Nick Melnychuk: Welcome, welcome ladies and gentlemen to the next episode of the in production podcast. Today we're meeting amazing Brian. Please tell us more about yourself and why should people listen?
Brian: Yeah, well, I'm excited to talk about AI today. Thanks for having me. I have been coding ⁓ somehow involved in the realization of code ⁓ for years now and really since last I haven't written a line of code because the world has changed. I think that's exciting for us at the moment. So I think how we bring code to market fundamentally different and things we're doing today are at the front of where those are going with enterprises that we're working with. And we're really getting agents and coding into production, not just a beta and proof of concepts. ⁓ These things are live. And so I think there's
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: you know,
Nick Melnychuk: It's amazing. And we will be covering your consulting side of work and career later down the episode. But I would love to start with your ad tech component that you have founded. And then you jump into the consulting. ⁓ was the main learning that you can get in the company, ⁓ doing and implementing it, that ⁓ you never really learn in the consulting? Was there anything that was your asset that you grabbed while you're time building the business in terms of AI that you would never really get into in the consulting in your current position.
Brian: Yeah, I think a few things. I one thing as a co-founder and that's the second or third company that I've co-founded. so these themes kind of run through, but you begin to look at things, I think a little more through a vision of ROI and EBITDA. And if I finish this project, what is this going to mean to the exit value of this company? What does it mean to how we staff hire and work with people?
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: if this thing comes to fruition or it works. And so some of those, you know, second level metrics of things that possibly when you're consulting, you're spending your time trying to, of course, advise on how to do the right thing. Always usually probably thinking about ROI, but maybe not as much about how much value this adds to, know, my exit criteria. If I own this technology and it's really functioning like it is, is it going to really do these things for the evaluation of this company that I really want it when it's done. I think those are hard things to grasp sometimes when you're consulting. And then also just the constraint of resources. When you're founding a company or trying to do a startup and bring it to market, some of the ingenuity and creativity is because you don't have the money. ⁓ so you spend a lot of time thinking of these ways to change
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: you accomplish the same thing, but for half the cost or a quarter of the cost or less. And how do I do that in days or weeks, but not months or years, because you don't have months or years. And so compressing these timeframes. And I think one of the biggest challenges to consulting is understanding that these entities are different. And so for me, into it that way, they need longer timeframes. They're not going to move at the same pace.
Nick Melnychuk: Mm-hmm. Mm-hmm. Mm-hmm.
Brian: so I think it's been two ways, right? I'm learning some new things and bringing some things that I've learned that help out.
Nick Melnychuk: Yeah, that's very cool. It's always interesting to see when you don't have a certain resource, how you turn curiosity or boldness or audacity into being that kind of resource that will help you maybe even compensate instead of having the deep pockets. By the way, I was reading the news that the team that you're currently with with Merlaps just became one of the Anthropics first cloud partner network members. that you were at the Partner Summit. I would love to use this opportunity to kind of pick your brain or just see what you've heard ⁓ you're talking with people that are building those frontier models. What are the conversations behind the closed doors that enterprise clients, that businesses don't usually hear of?
Brian: Yeah, that was an amazing opportunity. I one of the things I took away the most is that this company, Anthropic, is growing so fast and it's such a crazy pace that it seems like it's even them some growth pain. Everything is about speed, right? It's more of a get really close, but we have to be really fast. And I that causes a certain number of challenges, but just the pace at which they
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: are moving things forward at the scale that they're at. It is really just, you know, on some level terrifying, right? There's still cases where they're on X and they're saying, look, we see someone recommend a feature on X and few hours later, that feature is live in the product it exists ⁓ Claude code is doing the coding for them. And so they're fully committed to the idea that Claude code is writing their code. and everyone's gonna do it that way and that this flywheel is just happening and that all these things around the model, like you just said, we think about them as a frontier model company, but I think one of the insights I got was like, that's part of it, but that may not even be the main part of it anymore. This infrastructure around that model.
Nick Melnychuk: Mm-hmm. Mm-hmm. Mm-hmm.
Brian: has so much more focus and attention of what they need to do way more than what I was giving it credit for or the way I was thinking it through. These extra tools are building it. And of course, they let us see some demos tools that aren't even out yet that are coming down the road.
Nick Melnychuk: Mm-hmm.
Brian: And those tools, you can just see that all the infrastructure and tooling around these models is just growing so fast and taking up so much time and energy. And then also, because they're using their own tools to do all their work, to build their tools, it's just a flywheel effect. It's just building on itself. And you get to really see those demos and hear from the people who are working on those things firsthand. And the pace was just
Nick Melnychuk: It is. It's a very cool. ⁓ Would you recommend to the people that are listening that are technical or who want to become technical? Because, mean, ⁓ are at the stage of the career where you don't have to code, right? But you still do it. even you're not, as you mentioned, you're not, since December, you're not writing a line of code because AI does it, but still you are vibe coding, you have the good understanding, you're building different POCs to support sales. I mean, within your team, within the teams of your clients. I mean, that's really unusual. you, ⁓ is the benefit that you have at this stage of the career when you're still hands on on everything? What's the benefit of doing it? And would you recommend the folks that want to do it? ⁓ should they start?
Brian: it's going to be unavoidable that everyone needs to embrace it on some level every day. They may not be writing code every day, but the amount of productivity that you can increase no matter what your business line is, no matter what your function is day to day, just getting exponentially out of control. So ⁓ last night, for example, I'm sitting on the couch with my dog. watching a show.
Nick Melnychuk: Mm-hmm.
Brian: I'm watching actually YouTube and the new model is out, right? Gemma four. And I'm like, that looks fascinating. And so what do I do? I pick up my phone. It's connected to clog code dispatch, right? Claw dispatch, right? So I just on my phone, I asked it to install that remotely on machine upstairs. And we go through a little bit of process. I just keep watching and I've got that up and running and ready to test. And I'm using it to test new features.
