Bridging Traditions and Technology: Bluon's AI Innovation Reshaping HVAC Industry

Embark on an extraordinary tale of transformation as we welcome Peter, CEO of Bluon, and Adam, SVP of Software and Data, who recount their journey from the world of chemical R&D to becoming trailblazers in the HVAC industry. Uncover the genesis of Bluon and the way these visionary leaders addressed the critical Freon issue, offering a doorway to a world where immediate access to retrofit instructions and comprehensive manuals is now at every technician's fingertips. Their narrative illustrates how innovation springs from the most unexpected places, reshaping entire industries and redefining the role of data and software in hands-on professions.

The age of artificial intelligence has dawned, and with it, our guests share the emergence of an AI system that's altering the landscape of tech support. Imagine a reality where the collective wisdom of seasoned professionals is encapsulated within an AI, offering instant, accurate responses to complex HVAC inquiries. Peter and Adam discuss the creation of an 'AI treasure trove,' the potential consciousness of such systems, and the profound implications for learning, problem-solving, and the integration of these advanced technologies within a traditionally skeptical trade community.

Lastly, we tackle the formidable challenges facing the trades industry today, from the decline in traditional apprenticeships to the generational shift towards just-in-time learning. The episode examines the revolutionary impact AI could have on HVAC technician training and the imperative for businesses to adapt or risk falling behind. Peter and Adam offer insights into how Bluon is equipping technicians and companies with self-directed learning tools, ensuring the vibrancy and sustainability of the trades for future generations. Join us for a thought-provoking discussion that bridges the gap between practical expertise and cutting-edge technology, setting the stage for a brighter, more efficient tomorrow.

http://bluon.com

Transcript

Speaker 1: 0:56

Welcome to the successful life podcast. I'm your host, corey Barrier, and today, folks, we have two people with me, which is not the norm, peter.

Speaker 2: 1:09

I even wrote it down, dude Cappuccino.

Speaker 1: 1:12

Man and Adam Curry, not the MTV Adam Curry, the other Adam Curry. What's up, fellas?

Speaker 3: 1:21

Good stuff, man, appreciate you having us on Excited to be here.

Speaker 1: 1:25

So you guys are Peter, the CEO of BlueOn, and Adam, you are. What of BlueOn?

Speaker 2: 1:35

SVP of Software and Data, but I'm basically the resident man at scientists. No most of the time, Peter and I both actually.

Speaker 1: 1:44

So scientists and the trades don't typically go together. In fact, I can't think of two more polar opposites. So walk me through how in the world two scientists by trade, so to speak got into working with technicians and dealing with the trades.

Speaker 3: 2:06

Yeah, for sure, I'll give the cliff notes version. So I think it really started. If you really Both Adam and I were looking for some purpose, that you can use your scientific endeavor perspective to solve real-world problems that are meaningful. And our original, like BlueOn, was not founded as a software company at all. It was founded as a chemical R&D company to solve the Freon problem, the R22 problem. This is back in 2012, when the first idea of BlueOn came to be and make a long story short we were literally guys in lab coats developing a new refrigerant. For four years. That's all we did testing, testing, testing. So we met all the techs. We were in the field, constantly testing literally hundreds, if not thousands of formulas. So we ended up producing a hell of a refrigerant.

Speaker 3: 3:08

But that refrigerant the production of it and the manufacturing of it and the selling of it is where we sort of intersected with the trades and started to figure out we had another journey to go down and it really started with us getting into the so back up a little bit. So my background I had a lot of commercial real estate contacts with me coming through this sort of journey that were ready to go on a lot of retrofit projects. They were ready to go as we were getting the product finally approved by EPA and ASHRAE, which is a whole other nightmare for another conversation of bureaucratic disaster, which is just horrific. But anyways, we got through that issue and we felt like, okay, we solved the world problem, we got approval, let's go. But then we get out in the field and we run into a much bigger problem, which were HVAC technicians who, as I always say, can't say yes, but they share as hell can say no, and they were thinking they were a block. They're like yeah, we've been burned by replacement refrigerants. We want no part of this. Like, get out of here, right? No, thank you. And we heard that over and over and over again. They get burned so many times.

Speaker 3: 4:14

And so we were stuck in this conundrum of this demand, this amazing product which we loved, and a very skeptical, if not challenging, hvac tech community who want to know part of it. So we had to figure out a way to get the techs as advocates or champions of the product overnight. How the hell do you do that? And so, by listening to the techs we had a few on staff. Their biggest challenge was that they never were told what to do. So they just handed a tank of gas and like, have at it, like, go for it, and they were breaking compressors and breaking systems, particularly in the light commercial and commercial space a lot, and they were completely left out to dry by the manufacturers. So we said, okay, well, let's build a database of every model we can think of. And back then we didn't realize there were millions of models. So our naive day was our source of strength back then.

Speaker 3: 5:12

And so we started on that journey of creating a database of about 40,000 models at the time. And then it was Adam's idea to build an app around it and put it out there in the community so technicians could look up the app, look up the model number and be told exactly what to do, what not to do. So it was basically a red, it was a green, yellow, red system. Green was straight shot and goal like do your normal, your normal that's like a third of systems, about a third or yellow, meaning you got to modify, tune, change, tweak something. If you don't, you're going to break it. So do this, this and this. And then red was like run like hell, do not put a replacement in this in the system, it will explode, and so we built that database.

Speaker 3: 5:50

In the process of building the database, we collected all of the manuals, wiring diagrams, service bulletins of those models and because we had them, we put them in the app. Now this is a brand agnostic app. So at the time, like I don't know, 60 or 70 brands, oems and the sub OEMs, and so we put it out there. At the time we had a test, like a 45 minute test I mean class and a test before you get access to the app. It was free but you had to get accredited.

