you when you show up in places and obviously you know I want to sort of keep this confidential, but when you show up places and people are like, oh shit, the AI guy's here.

Speaker 1:

Do people think about their jobs? Do they see the net positive of maybe their job being easier, or do they see it as a threat?

Speaker 2:

I think it depends on who we talk with. If we talk with the people whose tasks the AI will do, they feel threatened by it, but then it's a matter of having them understand. Is it really you and your job filling in spreadsheets, or you're here for more than that? So having them understand that filling a spreadsheet is not a high value thing you could be doing.

Speaker 2:

So there are more things. It is not a high value thing you could be doing, so there are more things. But what we've seen change since two years ago when we started is then people started to hear about AI. They were concerned because all the things that they were seeing in the news is that things are hallucinating, that things are stealing your data. That was everything that was on the news. Now, when we talk with people, they're way more interested, because this thing has been happening for three years now. So more and more of my competitors are using it. More people are talking about it. So I want to know about it. Maybe not start using it, but at least I want to understand what this thing is and, if it can help me, how it can help me. So it changes and even maybe the people who felt threatened in the beginning now they're starting to see.

Speaker 2:

Well, maybe this is the way in which I can differentiate myself and the job site yeah, that makes sense, if I take a tool, then I can be more productive, more efficient and my boss will be happier with me yeah, that's cool.

Speaker 1:

Yeah, I think what we uh, it's going to be one of those things where you see the, the sort of larger companies doing the tech adoption and then sort of moving their way down. I had a podcast I did in Toronto with one of the guys from Lafarge Concrete Company and he's like the transformation guy and he's doing he's making agents, like you are, for them internally, and so it's an interesting path that we're at. Now. Let me ask you this At what point is making agents agentable? What do you mean by agentable? Well, at what point agentable? What do you mean by agentable? Well, at what point? I mean so to make in order to.

Speaker 1:

At what point can you ask an agent to make the agents like you're talking? Do you know what I'm saying? Yes, like, at what point does that? Because we're looking at that in software in general. Right, at what point do you get AI to make the software for you? Well, it's multifaceted in that point, right, because it's complicated-ish, but it can get easier Once one does it. Once, then you can start spitting this stuff out pretty easy, right, it's just the one that's going to be. So, you know, if I look at Sitemax, for instance, I go okay, well, what AI can just make our whole platform? Well, not that easy to do today, but once one person does something like SightMax, it's going to be easier and way easier and way easier and way easier. And then suddenly that's the watershed moment, right? But yeah, do you think about that in terms of at what point is the action become? Does it self-cannibalize itself?

Speaker 2:

So, for example, if you look at Zapier right now, they actually have that feature now where you describe your agent and it will try to find what's the best way to build it based on what they have right.

Speaker 2:

So that's already happening. But there are two pieces to the agent. One is what is the agent interacting with? So if you don't have those integrations, it doesn't matter what you ask the agent, it won't be able to pull the data or push the data. So that's one piece. The second piece is the prompt, which is what makes the agent act. So I think the prompt is easier-ish to do, as long as the agent, or this AI, is capable of asking the relevant questions and asking for the relevant information to then be able to generate its own prompt. The integration is where it becomes harder, because someone might be using a software you've never heard of and it's going to ask for that integration. So today you won't be able to do that. So I think that's really where the challenge is, because there is no universal API to connect to whatever is out there, so it won't be something that can integrate with all your tools whenever you need it.

Speaker 1:

I mean, it's getting, it will get that way, but it's going to take some time, yeah, but yeah, that's pretty cool, all right, well, that's pretty cool, man, um. So yeah, well, you did a good thing at the show here. That's been awesome, and I look forward to having more chats with you. Maybe this is just the first one. I think it might be, yes, um, but but yeah, we wish you luck with everything and yeah, let's keep our conversations going.

Speaker 2:

Perfect Love, it All right, well, thank you.

Speaker 1:

Thank you very much, Bye. Well, that does it for another episode of the Site Visit. Thank you for listening. Be sure to stay connected with us by following our social accounts on Instagram and YouTube. You can also sign up for our monthly newsletter at sitemaxsystemscom slash the site visit, where you'll get industry insights, pro tips and everything you need to know about the site visit podcast and Sitemax, the job site and construction management tool of choice for thousands of contractors in North America and beyond. Sitemax is also the engine that powers this podcast. All right, let's get back to building.