Nick Melnychuk: Mm-hmm. Mm-hmm. Yeah, Mm-hmm.
Brian: in an application that I was working on today, but I did that for my couch. I would have had to go sit somewhere and do that work before in a way that would have taken a lot more time. And now I just delegate that work and it actually works. comes back working. And for people who are writing code, they're able now to have the accessibility to create new tools that can even be throwaway tools. Like you can use some of these tools one time to
Nick Melnychuk: Mm-hmm.
Brian: Maybe map data from A to B just to do something for you or write hooks to Slack that automatically let people create tickets from Slack and into Jira automatically for them and just queue up these processes and just start removing friction across the board in ways that they know in their domain. But I might not know from where I sit as a developer. So putting these tools in their hands, right, is just enabling them
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: to create all these efficiencies that they really was not available to them before. They would have had to put in a queue somewhere, wait for someone to do this work and maybe get back to them later. Or it would just be cost prohibitive. It just would be impossible. You just wouldn't do it at all. And so I think on some level, getting into these tools doesn't necessarily always mean writing a full application code. It doesn't mean you gotta be making a new mobile app. But people are using these tools and ways to automate their marketing efforts, to work on different content. All these things are coming to fruition and it takes that light bulb moment where you realize how that productivity is going to switch over. then the flywheel just goes and you just start realizing all these things that you can do that were just never possible before.
Nick Melnychuk: Mm-hmm. Yeah. Yeah, it's very cool when you a of people are being stopped at the personal level in implementing this thing because they just simply don't know that they can do those. And for them, my like my simplest advice is just go to cloud and start the chat and then ask it and ⁓ what you do or maybe just use your history of what things that you've been Googling with them. It knows a lot about you. So just ask them, Hey, I'm doing this and this. This is my project is my priorities. Like, how can I make my life easier? Today, I learned about like called scheduled tasking cowork that it can perform something and like, damn, wow, this is amazing. Like how can I do it? But this is all nice like on the personal level. I'm curious about ⁓ company-wide initiatives. Whenever teams are launching AI pilots, you've seen a few of them, a bunch of them. ⁓ And most of them, like at least from what I see from the ⁓ ⁓
Brian: you
Nick Melnychuk: it's very hard to run a successful AI pilot that then will be scaled to the whole organization. So in your experience, if you talk to someone and you see that the pilot failed, what is the single most common reason that it failed? What are the things that folks are just not noticing in the early stages?
Brian: Yeah, is quite a few things, quite frankly. So our projects, by the way, the ones that I'm working on, we're going live. So so these things go live. ⁓ that they're not always the same project that they originally had wanted, or it's not approaching it quite the same way they wanted to approach it. So I there's a few things that stop some of these from going ⁓ One is think everybody thinks it's free. And then they find out there's a bill and there's a bill to scale it.
Nick Melnychuk: Mm-hmm.
Brian: And so some of those economics haven't really been thought out upfront. Like they thought that we're to do a POC. They thought it might be cheap. It was hard to figure out how much it might cost. And then you look at it and you go, well, this is actually more expensive than what we used to do. Like this is not worth it. So that definitely, that definitely happens. think people get in and find out that the data that they needed to feed these applications just isn't in the state. needs to be.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: And they don't realize how much the human was actually, you know, overcoming these problems. And so this human was sitting there like toiling through this bad data or this not quite perfect data. And the AI, just ends up making mistakes or delivering poorly. And too much expectation that this AI is somehow ⁓ magically your bad data. And ⁓ that spoils a lot of projects, obviously.
Nick Melnychuk: Mm-hmm.
Brian: Right. It just goes downhill from there. So, you know, we try to talk to like your data just has to be reliable for making these decisions or you have to build into your process. And so, you know, one more thing I think people really don't ever do is realize how many errors the humans were making. Right. So the AI makes errors. The human makes errors and everybody keeps like approaching this like the AI is a calculator. It's not a calculator.
Nick Melnychuk: 100%.
Brian: It's like a mini human with a college degree that hasn't done anything yet and can make a lot of mistakes. so enterprises have process in place to deal with these error rates. But somehow they think when they use AI, they just can throw that away and it'll just be perfect. Like this, whole paradigm. They hope for the magic bullet. It's like, that's not going to work. And if the error rate on the AI is 5%, but the human was 10 and it was cheaper, like that's an improvement.
Nick Melnychuk: Peace. Yeah, yeah, they hope for the magic books.
Brian: You did good. know, put in a process to deal with the 5 % and move forward. And I think they just really struggle with the idea that the AI is not a calculator and that it makes mistakes and that's okay.
Nick Melnychuk: Well, that's a lot of failing hopes and becomes especially like my day to day in my experience because we work with the highly regulated industries ⁓ you the same. You work with healthcare claims, with financial services, infrastructure, with manufacturing, with government. In your which industry has the biggest gap between executives are overexcited. They're like, ⁓ yeah, we're so pumped. We need to have this thing like yesterday, but they actually do not realize that their data or infrastructure or like that their homework is not done. Yes, the promise is there for the I and the eyes like can be really revolutionary for their business. But gap between where they think they are and where they should be is too big. So in your experience, which industry maybe it's not even ⁓ of those four that I mentioned to you. What is the industry where it's like the hardest way to get AI to the level where it needs to be? So leadership just simply don't think about it.
Brian: No, well, things I think are ⁓ in some of these industries, ⁓ like the finance sector, we with some large financial institutions and they were cutting edge with machine learning before there was AI. And so they're kind of funded and ready to go. They know what they're, really, they're into it. It's important to their business. Then when you shift to like manufacturing, the margins such that these look like large enough expenditures. you know, that it gets really challenging to figure out how to move forward. that's kind of like a different problem on sort of like, how can I fund this? How impactful is it really going to be? And more slow to looking for solutions that are are canned or proven, right? But then I think to answer your question, the biggest opportunity is really in health care. If you really look at what can be done with AI in health care, it's just it's ⁓ massive. What be done?