Speaker 3: 6:16

So it was a bit of a wall right for a tech to climb over and we were blown away by how many techs went through that process, not because they wanted the retrofit instructions, but because they got word that you get a brand agnostic information repository of every model you can imagine and you could pull up the manuals very quickly because Adam built, they built a great app and that that started the journey of us realizing there was this huge thirst for information on demand in the field brand agnostic quick, easy, no BS. Give me access to it when I need it, don't give me a lot of a headache, and you know, within a couple of years, we had 100,000 techs on the platform gobbling this information. I'll be 170,000 now gobbling this information up because no one had ever put it in. Why would you? A brand agnostic database? That was, you know, super fast and super easy to digest.

Speaker 3: 7:09

So that's how we ended up, adam and I, in this space. We were not planning on getting. There was a. It was a journey from chemistry to need to oh my God, we're going to build that. And then you know, adam having the idea to build the app and back then it was a little bit Adam and one outsourced Dev and Bulgaria did all the development for the first couple of years, right?

Speaker 1: 7:30

That's right. It's a lot of work. I mean, like now, right now, you can pull all that manufacturer data into and, adam, I'm sure you'll speak to this using AI and maybe you're using some form of it then but to normal people like myself, like AI has only been around. It's been around for years, but to the normal person, it's only been around for a very short amount of time. So how did you aggregate all that data without having the tools that are available now?

Speaker 2: 8:04

Mostly by hand. So one thing that we had done is, yeah, that's the right reaction, that's wild. You know, we had built the team of, on the one hand, experienced HVAC technicians you know, 10 years plus field experience and on the other hand, had computer chops and their task full time. Full time job was to go through the scour, the internet, basically looking for HVAC models, and pull that information, find the manuals, go through the manuals, make sure everything's correct, put that into a, an online Excel spreadsheet program, basically, and then from there we turn it into into the database. But we did that by hand for years.

Speaker 2: 8:57

We have we a couple years ago acquired a company called X ref which was the only other sort of database of models and parts, and once we had that, then we were able to, we had this sort of go through and clean it up, which is this whole other side adventure. But that also began our, our, our move into parts as well as manuals and models. Yeah, but yeah it's mostly been by hand. You know the AI stuff is is a year old.

Speaker 3: 9:29

What's important to realize is that, you know, when we acquired X ref so X ref had had also been doing it by hand since like the 80s, right, and so their data goes way back and now we've incorporated it all.

Speaker 3: 9:41

Now we have about, you know, half a million models for models. You know maybe 20 million extended models, so pretty much the market. But all that data that we acquired next ref, even though it was great, it was messed up, right it was it was heavily corrupted in terms of you know they had, because they built the database in the 80s, right, so they had data restraints, like data constriction, you know, contract contraction, removing hyphens, doing all these things you would never do now. They did because they have to save space and we had to go back and undo it all, like an Adam and a team of you know a dozen of these unicorn HVAC tech database guys spent, I think we said, like 70,000 man hours going through this data and extracting it and putting it back together in a way that was useful. We have all that. We have all the part numbers, all of the cross reference part numbers, all the specs. It's like 60 million connections. You know five million skews. You know half a million models. So the ton of data and it's the way it connects together which is the magic, because you can find compatibility, which is where the industry really struggles.

Speaker 3: 10:40

But yeah, turned into a you know this, this escapade into this incredibly fragmented data which nobody has all of it, we're the only people now it has. Everybody's got a little piece of it right, every little chunk, and you can't Google it. You know things before 2225 or just not there, you can't find them. So we have like the only source material of some of that stuff. Oems ask us for their data back that they no longer have from like 95, 2000, you know that, those kind of time frames.

Speaker 3: 11:08

So it's amazing how fragmented and how you know uneven that playing field was and nobody having really the you know hand on the data and it's so valuable. It's remarkable to me that no one did this and we talked to folks like that, some of the big aftermarket manufacturers who would be in their interest to do this work, and they, to a T, told us yeah, we try to do hard, can't do it, as we were told that from like three of the big boys, yes, basically looking at saying you can't do it like well, we did it. It was an ungodly amount of effort, but we pulled it off. So we're pretty. We're pretty stoked with how the data came out and its users are, just, you know, everywhere.

Speaker 2: 11:45

I think that you know, it was the right thing to because it was so hard, it kind of made it the right thing to do, or it signal that it was the right thing to do, because it's one of those things where One key to this success and your start-up whatever is find the hard thing to do and do that. Not all the time, but find one of the hard things to do and do that because you're going to unlock value that no one else has unlocked. And, particularly in our case, everybody knew that if you could get a database like we'd built, it would be unbelievably valuable. Everybody knew that and they all told us that. But they also at the same time said it was too hard, right, right. So one of the reasons why it was too hard is because in the trades you're not really a software company usually and in software companies you have no idea about anything.

Speaker 2: 12:38

Blue collar, right. So it was up to. It was basically Blue. Honor is the only group that could do it, being composed of people in the trades and people in software, kind of sitting in a room together every day.

Speaker 1: 12:53

I think that's the part that you for sure did right is involving people within the industry along beside yourself that understand software. We do some AI automations and not anything compared to what you guys are doing, but we do have an HVITE technician knowledge bot, if you will, and the reason we created that is very similar to what you're talking about is what I find is technicians are in the field, they need a question answered and what do they have to do? They have to call back to the service manager or the sales manager and wait on that guy to call back, or maybe they don't call them back and they're standing there by the customer, and so the reason I bring that up is it's been a nightmare trying to get most people to use it and it doesn't really. It didn't cost anything because it's built on. I can't remember how you built it. I have an engineer that works for me that builds all the stuff.

Speaker 1: 13:54

I'm not the software guy, but the point is, I guess, is that's what you did. Right is bringing those two people together, because it's really hard as a software person, or even an outsider, to communicate the way trades people need to be communicated with. It's just different. It's just a different breed. That's all I mean.

Speaker 3: 14:17

Well, I think that's where the story becomes more interesting for this conversation. So, as that, we went through how we became a software company. It's phase one, so then phase two we sort of leaned into the need, the support, this huge thirst, and so back in 20, I think it was 2019, we launched our tech support, which was a live human being that could answer a question, which took off. Also, of course, we did it for free at the time, and so we were getting ungodly amount of. We were doing like six, seven, eight, nine, seven hundred questions a day with a team of 30 techs, all of which had 30 years experience of more. These are the cream of the crop techs, amazing, amazing people.

Speaker 3: 14:55

But we were recording all of these Q&As for the last five years, about 100,000 calls, and so those calls became the source material for the AI that Adam built, which is a whole other story, which is interesting what you said. So we had this database of this treasure trove of practical question and answers that were directed by a tech to a tech in a way that you can't look in a book. So now we can query that in a large language model AI and pull out the nuggets that are specific and practical. So techs like using our AI because it speaks their language. It's written by them effectively in a way that you could not do otherwise. You would need to go out and offer free text for five years to get the data in order to build the AI we always used to talk about.

Speaker 3: 15:44

How do you get the wisdom out of these techs that are just these walking knowledge machines. How do you get it out of there? And we did it accidentally. We just recorded all the techs' work calls and then AI came along and it proved to be the magic to pull it out. But yeah, totally right, you could never do it because techs are incredibly. They don't like academic writing at all. They despise it, as you know, and so you really can't just give them an internet source, chat, gpt source to answer. This is too technical and too academic. They're like it doesn't hit the nail in the head with the practical answer. So I think we sort of accidentally fell into this AI treasure trove based on Adam's midnight to 6 AM working for six months when no one knew what he was doing, and came out one day about nine months ago. Hey, check this out. We're like what, what is this? The AI was born and so from there it's been evolving. But yeah, it was all about the techs' work calls.

Speaker 1: 16:40

So would you all right?

Speaker 1: 16:41

So let me ask you both something.

Speaker 1: 16:45

In my opinion, my experience with working with AI which I've done for you know, since chat GPT came out, it was like fireworks going off in my head and I've just been knee-deep in it, and so I want people to really understand that this kid, this is just my opinion, this is how I've used it.

Speaker 1: 17:04

If there's something I don't know, I can ask chat GPT whatever that thing is reverse, engineer it and I can understand it at an astronomical pace compared to before chat GPT was born, and I wish people would really get that concept because you, like, I believe that you can teach yourself anything you want to know about. Now you have to be careful, because some of the information is not exactly right but depending on where you're getting, where it's sourcing it from and now that it's tied in with GPT four is tied in with the internet you know that could give you more capability, but also could give you, you know, skewed answers also, but for the most part, my experience with this is that you can skyrocket the amount of information that you want to learn about something. Would you agree with that?

Speaker 3: 18:25

Yeah, I would.

Speaker 2: 18:27

Absolutely. It's like strapping a rocket packed onto whatever you want to do, whatever you want to learn, whatever you want to accomplish. It's an unbelievable tool that is game changer in any discipline.

Speaker 3: 18:41

Yeah, yeah, I mean what you said earlier is, you know, getting people to use it is the key, right? So we've been really trying to make techs understand the availability of it, the ease of it, get them into it, train them. Once they see it, they're in. But it takes some doing. They're naturally or inherently skeptical of that kind of information. So once they realize it came from techs, it's their peers who created the content. There's a more of like a, you know, gradual acceptance of it, but that's the biggest barrier is just adoption 100%.

Speaker 1: 19:15

So, adam, I want you to dive into the conscious. You mentioned consciousness earlier. I think a lot of people believe, and maybe this could be true to an extent, I suppose, depending on what level of AI you're talking about. I know the military's had some, maybe a couple of mishaps, but can you dive into, you know, I guess the I don't know if there's I wouldn't say there's consciousness. I actually don't know, I don't even know the question I'm trying to ask, but just dive into consciousness as it pertains to AI, if you would.

Speaker 2: 19:51

Sure, yeah, now you can go by it, you nailed it.

Speaker 2: 19:55

Yeah, when people say consciousness, that tends to mean different things to different people. But in this context we mean is it sentient? Is the artificial intelligence awake and aware? Does it have a first person experience? Is it alive? Is maybe another way of saying it? Is it awake and alive?

Speaker 2: 20:18

And you know, this is something that people discuss and have discussed for a long time, going back even to the 50s, in which Alan Turing developed his Turing test and he said that you could consider a machine conscious if it can, if one person can interact with both a machine and a human and not know which is which right. Well, the problem with that is that with chat, GPT and other types of AI systems, we're pretty much right there. A person might not know whether or not they're interacting with a machine or a human, but that doesn't mean that the AI has consciousness. In fact, I don't think it does. So that's sort of where we are right now is. Ai can resemble humans very closely to the point where we don't know which is which, and we actually don't yet have a good heuristic or toolkit by which to measure whether or not these things are conscious or not. I will say this as a general comment People tend to underestimate consciousness and overestimate AI.

Speaker 2: 21:33

They underestimate consciousness because consciousness is a mystery and most people don't realize how much of a mystery it is. It's not just an illusion produced by the brain. It's something more than that. There's a lot of research that shows that it can do some pretty exotic things as well, even outside of the brain, and so we underestimate it. We overestimate AI mostly because it's so damn impressive that it seems like magic. But if you were to go down a couple of levels into how it's actually working, you can kind of see the man behind the curtain and you realize that, as impressive as it is on the outside, what's really going on is a whole lot of statistics, basically just a whole lot of layers of deterministic probabilities. It's awesome, but that's all it is. It is not the sort of the magic of a human being or an animal or something like that, quite yet.

Speaker 1: 22:44

I wonder if a lot of the perception around whether it's conscious or not here's what I think, because it can build AI, can build on AI, and I think that's where a lot of people, maybe a lot of people, think that, well, there has to be a level of consciousness. So I agree with you, I don't believe it's conscious, but I do believe it resembles a level of consciousness that, as you kind of alluded to, it's deceiving.