Nick Melnychuk: Mm-hmm.
Brian: but the resistance to it is majorly high, right? It's just overwhelming. And part of its regulation, part of its lives are at risk, right? We get that. the ⁓ to improve healthcare, improve lives, have AI, have an extremely positive impact on outcomes, ⁓ just on a high level on a system that, you know, an opportunity for these extra costs to really be taken in.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: be taken out of it. I do think over time we're going to see massive improvements there. But you know if you were starting today and trying to do something and you were trying to get into that industry especially with a startup or consulting or something these are you know really challenging environments that are adverse to the change even if individuals, individual doctors maybe or individual administrators want to help advocate these changes. As an organization, the organization is built to resist that change and sort of defend the status quo and patient safety. And we get that. But on some level, we to work to push these things forward and help. so maybe you have to start at an administrative level. How do we improve the ⁓ paperwork of this process in a way ⁓ that with the administration side so that the system can still free up cash flow?
Nick Melnychuk: Mm-hmm.
Brian: to maybe invest and try things in different areas. How do we do pilots in a way that helps with that risk aversion, but still move things forward? And it's just very different environment that I think ⁓ is extremely challenging, but has so many opportunities.
Nick Melnychuk: I'm curious here. So do you think that in this industry, so within in this environments, will this change and the slow adaptation towards AI? Will it come from new teams from the market like from startups like ours? Or will it be something that the company will ⁓ test and deploy internally?
Brian: amazing. You know, I get a lot ⁓ customers or prospects right now that we're talking to that are really ⁓ us to come to them with an actual solution. So think... ⁓
Nick Melnychuk: like, like the one day that you can build it for them or why don't you come in and you say buy this, this, this and it's going to work.
Brian: Not that you can build it, but like you already, exactly, like you've either pre-packaged these things you would buy, or you have already started to build this thing that would nearly be done for them, and you know how to do this part of their business as well as they do, and you've already thought about this solution, and you're down the road on it. And so I think these companies that have this vertical knowledge, they are going to be able to work into companies and succeed. On the flip side, think companies that have their own proprietary data that no one else has, they're going to have to in some way, they're going to be developing their own solutions and they're going to be extremely successful long term. Right? Like I worked with a company that does have its own data. We helped AI enable some of this data and with seven or eight days of demos, they sold 800,000 in new revenue.
Nick Melnychuk: Mm-hmm. Nice.
Brian: explosive like that changes a small company and how they work and how they go to market but if you didn't have this proprietary piece
Nick Melnychuk: It is. Yeah.
Brian: Right? No one else could sell them a canned solution because no one has the data and the thing they have. they had to bring to market. And when they did, it's very good status for them. I think is sort of going to be the challenge. It's ⁓ which things my company buy and which things should they build? And I just think that equation has changed. I think that dynamic has always been there. We've always talked about that for the last 30 years and we had shifted, I think, a long way towards SaaS.
Nick Melnychuk: Mm-hmm.
Brian: And I think with the ability of claw code to really just recreate these very specific solutions to your industry, think some of that will shift back over time. But I don't think, you know, I'm not, not a, you know, a sass doomer or anything. I don't think that the big banks are leaving Salesforce anytime soon. Like they're there, but we've got to which ones. And I think that dynamic is changing dramatically towards
Nick Melnychuk: Mm-hmm. Mm-hmm. Just do it. Yeah. Mm-hmm. Yeah, me neither.
Brian: of doing it in-house again and being able to.
Nick Melnychuk: Okay, a follow up question here. So it still doesn't matter. Like the problem is gonna be the same whether the company decides to build something on their own or they decide to buy somebody. But whenever you, and you mentioned it already, like whenever you ⁓ into the process, there is the process and then there is the way that humans are actually doing the work. And sometimes those are two friends. What did learn about... how humans actually do the work, do they do exceptions, how ⁓ they find creatively workarounds, how do they create those workarounds, what is your experience around that?
Brian: On some level, right? It's frightening how little the business knows how the business operates sometimes Right. It's fascinating. I think we've seen some workforces where let's say you have a few hundred people doing a job That's done by humans
Nick Melnychuk: Mm-hmm.
Brian: there is, like you said, the sort of documentation that is out of date. Like you can't look at the documentation and get the answer. And if you talk to the human, they'll say, well, this data really means like this. for this exit criteria to be exist properly the right way, you actually do it this way, even though normally you would do it that way. And so the happy path is kind of where the corporate documentation tends to be. And then it's sort of like these exception cases, are completely undocumented.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: going to automate that, that needs to actually work over time, right? Like we actually need to know what those are. Um, I think that represents a huge challenge when you have just masses of people who sort of the humans are learning over time from other people or experience, they're making some failures, they're learning to do it their own way. Kind of exactly tribal. And they're not going to turn that back into some document, you know, somewhere. So, so we've actually implemented some projects where the AI is basically watching the logs of what the humans are doing.
Nick Melnychuk: the tribal knowledge, we.
Brian: and that we record that and we go through with AI machine learning to say, hey, in the future, maybe this should be something, a piece of knowledge that we've learned.
Nick Melnychuk: ⁓ huh.
Brian: now that we know this exists. And then the system can actually apply that quote unquote, let's call it an edit for the moment in a future iteration and then still put it in front of a human. know, humans always in the loop in these spaces, they'll put it back in front of the human and it'll just already be edited. And it says, Hey, this is why we edit it kind of thing. And now the human, that human may not have ever learned that lesson yet, but they're getting the edit and they can choose to accept it or not.
Nick Melnychuk: Mm-hmm.
Brian: But these types of things, we've really had to think through, sort of surprising. We sort of thought some of these things, you know, would be well documented. But if they are well documented, the tribal knowledge is ⁓ really and it's hard to learn. ⁓
Nick Melnychuk: Yeah, I can tell you a story later about the company that is building this way to observe what the employees are doing and then build the map out of it. But I'm also curious about the companies, say mid-market companies that ran a successful AI pilot. What's usually the biggest blocker then? So for example, the company did the pilot, everybody's happy, it's successful. It ⁓ ⁓ it had to prove. What is usually the moment where the whole program goes into the garbage? Or is ⁓ there anything that, or like... Generally, if the AI pilot is proved the program to be working, going to be implemented like 100 % of times.
Brian: no, I think if get tripped up, it's always like we'll a PLC and that's great. The PLC shows that this would work, but then there's some sort of security check regulation review. Can we use that data? Is there bias? Is this allowed? These types of things that can put sort of some larger delays into things, but I think
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: If the POC is successful, right, people try to move to roll out. I think those rollouts take longer than people expect is the challenge, but like we're getting through that process. But, but what type of infrastructure is actually needed to go live? ⁓ it's much like when we were kind of rolling out.com and everyone thought.com was free and it's just some web servers somewhere. And they're like, why is this bill so big? It's like, No, no, there's like servers and they have to do different things and there's a whole deployment pipeline that has to be built that makes this work. And it's just, think that that effort to deploy is a lot longer than people expect. And you have to be prepared to go through all those hurdles, right? Everybody wants to release something secure. So we've to make sure we go through the security review. Everybody wants it to scale. So how are we going to do it? Pilot. And then there's going to be problems.
Nick Melnychuk: Mm-hmm.
Brian: And so things get stopped at a whole bunch of blocks. Like the first time a person gets an answer, probably an executive, they didn't think should have been answered that way. It's like, do we freeze everything? Do we turn it off? Or can work through why wasn't accurate? How do we fix it ⁓ and make this fast? ⁓ has turn around fast, but it has to be accurate again. And can we do that? Can we respond in that way? But I think those things slow.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: at least for us, our customers down, but haven't stopped from going live, which is good. But I can see how you can easily get tripped up ⁓ decide to give up.
Nick Melnychuk: Mm-hmm. Okay. So it's very interesting. by the end of the episode, I left the interesting question for the desert. So listeners and viewers, live with a very interesting insight. You in the beginning that you have been ⁓ actively and building for essentially years. That's a lot of time. ⁓ And that time,
Brian: You
Nick Melnychuk: you've seen a bunch of hype cycles and those hype cycles they come and then they dissolve, they disappear. within like the hype is not ⁓ usually not built from nothing, especially right now in AI there is a lot of hype, there is a lot of noise, but it's because there is also a lot of signal ⁓ in what's how the world is changing. So I'm curious in ⁓ your experience working within the hype cycles, and then digesting what is signal and what is not signal. Because for example, the company can implement the AI in the sense of, it just summarizes me my morning emails. That's nice. But it's not the experience like you mentioned about that company where you implemented AI and they made like 700k in in terms of sales. ⁓ that the company was changed by AI. So how do you understand from news or just from any other indicators that the company is actually implementing AI and AI is going to change this company. It's not just like to sprinkle on top and say, we're AI.
Brian: ⁓ anything someone puts out in the news I assume is a lie. ⁓ just start there. So everything ⁓ that regard is why I have to stay hands on. It's why I had to get Gemma installed last night ⁓ and to run OCR examples on it this morning. Can you really do that on a Mac mini? Turns out you can. Wow, okay, pretty good. ⁓
Nick Melnychuk: Okay. Mm-hmm.
Brian: But I need to know that it's real. And when I look at the companies and what they put out, you can see that a lot of them are putting out information that say, you know, we're doing AI, where we've implemented this very well, but it can't quite be at the level they're saying, right? Like it has to be some height, because if it were at the level they're saying, you would see it in the financials in a quarter, right? You might not see it right away. And I get that a lot of stuff is new, but you would start to see. know, increases in gross margin, you would start to see these companies talk about these massive productivity games that would either come in new sales, like the company we just talked about, you know, we exploded the sales growth and that would show up in the financials, ⁓ or saved a whole bunch of money in cost and that would show up in the financials. And so ⁓ I what we're seeing right now is a of investment being made, a lot of people ⁓ still rolling out.
Nick Melnychuk: Mm-hmm.
Brian: partially, but you know, we haven't quite figured out how to optimize what we're doing and I don't think it's that much different than when the internet came out, right? When the internet came out everybody sort of like put up a website and they put up a store and they tried to do some things and some of them worked and right some of them didn't and we did, you know, reverse logistics bidding and all these things to try to understand like which thing is gonna work for my business so I still think in this part of hype cycle, that's where we're at.
Nick Melnychuk: Mm-hmm. Mm-hmm. Mm-hmm.
Brian: these companies with all these announcements, they have a whole bunch of people. And if you can imagine them just like running all over the place. And so now they're running all over the place really fast. It's not like we're going somewhere. know, that vision hasn't sort of coalesced to where 10 % productivity in every employee we have turns into 10 % productivity for the company yet, because it's too scattered, too not focused just yet. But that will come with time as these
Nick Melnychuk: Mm-hmm.
Brian: everybody sort of gets used to the tool and ⁓ to go sort of in the same corporate direction as they sort of start to understand, how I bring these things to market in a way that, you know, what person A is doing and person B is doing isn't conflicting anymore. Like it's building on each other. You know, think we're going to see those flywheels start to happen ⁓ and then going to see the news, but you're going to see it in the financials. You're going to see it when they do their quarterly announcement. You're going to see.
Nick Melnychuk: Mm-hmm.
Brian: know, the productivity or you're or you're or you're not right. It didn't work, but it most likely it's going to go up once these individuals get aligned. Like we're just seeing individuals hit these individual flywheels of productivity where they're just pumping out ⁓ much more than they could ever do. Right. People having like AI log into JIRA and bring all their tickets together and start to format them in ways that are ready for coders in different ways that would have taken them hours and hours of going through all these submissions and AI prepping all that work so they can put it out the next day. Weeks of work being done overnight. And that doesn't change the rest of the company's delivery cycle yet, but as it hits the rest of the company, it will.