Speaker 2: 23:19

One of the reasons for that is because we've built it to resemble human behavior. So, up until about 2018, the way that artificial intelligence was being approached was by creating ever more sophisticated algorithms, like, basically, computer programs like you normally think of them, and 95% of AI scientists believe this is the way towards AI, and 5% said no, I don't think this is going to work. And they were proponents of what's called neural networks and tensors, and basically all that means is it's a totally different way of developing the AI. In terms of software, the neural networks are basically like levels of probability that get updated and go really deep, like billions of levels deep, and it wasn't until about 2018 that these 5% of AI scientists had the cloud computing resources necessary because the economies of scale have brought the cost down in order to build models large enough to test their theories. And then it turns out that these theories over delivered and nobody really questions that anymore and, to the credit of the rest of the academic world, which is very rare, they said we were wrong and that tensor model with neural networks is the technology behind GPT, chat, gpt.

Speaker 2: 24:55

So what does that mean? Well, you have a large language model, which is basically you take great gobs of text from the internet, mostly in English, and you train these neural networks on okay, if, given this word, what's the likelihood that this word follows? If, given these two words, what's the likelihood that the third word is X versus Y, and so forth, on billions and billions and billions of pages of text. Okay, so you develop these probabilities of what word comes next, and that's the magic behind chat, gpt. So of course, it's going to resemble how a human talks and interacts, because that's literally its training set. You know English, english language. Now it turns out that this is a. Tensors are widely applicable to other types of things, like self-driving cars and so forth. We're still sort of figuring that out, but you know it does more accurately mimic the human brain. As we understand it, though, neural networks are much more like neurons versus. The previous approach was algorithms, which, you know, you don't really see so much in nature. Nature isn't so deterministic.

Speaker 1: 26:09

Well, to be able to answer questions, you know the way people feel is super important and the way the words that come back to me make me feel can change my perception of that information. Right, and I think that's where. What did you? You didn't say sentiment analysis, you call it something similar to that, would you call it? A second ago Sounded like sentiment analysis, but it wasn't. It was, oh, sentiment Sentiment. So sentiment is embodying the feeling of what it's trying to spit out, right? Do I understand that? Right?

Speaker 2: 26:50

Yeah, you know, and I think that's the magic behind master mechanic Luan's AI Is it because it's training set? You know it's soul, if you will, is tech conversations to other techs in plain English, all the colloquialisms about what they're seeing and you know how to diagnose it. When you particularly if you are tech and you start talking to this thing like you, normally would the answers come back, like you would expect them to come back, and they come back to you in very practical ways. Part of that is literally that the way that master mechanic is built, it is built upon the like, the meaning or the sentiment behind the words that went into the, that went into its training. We can get into that if you want, which is well, so that's it.

Speaker 1: 27:44

Yeah, I would love for you to get it. And those words are coming from the recordings that you built over the five years. Like it makes complete sense to me, it makes total sense why it operates in. For me, anyway, it makes complete sense why it spits out that verbiage Like that was genius.

Speaker 1: 28:00

And imagine all the companies out there that have all these calls, since service Titan or whoever, or Billy over at Sarah, they record all of these calls and most of these guys don't do a damn thing with it. It's like a treasure trove of data. And the reason I get amped up about this is we we developed, we developed an AI that would you know. It would tell you how the sentiment analysis, so to speak, of the call is going to be for the CSR or CCR and the customer, to where you wouldn't have to go back and listen to the call. Right, and I know that makes sense to you guys. And we're in the. We got the MVP portion of it bill and we're in the process of building it out for a company. But I think it's, I think it's, I think it's next level. That's what I think.

Speaker 3: 29:00

Amen Go through the.

Speaker 2: 29:01

It's pretty cool.

Speaker 3: 29:02

Go through Adam how you you know, hold it apart and yeah, okay.

Speaker 2: 29:08

So you know, blue on, we've got 100,000 tech support recordings, right. We picked 50,000 of the best. So the 50,000 recordings we converted into text, right, so we transcribed them into text, okay. So how do you get this into? How do you build an AI on top of this stuff? Well, one of the difficulties here, that one of the main problems to solve, is, if you just approach this like you would a normal search engine, meaning you just put it into a database and allow people to search it, it's not going to work. And it's not going to work because the way that the conventional search, like Google, works is it matches the characters that you type to characters in the database. So, for example, I'm a tech and I type in what it. What does error code 37 mean? Well, the database might be able to find err or space 37. But what if I, what if? What if I typed? Or what if the database had fault code 37, fault 37, there's no match. So it's, you can't match the characters. You have to somehow match the intention of of what is being said. You have to match something that is not captured in the words but between the words. It's this like invisible meaning, right?

Speaker 2: 30:41

So it turns out that there's, there's another branch of computer science that has to do with meaning, capturing meaning, and it's called vector space. And so in vector space what you do is you take a chunk of text and you take each word and, if you will, you put that word on an x y coordinate. Okay, so you put one word here and that's got you know, 00. And then the next word. Let's say, if it's, if the first word is dog and the next word is cat, you probably put those close to each other. But if the third word is zebra, it is an animal like a dog and a cat, but it's not a pet, right? And therefore it's a little bit further away in your x y coordinate. Now, with dog, cat and zebra, it's x, y and z, right? So then your fourth word.

Speaker 2: 31:43

You would plot on this, on this graph, but now it's a four-dimensional object. So essentially, what you do is you create an n-dimensional object where the distance between words is captured in the angle and the length between them, and that angle and the length ends up being the same as human meaning, once you can, once you can represent it all in this object. So essentially, what we did is those 50,000 conversation transcripts. We chunked them up into smaller sections and then we used a method to convert those, the text in that sections, into a 1500, approximately 1500-dimensional geometric object. So every word has an angle and a distance between them. And when you represent that in, like in flatland, to a computer all it is is just numbers because it's just a length and angle. Length and angle. Length and angle. So a computer can understand 1500 dimensions super fast, whereas anything over four dimensions just boggles our own human mind, like we can't even imagine that right.