Nick Melnychuk: Well, this is amazing. It was a great conversation and thank you so much for sharing the best of the best insights from your day to day with us and with the viewers. I'm sure you're going to have a few questions coming from the audience, so ⁓ I'm to leave your contact. If folks want to reach out or they're to be interested in the case that you described with that explosive revenue. going to be a very interesting chat for you.
Brian: and I'm happy to ⁓ up and have conversations. And I'm glad that we had the opportunity to talk today. I really appreciate it.
Nick Melnychuk: It has been a pleasure. Thank you so much for coming on the podcast and thank you so much dear listeners and viewers for attending.
Brian: Yeah, well, I'm excited to talk about AI today. Thanks for having me. I have been coding ⁓ somehow involved in the realization of code ⁓ for years now and really since last I haven't written a line of code because the world has changed. I think that's exciting for us at the moment. So I think how we bring code to market fundamentally different and things we're doing today are at the front of where those are going with enterprises that we're working with. And we're really getting agents and coding into production, not just a beta and proof of concepts. ⁓ These things are live. And so I think there's
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: you know,
Nick Melnychuk: It's amazing. And we will be covering your consulting side of work and career later down the episode. But I would love to start with your ad tech component that you have founded. And then you jump into the consulting. ⁓ was the main learning that you can get in the company, ⁓ doing and implementing it, that ⁓ you never really learn in the consulting? Was there anything that was your asset that you grabbed while you're time building the business in terms of AI that you would never really get into in the consulting in your current position.
Brian: Yeah, I think a few things. I one thing as a co-founder and that's the second or third company that I've co-founded. so these themes kind of run through, but you begin to look at things, I think a little more through a vision of ROI and EBITDA. And if I finish this project, what is this going to mean to the exit value of this company? What does it mean to how we staff hire and work with people?
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: if this thing comes to fruition or it works. And so some of those, you know, second level metrics of things that possibly when you're consulting, you're spending your time trying to, of course, advise on how to do the right thing. Always usually probably thinking about ROI, but maybe not as much about how much value this adds to, know, my exit criteria. If I own this technology and it's really functioning like it is, is it going to really do these things for the evaluation of this company that I really want it when it's done. I think those are hard things to grasp sometimes when you're consulting. And then also just the constraint of resources. When you're founding a company or trying to do a startup and bring it to market, some of the ingenuity and creativity is because you don't have the money. ⁓ so you spend a lot of time thinking of these ways to change
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: you accomplish the same thing, but for half the cost or a quarter of the cost or less. And how do I do that in days or weeks, but not months or years, because you don't have months or years. And so compressing these timeframes. And I think one of the biggest challenges to consulting is understanding that these entities are different. And so for me, into it that way, they need longer timeframes. They're not going to move at the same pace.
Nick Melnychuk: Mm-hmm. Mm-hmm. Mm-hmm.
Brian: so I think it's been two ways, right? I'm learning some new things and bringing some things that I've learned that help out.
Nick Melnychuk: Yeah, that's very cool. It's always interesting to see when you don't have a certain resource, how you turn curiosity or boldness or audacity into being that kind of resource that will help you maybe even compensate instead of having the deep pockets. By the way, I was reading the news that the team that you're currently with with Merlaps just became one of the Anthropics first cloud partner network members. that you were at the Partner Summit. I would love to use this opportunity to kind of pick your brain or just see what you've heard ⁓ you're talking with people that are building those frontier models. What are the conversations behind the closed doors that enterprise clients, that businesses don't usually hear of?
Brian: Yeah, that was an amazing opportunity. I one of the things I took away the most is that this company, Anthropic, is growing so fast and it's such a crazy pace that it seems like it's even them some growth pain. Everything is about speed, right? It's more of a get really close, but we have to be really fast. And I that causes a certain number of challenges, but just the pace at which they
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: are moving things forward at the scale that they're at. It is really just, you know, on some level terrifying, right? There's still cases where they're on X and they're saying, look, we see someone recommend a feature on X and few hours later, that feature is live in the product it exists ⁓ Claude code is doing the coding for them. And so they're fully committed to the idea that Claude code is writing their code. and everyone's gonna do it that way and that this flywheel is just happening and that all these things around the model, like you just said, we think about them as a frontier model company, but I think one of the insights I got was like, that's part of it, but that may not even be the main part of it anymore. This infrastructure around that model.
Nick Melnychuk: Mm-hmm. Mm-hmm. Mm-hmm.
Brian: has so much more focus and attention of what they need to do way more than what I was giving it credit for or the way I was thinking it through. These extra tools are building it. And of course, they let us see some demos tools that aren't even out yet that are coming down the road.
Nick Melnychuk: Mm-hmm.
Brian: And those tools, you can just see that all the infrastructure and tooling around these models is just growing so fast and taking up so much time and energy. And then also, because they're using their own tools to do all their work, to build their tools, it's just a flywheel effect. It's just building on itself. And you get to really see those demos and hear from the people who are working on those things firsthand. And the pace was just
Nick Melnychuk: It is. It's a very cool. ⁓ Would you recommend to the people that are listening that are technical or who want to become technical? Because, mean, ⁓ are at the stage of the career where you don't have to code, right? But you still do it. even you're not, as you mentioned, you're not, since December, you're not writing a line of code because AI does it, but still you are vibe coding, you have the good understanding, you're building different POCs to support sales. I mean, within your team, within the teams of your clients. I mean, that's really unusual. you, ⁓ is the benefit that you have at this stage of the career when you're still hands on on everything? What's the benefit of doing it? And would you recommend the folks that want to do it? ⁓ should they start?