Speaker 2: 33:16

So each chunk of text from our call recordings is this dimensional object that we then store in a specialized database and you build an index on that. So now if I'm the technician and I type in what is fault code 37, it's going to take that text, it's going to turn that into the dimensional object and it's going to search the list, our database index for the object that is the closest fit to it, meaning, if you were to represent it on the coordinates, it's the closest to those chunks of text. That means that the closer the shape, the more similar the meaning. So I said fault code. But if one of these texts was talking about error code 37, fault and error are represented very close meaning. The meaning is the same. So then it takes the top end of those conversation chunks in which fault code or error code 37 or something like that was mentioned, takes that, converts it back into language like English and then it sends that to an LLM just to synthesize that into a proper answer and then that's what you get back as a result. So all this happens within a few seconds.

Speaker 2: 34:51

It's really combining that approach with the conventional LLM GPT approach where I think there was really the master mechanic breakthrough and then we did a number of passes through, like we rebuilt the code several times to optimize for this or that, and we had some of our own team of our top experts in HVAC kind of go through it, rank the answers, rate them, develop a way to ensure that it's getting better. So this is human feedback, like reinforcement learning, right, and then kind of updated the algorithms from there and from that point on it's been basically a lot of feature updates. So we gave it the ability to understand different languages. So if you type or speak Spanish into it, it will recognize it's Spanish, it will convert that to English. Then go find the corresponding conversation snippets, do the whole thing and then convert it back to Spanish or any other language that you talk into it. And yeah, and so that's the current state of master mechanic.

Speaker 1: 36:04

And master mechanics the name of the engine, so to speak, inside of the BlueON app right.

Speaker 3: 36:13

It's the AI that's there for specific questions.

Speaker 1: 36:18

OK, perfect. Ever in a like could you even imagine that this would have been a thing? Like both of you are like it's hard for my brain to even comprehend what you just said. Like I don't know if I want to comprehend exactly what you just said. I'm just super glad it works. Like that's really deep shit that you just explained. It's mind boggling.

Speaker 2: 36:51

Yeah, it's cool stuff, it's great stuff.

Speaker 3: 36:53

And, of course, explained to me the vectors based up. I mean, it makes sense, like inherently, like wow, that makes so much sense. You know, I think having a physics background helped, but to get there, yeah, that's the big question, like, who got there? Like that's an amazing idea that requires insane computing power to deal with, right, both to set and to analyze. But it makes total sense, right, when you everything has gotten back to a geometric comparison where you can get so much more nuance right in that, in that evaluation, which you just can't do right in a normal application. So it makes total sense. But man, yeah, I don't think I would have imagined that before I was told.

Speaker 1: 37:35

So where do you think? Where do you think you know? Where does this, you know? Where are we at the end of 2024 with the amount of speed that we can do things now, whether it's marketing stuff, or whether it's what you're talking about or really you could do if you have the right information, you could do just about anything 100 times faster than you could do it now. Where does the industry, as we are talking about HVAC or plumbing where, where does this lead the industry? Because it's I'm a little nervous about that. I'm a little bit nervous that we are. We are advancing so fast and let's just be honest, this is not a knock against the trades, but like technology is like 10 years ago, like that's where everybody's sitting, and this is like 20 or 30 years ahead. So how do we bridge that gap? And where do you see? Where do you see that going?

Speaker 3: 38:39

You want to jump in Okay. I got a couple of go Go.

Speaker 2: 38:41

Okay, okay, a couple of things to say there really quickly. One is for the HVAC industry, and for techs and contractors in particular. I think this is the golden age for them. As advanced as AI and robotics are going to become, it will never replace the guy who needs to get up on the roof, right. And so what you've got instead is this unbelievable tool that can help make your job easier and better and faster and less of a pain in the ass, right, that's great.

Speaker 2: 39:19

Other people in other professions, not so lucky. They got to scramble In terms of the industry being behind technologically. I mean, that's one thing that we're tackling at BlueOn. I think of it maybe, as, like you look at the continent of Africa, it kind of missed out on the last 100 years of tech development, and so it didn't lay copper phone cables, it didn't lay fiber optic cables, but then, once satellites were in place, they went straight to mobile phones, right. So there's some capabilities for the laggards to basically wait long enough to benefit from the new technology infrastructure without having to go through the innovators dilemma or whatever of adopting these types of things.

Speaker 3: 40:11

Yeah.

Speaker 3: 40:11

I think Go ahead, peter, sorry, go ahead I was going to say just real quick, because I think I talked about this a bunch, is HVAC. It goes to exactly what you were saying. Hvac is still stuck in the 90s in terms of how it trains technicians right, and still trying to get people to memorize stuff and to get experiential training which they can take to the field. And if you ever talk to a Gen Z, that's not how they do anything right, it's all indexed. Tell me where to find it, I don't want to know it. Right, and that approach really is synergistic with what we're talking about, right?

Speaker 3: 40:50

So now, with the AI function, you can be very much opportunistic in the just-in-time training you need at that moment, as long as you've got a good head on your shoulders to take advantage of it. And that meets them where they are, where they're now being sent to like 3A course, where I don't want to be here. This sucks, I don't want to memorize those crap. So I think we're allowing that 22-year-old kid to be effective without the requirements of the historical training apparatus or the requirement of a mentor, which is now gone in the industry for the most part. So the industry needs this more than ever because it's losing all of its wisdom, but it has no means of educating the new people because they're still doing it the wrong way. But this just-in-time method, I think, is the secret sauce to bring in a new wave of folks who can take advantage without having to go through that historical apparatus.

Speaker 1: 41:44

How are you educating? How are you getting this out? You have a significant amount of people on your platform, so you have the ability to have a lot of eyeballs on what you're doing, which means that comes with a lot of responsibility. That comes with-because really outside of you all and maybe I'm speaking at a turn here, I don't know anybody else that has a platform like you do that can affect the change like what we're talking about. So how are you guys making sure that you do affect the industry and get this education out? Yeah, how do you do that other than coming on my podcast?

Speaker 3: 42:26

No, it's a great question.