Brian: it's going to be unavoidable that everyone needs to embrace it on some level every day. They may not be writing code every day, but the amount of productivity that you can increase no matter what your business line is, no matter what your function is day to day, just getting exponentially out of control. So ⁓ last night, for example, I'm sitting on the couch with my dog. watching a show.
Nick Melnychuk: Mm-hmm.
Brian: I'm watching actually YouTube and the new model is out, right? Gemma four. And I'm like, that looks fascinating. And so what do I do? I pick up my phone. It's connected to clog code dispatch, right? Claw dispatch, right? So I just on my phone, I asked it to install that remotely on machine upstairs. And we go through a little bit of process. I just keep watching and I've got that up and running and ready to test. And I'm using it to test new features.
Nick Melnychuk: Mm-hmm. Mm-hmm. Yeah, Mm-hmm.
Brian: in an application that I was working on today, but I did that for my couch. I would have had to go sit somewhere and do that work before in a way that would have taken a lot more time. And now I just delegate that work and it actually works. comes back working. And for people who are writing code, they're able now to have the accessibility to create new tools that can even be throwaway tools. Like you can use some of these tools one time to
Nick Melnychuk: Mm-hmm.
Brian: Maybe map data from A to B just to do something for you or write hooks to Slack that automatically let people create tickets from Slack and into Jira automatically for them and just queue up these processes and just start removing friction across the board in ways that they know in their domain. But I might not know from where I sit as a developer. So putting these tools in their hands, right, is just enabling them
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: to create all these efficiencies that they really was not available to them before. They would have had to put in a queue somewhere, wait for someone to do this work and maybe get back to them later. Or it would just be cost prohibitive. It just would be impossible. You just wouldn't do it at all. And so I think on some level, getting into these tools doesn't necessarily always mean writing a full application code. It doesn't mean you gotta be making a new mobile app. But people are using these tools and ways to automate their marketing efforts, to work on different content. All these things are coming to fruition and it takes that light bulb moment where you realize how that productivity is going to switch over. then the flywheel just goes and you just start realizing all these things that you can do that were just never possible before.
Nick Melnychuk: Mm-hmm. Yeah. Yeah, it's very cool when you a of people are being stopped at the personal level in implementing this thing because they just simply don't know that they can do those. And for them, my like my simplest advice is just go to cloud and start the chat and then ask it and ⁓ what you do or maybe just use your history of what things that you've been Googling with them. It knows a lot about you. So just ask them, Hey, I'm doing this and this. This is my project is my priorities. Like, how can I make my life easier? Today, I learned about like called scheduled tasking cowork that it can perform something and like, damn, wow, this is amazing. Like how can I do it? But this is all nice like on the personal level. I'm curious about ⁓ company-wide initiatives. Whenever teams are launching AI pilots, you've seen a few of them, a bunch of them. ⁓ And most of them, like at least from what I see from the ⁓ ⁓
Brian: you
Nick Melnychuk: it's very hard to run a successful AI pilot that then will be scaled to the whole organization. So in your experience, if you talk to someone and you see that the pilot failed, what is the single most common reason that it failed? What are the things that folks are just not noticing in the early stages?
Brian: Yeah, is quite a few things, quite frankly. So our projects, by the way, the ones that I'm working on, we're going live. So so these things go live. ⁓ that they're not always the same project that they originally had wanted, or it's not approaching it quite the same way they wanted to approach it. So I there's a few things that stop some of these from going ⁓ One is think everybody thinks it's free. And then they find out there's a bill and there's a bill to scale it.
Nick Melnychuk: Mm-hmm.
Brian: And so some of those economics haven't really been thought out upfront. Like they thought that we're to do a POC. They thought it might be cheap. It was hard to figure out how much it might cost. And then you look at it and you go, well, this is actually more expensive than what we used to do. Like this is not worth it. So that definitely, that definitely happens. think people get in and find out that the data that they needed to feed these applications just isn't in the state. needs to be.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: And they don't realize how much the human was actually, you know, overcoming these problems. And so this human was sitting there like toiling through this bad data or this not quite perfect data. And the AI, just ends up making mistakes or delivering poorly. And too much expectation that this AI is somehow ⁓ magically your bad data. And ⁓ that spoils a lot of projects, obviously.
Nick Melnychuk: Mm-hmm.
Brian: Right. It just goes downhill from there. So, you know, we try to talk to like your data just has to be reliable for making these decisions or you have to build into your process. And so, you know, one more thing I think people really don't ever do is realize how many errors the humans were making. Right. So the AI makes errors. The human makes errors and everybody keeps like approaching this like the AI is a calculator. It's not a calculator.
Nick Melnychuk: 100%.
Brian: It's like a mini human with a college degree that hasn't done anything yet and can make a lot of mistakes. so enterprises have process in place to deal with these error rates. But somehow they think when they use AI, they just can throw that away and it'll just be perfect. Like this, whole paradigm. They hope for the magic bullet. It's like, that's not going to work. And if the error rate on the AI is 5%, but the human was 10 and it was cheaper, like that's an improvement.
Nick Melnychuk: Peace. Yeah, yeah, they hope for the magic books.
Brian: You did good. know, put in a process to deal with the 5 % and move forward. And I think they just really struggle with the idea that the AI is not a calculator and that it makes mistakes and that's okay.
Nick Melnychuk: Well, that's a lot of failing hopes and becomes especially like my day to day in my experience because we work with the highly regulated industries ⁓ you the same. You work with healthcare claims, with financial services, infrastructure, with manufacturing, with government. In your which industry has the biggest gap between executives are overexcited. They're like, ⁓ yeah, we're so pumped. We need to have this thing like yesterday, but they actually do not realize that their data or infrastructure or like that their homework is not done. Yes, the promise is there for the I and the eyes like can be really revolutionary for their business. But gap between where they think they are and where they should be is too big. So in your experience, which industry maybe it's not even ⁓ of those four that I mentioned to you. What is the industry where it's like the hardest way to get AI to the level where it needs to be? So leadership just simply don't think about it.
Brian: No, well, things I think are ⁓ in some of these industries, ⁓ like the finance sector, we with some large financial institutions and they were cutting edge with machine learning before there was AI. And so they're kind of funded and ready to go. They know what they're, really, they're into it. It's important to their business. Then when you shift to like manufacturing, the margins such that these look like large enough expenditures. you know, that it gets really challenging to figure out how to move forward. that's kind of like a different problem on sort of like, how can I fund this? How impactful is it really going to be? And more slow to looking for solutions that are are canned or proven, right? But then I think to answer your question, the biggest opportunity is really in health care. If you really look at what can be done with AI in health care, it's just it's ⁓ massive. What be done?
Nick Melnychuk: Mm-hmm.
Brian: but the resistance to it is majorly high, right? It's just overwhelming. And part of its regulation, part of its lives are at risk, right? We get that. the ⁓ to improve healthcare, improve lives, have AI, have an extremely positive impact on outcomes, ⁓ just on a high level on a system that, you know, an opportunity for these extra costs to really be taken in.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: be taken out of it. I do think over time we're going to see massive improvements there. But you know if you were starting today and trying to do something and you were trying to get into that industry especially with a startup or consulting or something these are you know really challenging environments that are adverse to the change even if individuals, individual doctors maybe or individual administrators want to help advocate these changes. As an organization, the organization is built to resist that change and sort of defend the status quo and patient safety. And we get that. But on some level, we to work to push these things forward and help. so maybe you have to start at an administrative level. How do we improve the ⁓ paperwork of this process in a way ⁓ that with the administration side so that the system can still free up cash flow?
Nick Melnychuk: Mm-hmm.
Brian: to maybe invest and try things in different areas. How do we do pilots in a way that helps with that risk aversion, but still move things forward? And it's just very different environment that I think ⁓ is extremely challenging, but has so many opportunities.
Nick Melnychuk: I'm curious here. So do you think that in this industry, so within in this environments, will this change and the slow adaptation towards AI? Will it come from new teams from the market like from startups like ours? Or will it be something that the company will ⁓ test and deploy internally?
Brian: amazing. You know, I get a lot ⁓ customers or prospects right now that we're talking to that are really ⁓ us to come to them with an actual solution. So think... ⁓
Nick Melnychuk: like, like the one day that you can build it for them or why don't you come in and you say buy this, this, this and it's going to work.
Brian: Not that you can build it, but like you already, exactly, like you've either pre-packaged these things you would buy, or you have already started to build this thing that would nearly be done for them, and you know how to do this part of their business as well as they do, and you've already thought about this solution, and you're down the road on it. And so I think these companies that have this vertical knowledge, they are going to be able to work into companies and succeed. On the flip side, think companies that have their own proprietary data that no one else has, they're going to have to in some way, they're going to be developing their own solutions and they're going to be extremely successful long term. Right? Like I worked with a company that does have its own data. We helped AI enable some of this data and with seven or eight days of demos, they sold 800,000 in new revenue.
Nick Melnychuk: Mm-hmm. Nice.
Brian: explosive like that changes a small company and how they work and how they go to market but if you didn't have this proprietary piece
Nick Melnychuk: It is. Yeah.
Brian: Right? No one else could sell them a canned solution because no one has the data and the thing they have. they had to bring to market. And when they did, it's very good status for them. I think is sort of going to be the challenge. It's ⁓ which things my company buy and which things should they build? And I just think that equation has changed. I think that dynamic has always been there. We've always talked about that for the last 30 years and we had shifted, I think, a long way towards SaaS.
Nick Melnychuk: Mm-hmm.
Brian: And I think with the ability of claw code to really just recreate these very specific solutions to your industry, think some of that will shift back over time. But I don't think, you know, I'm not, not a, you know, a sass doomer or anything. I don't think that the big banks are leaving Salesforce anytime soon. Like they're there, but we've got to which ones. And I think that dynamic is changing dramatically towards
Nick Melnychuk: Mm-hmm. Mm-hmm. Just do it. Yeah. Mm-hmm. Yeah, me neither.
Brian: of doing it in-house again and being able to.
Nick Melnychuk: Okay, a follow up question here. So it still doesn't matter. Like the problem is gonna be the same whether the company decides to build something on their own or they decide to buy somebody. But whenever you, and you mentioned it already, like whenever you ⁓ into the process, there is the process and then there is the way that humans are actually doing the work. And sometimes those are two friends. What did learn about... how humans actually do the work, do they do exceptions, how ⁓ they find creatively workarounds, how do they create those workarounds, what is your experience around that?
Brian: On some level, right? It's frightening how little the business knows how the business operates sometimes Right. It's fascinating. I think we've seen some workforces where let's say you have a few hundred people doing a job That's done by humans
Nick Melnychuk: Mm-hmm.
Brian: there is, like you said, the sort of documentation that is out of date. Like you can't look at the documentation and get the answer. And if you talk to the human, they'll say, well, this data really means like this. for this exit criteria to be exist properly the right way, you actually do it this way, even though normally you would do it that way. And so the happy path is kind of where the corporate documentation tends to be. And then it's sort of like these exception cases, are completely undocumented.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: going to automate that, that needs to actually work over time, right? Like we actually need to know what those are. Um, I think that represents a huge challenge when you have just masses of people who sort of the humans are learning over time from other people or experience, they're making some failures, they're learning to do it their own way. Kind of exactly tribal. And they're not going to turn that back into some document, you know, somewhere. So, so we've actually implemented some projects where the AI is basically watching the logs of what the humans are doing.
Nick Melnychuk: the tribal knowledge, we.
Brian: and that we record that and we go through with AI machine learning to say, hey, in the future, maybe this should be something, a piece of knowledge that we've learned.
Nick Melnychuk: ⁓ huh.