Speaker 3: 42:28

I mean, man, I think it's-well. Making sure that the information is right is mainly-that's a big one. We took that really seriously, probably over-serious, with our guys being incessant about checking it. Just as a quick data point, we got to 50,000 questions on the AI in like two months. It took five years to get there with text work calls. So people are way more adept at using it in terms of getting information. So that's because it doesn't have the-particularly with this generation.

Speaker 3: 43:02

It's truly anonymous. No one's judging you for asking a stupid question. There's no human being there to reap their judgment on you, which I think is a huge deal, particularly if you're a young kid and you don't want to look stupid. So it's a great way to get information without that impediment, and we're even building our tools of feedback to preserve that anonymity. So the tech can always be anonymous, but the information can be aggregated so that the training can be appropriate for their team. But that anonymous thing has to be kept sacred.

Speaker 3: 43:35

But it's a great question. I mean, besides making sure it's right, making sure we get feedback, listen carefully to what the community is saying and make sure we react appropriately. I've only thought about it in a deeper way of what can we be doing here to make sure it's going forward, except for just serving the need as you see it, in the best way you can that leads to the best answer, and respect the trade and the guys doing it, knowing that our main purpose is to make their lives better and their existence better, which is our primary mission. If we make money along the way, great and really that's what fuels everybody here is making those folks have a better way to get it done. Yeah, it's a great question. I haven't thought about it outside of those characteristics.

Speaker 1: 44:20

I think it's probably, I think it's vitally important to the success of the current companies that are out there, the ones that do embody this and embrace the technology.

Speaker 1: 44:36

Look, everybody I shouldn't say everybody a lot of people in business are, they're building and they're growing, and there's some big players out there, and even the big players don't even know what the hell we're talking about, and so, but when they find out, it's the likeliness of them being able to gobble up the five million dollar guy, the 10 million dollar guy, I mean, not to mention the one to two million dollar guy is going to be.

Speaker 1: 45:06

I mean, that's such a competitive advantage just knowing that you can train, that you can use this technology to train a person completely off the street, and they don't have to have an ounce of knowledge, they don't ever have to go to a class, they don't do anything and, quite frankly, they're probably going to get better information than they would if they did go to a class or did take an online course, or even read a book, for that matter. So you know, I think it's important that this education does get out to the masses. But you know, I don't know how, and I've wracked my brain about this question. I just don't know how to help people other than to keep talking about it, how to help people to understand the importance of this. I mean, it's here, it's not going anywhere and your business is going to rely on it, or you're going to be out of business period. I don't know how soon that's going to be, but it is coming.

Speaker 3: 46:10

Yeah, I agree, yeah, it's. I think Adam had a big point, which is that the industry is so technologically behind the curve, so to speak, that there is potentially a very nuanced answer here. We'll see how it shakes out. I mean, there's a lot of stories of you know, if you remember, there was like you know, we're going to do everything with VR and then Google Glasses, and none of that panned out because techs were like no, I'm not, I'm not wearing that, all right, so you have to meet them in a place that's not only beneficial but viable to how they operate. Right, you know, to us, the biggest thing we solved and the reason why the app took off is because the techs liked using it. Right, it was the first app that they self-adopted. No one told them to do it Like, they just did it. No service manager owners say, hey, go get that. They just did it because their buddy told them about it and it works well. Right, I think that's a text.

Speaker 3: 47:07

You know, they're ironically, they're not technologically looking for solutions. Right, they, they, they most, most, all the younger guys are. But, as for, the majority of the techs would rather not go there, but it's so. It's got to be really easy, very easy to use and has to pass their skepticism test pretty quickly. So it's weird dynamic because you have these young guys that are all for it, but then the majority of the guys that run the show you know not so much right. So there's a there's a big chasm to the bridge there, for sure.

Speaker 1: 47:38

It also solves. This also solves, in my opinion, or will solve, the employment drought right? I believe that this is the solution for the future trades, because right now I'm sure you all know this there's, I think, the people coming into trade is at 0%, or at least it was last year, I'm pretty sure. Dude, what happens in 10 years when you know your dad or your grandpa, or both, are retired or they can't crawl into their house, what happens? Like there's nobody coming in, so they've got? These owners have got to think about how do I innovate to make sure my legacy moves on or make sure my, my business continues? And without this, I don't know. I don't know what they're going to do. I really don't, yeah.

Speaker 3: 48:29

Yeah, I totally agree. That's what I was saying earlier. Like the industry just has not really just sunk into this new generation and how they deal with them right, and that's a big problem. You know, I talk about this all the time. It's like you've got to figure out. You cannot train them like you trained their father not going to work right. You will get bad output, right Period and so I think that you're right. This is the solution that not only gives them the information, but it does allow them to train themselves in a way that suits their learning. You know abilities right, they're learning capacity. So totally, totally agree. But our big journey now is to, like you mentioned, is to get the owners aware right To stop thinking about things like they used to and start thinking about things in a new capacity. Because, you're right, if you don't bring on new tax, you're going to die on the vine and there ain't, I guarantee the number of taxes shrinking, not growing, for the foreseeable future. That is going to be the case.

Speaker 1: 49:25

Right, I'll throw one more thing out there that I'm sure that you've probably recognized by now working with as many taxes as you have is most people in this industry, me included, have a form of ADHD, and so I believe you know we hated school. That's why most people are technicians. I'm not a technician, but I might as well be right. I hated school. I wasn't good in school, like it was the worst thing in the world. I had to do is go to school and sit still, and I've used this. Well, like I said earlier, I've used this to teach myself things that I would have never done. Right, and you mentioned we mentioned data earlier. Good data in good data out. If you know the right questions to ask and you ask it in a way, it will answer the question and it's pretty damn spot on. So is blue on available to anybody? Is it a private thing? I know you all have something with Sarah, if I'm not mistaken, a partnership of some sort. So, yeah, who has access to this?

Speaker 3: 50:34

So any tech in the world can open a blue on membership account and have access right now. Right, the access to the platform is currently unlimited, except for live tech expert, which you have to pay for. But and then, right now, the AI is open for everybody. Go nuts, now that's going to come. That's going to change. We get into Q1 where three members will have a finite access, like a handful of questions a month they can ask to it. And then the paid guys that are part of a contractor, who have access unlimited access will have a better service. Limited access, have live tech support, have the cross reference, but in terms of getting the basic app which gives you the models, the manuals, the part numbers, all that stuff, as well as the AI that's free of charge, remain, remain so, but it'll be limited. And how often you can use it if you're not part of a paid group? But everybody can use it in terms of accessing it right now.

Speaker 1: 51:34

And what do you want me to ask him? What's the difference? Obviously I understand free. What's the paid account? What does that look like for an individual? If I'm just a technician that wants to learn more, can I purchase that paid, or does it need to be through a company, or could it be both?

Speaker 3: 51:51

It's both. So the right now we call blue on for business is a entity, a company, coming on board and that's you pay by size. So if you're a small shop, you'll pay a couple hundred bucks a month. If you're, let's say, a 15 person shop, you'll pay like 600 bucks a month for that. Access gives you all your texts on limited access, gives the home office the database for cross reference, which is really valuable. And then if you're an individual starting in like March, we're going to open up. If your company won't go that way, as an individual you'll be able to do a monthly subscription for like 999 as an individual to get access on, you know, outside of a company perspective, In my opinion Get there.

Speaker 3: 52:35

But the, yeah, the big issue for us is, you know, we've been historically, you know, live text board is expensive and so we've been, you know, trying to have an alternative where we can still have unlimited access and then make sure that the, if you're going to go, go talk to one of our guys, which is still available, you got to pay because it's just, you know, costs us like on average, about 30 bucks a call just to finance it, you know, so we have to obviously monetize that, at least to a break, even on the text board. But but luckily, like we mentioned earlier, all that text board that we did offer for free for five years has allowed this AI to be born, which folks can take advantage of, you know, without having to pay the freight.

Speaker 1: 53:15

That's pretty astronomical. I mean that's, that's phenomenal, like that's really in for anybody, listed as this. Like, like I understand, like you still like running, like when we run some of our GPT's and whatnot, like there is a finite cost to it. But there could be an astronomical cost to it depending on and I may say this wrong, but how many? Maybe API calls that sound right, right, so how many so like so? So it's not like you all have. It's not like this is completely free. It's costing you money every time somebody uses it.

Speaker 3: 53:52

Yeah, oh, yeah, yeah, yeah, it does. It costs, it takes money, it takes time and effort to maintain it Right. So the maintenance is not not simple. I won't think we do so. You know the equipment is changing every day so the live text work calls are feeding it AI. So as things get progressed we are always keeping up to date by feeding it the most current information which you have to do. Otherwise you fall behind like quickly, particularly in the current, you know, evolution of products changing rapidly with CR2 and A2L is not like craziness going on. So you got to be current with your dataset, Otherwise AI is not going to be much good.

Speaker 1: 54:29

Right and you're updating it practically instantly.

Speaker 3: 54:33

Right, yeah, it comes from the source, right. And then I'm doing all kinds of other work by integrating other sets of data to augment even further along the line. So it's a never ending progress, right, you're just constantly adding to it to make sure you can do as much as you can. The key is to try to make it efficient and affordable, because knowing a lot of text work call, which is great, is just not affordable five times a day, right? So we gotta get a way to do the best we can in a way that's almost as good even not better, I would say in some cases to answer those questions in a way that's very fast and extremely practical, right, doesn't give you some BS, academic answer, but usable information.

Speaker 1: 55:13

Well, it's also not affordable for the company. That's got the guy waiting on the phone for tech support or on the phone right. I mean that costs the company money, so it's costing both ways huge.

Speaker 3: 55:26

But they don't recognize it. They don't recognize it. I mean, when you show people the mail on what the time saving is, they're like that can't be right, like it is right. They walk you through it Like holy shit, they didn't realize. You know, because you got a billable most guys are billing at 350, 450 billing and then they got their sub costs of like 100. Do the math. That's a double hit right. You got a one add one subtract. It's a big number really quickly.

Speaker 1: 55:49

Right, it's easier just not to look at for a lot of these guys. I mean, just be honest, I mean, but I'm doing it.

Speaker 3: 55:59

So it's your order, or they're calling another tech, which happens all the time. They're calling the senior tech, taking him out of the job. So you got two guys talking and that's a disaster. Right now they're both off the clock.

Speaker 1: 56:10

Yeah, 100%, and the customer's waiting Right.

Speaker 3: 56:13

Waiting. That was all day, everyday.

Speaker 1: 56:15

Yeah, 100%. So what is there anything other than the small price change coming out in 2024? Like, do you all have something? I mean it's not, like this is not big enough. If you said we're not doing anything else, you've already solved a million problems. But what else do we have to look forward to coming out, potentially in 2024, that maybe you guys haven't necessarily released yet?

Speaker 3: 56:42

I don't know if I should ask that, because I don't need to tell you, but maybe I did a one thing that it will be coming out in either late Q on early Q two, and so we talked about, you know, our ability to search any model, find the build materials, find the part number, find the compatible part number. You just pour that part aftermarket and OAM cross just massive. So now what will happen? Let's say, march time frames, you'll be able to click on any of those parts in the you know that are compatible and just buy it with one click right Through partners of ours that will be priced right there in real time. So you'll be able to have the ability to now, you know, figure out what you want and then just boom, take care of it right then and there with that interface which you know right now.

Speaker 3: 57:27

If you're, you know if you're a contract, you got to depend on your distributor to do that research for you. So now you do your own research. Takes you one second by the part. You want cheap, expensive, oem aftermarket. Your call one click, done right. So that's what we're working on also, which I think is a you know it's the first time that capability has been in the hands of the contractor ever and it's a. You know it's a big shift. Distributors aren't exactly happy about us moving it over there, but it's a it does offer. You know the buyer should have the data. Let's call it a day, right. The buyer needs the data, not the seller, and then it's been in the wrong hands for a long time.

Speaker 1: 58:05

Wow, that's pretty, that's pretty impressive. And the first thing I think of is, well, a lot of these technicians. You know they use having to run to the supply house or the parks house as a break, sometimes three times a day, right? So how you know, have you had to overcome the technicians being well, I guess it hadn't come out yet. So do you anticipate there being an aversion to that specific part, maybe because of the lack of better terms? Then burning up the clock, doing nothing?

Speaker 3: 58:39

You know I used to think that was a big problem. But as we've been speaking to, you know hundreds, if not thousands, of contractors. You know mostly guys that are, let's say, seven, eight trucks or bigger shop. Most of them not all of them, but most of them really don't allow or enable their techs to go to the supply house, right, they have a runner doing that. They keep them out of there, right, and so that's sort of a purposeful process. So there's really, and the small guys different, but in the bigger guys there's really not a conflict.

Speaker 3: 59:09

In the smaller shops there is a conflict for your purpose exactly, but a lot of times they're closer to the money and therefore they have a more direct relationship with the efficiency. So there's another play there. But you're totally right, there is a built-in bias to wanting to have that break. Now COVID broke that a bit, you know, broke that relationship a bit, but it's still there. I mean there's still, and there was distributors that are like putting out, like you know, lunch and learns and pizza parties and, which is insane to me, like you do not want your tech in the spiles, like it's a bad, bad thing, but yet come on in, you know it's a whole thing is backwards, but there's a big challenge there, for sure.

Speaker 1: 59:45

So if I were tech and I go on and I order the part and pay for it, hypothetically what happens next?

Speaker 3: 59:54

So either it would be delivered to you tomorrow morning or you'd go pick it up at will call Nice, or your runner would pick it up yeah.

Speaker 1: 1:00:03

Or you can have it delivered.

Speaker 3: 1:00:04

So partnership with Coury, which would actually bring that last mile delivery which you could track like you do on, you know, postmates or you know on Uber Eats, so you'd be able to pick it up and you could have it delivered right to the site if you wanted it, like within 35, 40 minutes from the when you picked it.

Speaker 1: 1:00:22

So I'm gonna give you the first objection that you're probably gonna get from this, because I and I know you could overcome it. Well, how much is that, Peter? How much does that go cost me, Are you? So how much now is my part gonna be if you do all that and we could just go to the supply house and pick it up ourselves? It's gonna have to be cheaper if we just go there to pick it up.

Speaker 3: 1:00:41

The part will be cheaper from us pretty much guaranteed, but the deliveries attack on so it's gonna cost you call it 25 bucks. Get that thing delivered, which is an add-on, so that would cost you. But you want to cost to save a tech an hour seems like the biggest no-brainer of all time.

Speaker 1: 1:00:59

Especially if you're paying him 50 bucks an hour.

Speaker 3: 1:01:01

I mean that's half the time right and you're not billing. That's it. I mean, you're definitely winning and it seems like most guys get that there's a desire to do it. There's just never been a system that is viable enough to deal with it right. And the bigger issue we have to overcome, which we've been noticing, is historically counterstaff. God love them, but they're an impediment because they're like a time delay so you can't do things that you can normally do electronically. You have to get around them with integration so you can not have this human being in between you and efficiency right. So that's kind of one of our big challenges is integrating directly around and not through the typical infrastructure which is human-dependent, with a counterstaff which I love those guys, but they're just inherently humans, like we all are, and they can't be an instant handoff right From demand to request. Yeah, 100%.

Speaker 3: 1:01:56

So world's changing.

Speaker 1: 1:01:58

The world's changing and either people can get on board or they cannot, and we'll see what happens, but I do. This has been such a I really didn't know. I didn't really anticipate this being such a great conversation. I didn't really know what to expect, which I guess that's good. This stuff's right up my alley. I love, love, love talking about this stuff and getting the word out about different AI stuff. It's just such an important thing. And I just want to say one last thing. People don't need to be scared about it. It's not anything to be scared about, right? I mean, there are always gonna be bad apples. There's always gonna be somebody that's gonna take AI and rob a bank with it or siphon money out of your grandmother, but guess what? Most people are not gonna do that. So just don't bring those people on your team.

Speaker 3: 1:02:49

Hey, man do that yeah.

Speaker 1: 1:02:51

So, yeah, guys, I really appreciate it when, if somebody's listening to this, they wanna find, they wanna get ahold of either one of you. The best way to do that would be how? And then how would they go and download the BlueON app, if that's what you want them to do?

Speaker 3: 1:03:06

Yeah, yeah, it is. Go into, whether it's the app store or the Play Store, just type in BlueON. There it is boom, download it. You'll go through a little onboarding process. We'll get a little information on you. That's it. You're in. You're ready to rock and roll. If you're a contractor and you wanna reach out to us, go to BlueONcom, go to Contractors. You can watch some videos there. You can also just hit a forum. We can set up a meeting, set up a presentation, or you can sign up right then and there, super easy, yeah, and you can contact us right through the website. No problem, we'll get back to you pretty quick from the forum there or the info directly on the website. So, yeah, reach out to us, we'll set you up. We can get guys into a presentation within a day or two and get you onboarding off to the races. But, yeah, text, get after it, get the app, check it out.

Speaker 1: 1:03:51

Perfect Adam.

Speaker 2: 1:03:55

And if you wanna try Master Mechanic, just go to BlueONai. That's B-L-U-O-Nai and you can give it a spin.

Speaker 3: 1:04:05

Yeah, you can check it there or on the app, you both can, or on the app. Our loud currently. Yeah, it's a good way to just go to the website. You can beat it up. That's what I was gonna say. Give it a beat it up, see if you can stump it. Good luck, love it.

Speaker 1: 1:04:19

Gentlemen, thank you very much. I appreciate your time.

Speaker 3: 1:04:22

Hey, thanks, corey.

Speaker 1: 1:04:23

My pleasure, you got it.

Speaker 3: 1:04:24

Cheers.

Speaker 1: 1:04:25

Thank you.