Brian: now that we know this exists. And then the system can actually apply that quote unquote, let's call it an edit for the moment in a future iteration and then still put it in front of a human. know, humans always in the loop in these spaces, they'll put it back in front of the human and it'll just already be edited. And it says, Hey, this is why we edit it kind of thing. And now the human, that human may not have ever learned that lesson yet, but they're getting the edit and they can choose to accept it or not.
Nick Melnychuk: Mm-hmm.
Brian: But these types of things, we've really had to think through, sort of surprising. We sort of thought some of these things, you know, would be well documented. But if they are well documented, the tribal knowledge is ⁓ really and it's hard to learn. ⁓
Nick Melnychuk: Yeah, I can tell you a story later about the company that is building this way to observe what the employees are doing and then build the map out of it. But I'm also curious about the companies, say mid-market companies that ran a successful AI pilot. What's usually the biggest blocker then? So for example, the company did the pilot, everybody's happy, it's successful. It ⁓ ⁓ it had to prove. What is usually the moment where the whole program goes into the garbage? Or is ⁓ there anything that, or like... Generally, if the AI pilot is proved the program to be working, going to be implemented like 100 % of times.
Brian: no, I think if get tripped up, it's always like we'll a PLC and that's great. The PLC shows that this would work, but then there's some sort of security check regulation review. Can we use that data? Is there bias? Is this allowed? These types of things that can put sort of some larger delays into things, but I think
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: If the POC is successful, right, people try to move to roll out. I think those rollouts take longer than people expect is the challenge, but like we're getting through that process. But, but what type of infrastructure is actually needed to go live? ⁓ it's much like when we were kind of rolling out.com and everyone thought.com was free and it's just some web servers somewhere. And they're like, why is this bill so big? It's like, No, no, there's like servers and they have to do different things and there's a whole deployment pipeline that has to be built that makes this work. And it's just, think that that effort to deploy is a lot longer than people expect. And you have to be prepared to go through all those hurdles, right? Everybody wants to release something secure. So we've to make sure we go through the security review. Everybody wants it to scale. So how are we going to do it? Pilot. And then there's going to be problems.
Nick Melnychuk: Mm-hmm.
Brian: And so things get stopped at a whole bunch of blocks. Like the first time a person gets an answer, probably an executive, they didn't think should have been answered that way. It's like, do we freeze everything? Do we turn it off? Or can work through why wasn't accurate? How do we fix it ⁓ and make this fast? ⁓ has turn around fast, but it has to be accurate again. And can we do that? Can we respond in that way? But I think those things slow.
Nick Melnychuk: Mm-hmm. Mm-hmm.
Brian: at least for us, our customers down, but haven't stopped from going live, which is good. But I can see how you can easily get tripped up ⁓ decide to give up.
Nick Melnychuk: Mm-hmm. Okay. So it's very interesting. by the end of the episode, I left the interesting question for the desert. So listeners and viewers, live with a very interesting insight. You in the beginning that you have been ⁓ actively and building for essentially years. That's a lot of time. ⁓ And that time,
Brian: You
Nick Melnychuk: you've seen a bunch of hype cycles and those hype cycles they come and then they dissolve, they disappear. within like the hype is not ⁓ usually not built from nothing, especially right now in AI there is a lot of hype, there is a lot of noise, but it's because there is also a lot of signal ⁓ in what's how the world is changing. So I'm curious in ⁓ your experience working within the hype cycles, and then digesting what is signal and what is not signal. Because for example, the company can implement the AI in the sense of, it just summarizes me my morning emails. That's nice. But it's not the experience like you mentioned about that company where you implemented AI and they made like 700k in in terms of sales. ⁓ that the company was changed by AI. So how do you understand from news or just from any other indicators that the company is actually implementing AI and AI is going to change this company. It's not just like to sprinkle on top and say, we're AI.
Brian: ⁓ anything someone puts out in the news I assume is a lie. ⁓ just start there. So everything ⁓ that regard is why I have to stay hands on. It's why I had to get Gemma installed last night ⁓ and to run OCR examples on it this morning. Can you really do that on a Mac mini? Turns out you can. Wow, okay, pretty good. ⁓
Nick Melnychuk: Okay. Mm-hmm.
Brian: But I need to know that it's real. And when I look at the companies and what they put out, you can see that a lot of them are putting out information that say, you know, we're doing AI, where we've implemented this very well, but it can't quite be at the level they're saying, right? Like it has to be some height, because if it were at the level they're saying, you would see it in the financials in a quarter, right? You might not see it right away. And I get that a lot of stuff is new, but you would start to see. know, increases in gross margin, you would start to see these companies talk about these massive productivity games that would either come in new sales, like the company we just talked about, you know, we exploded the sales growth and that would show up in the financials, ⁓ or saved a whole bunch of money in cost and that would show up in the financials. And so ⁓ I what we're seeing right now is a of investment being made, a lot of people ⁓ still rolling out.
Nick Melnychuk: Mm-hmm.
Brian: partially, but you know, we haven't quite figured out how to optimize what we're doing and I don't think it's that much different than when the internet came out, right? When the internet came out everybody sort of like put up a website and they put up a store and they tried to do some things and some of them worked and right some of them didn't and we did, you know, reverse logistics bidding and all these things to try to understand like which thing is gonna work for my business so I still think in this part of hype cycle, that's where we're at.
Nick Melnychuk: Mm-hmm. Mm-hmm. Mm-hmm.
Brian: these companies with all these announcements, they have a whole bunch of people. And if you can imagine them just like running all over the place. And so now they're running all over the place really fast. It's not like we're going somewhere. know, that vision hasn't sort of coalesced to where 10 % productivity in every employee we have turns into 10 % productivity for the company yet, because it's too scattered, too not focused just yet. But that will come with time as these
Nick Melnychuk: Mm-hmm.
Brian: everybody sort of gets used to the tool and ⁓ to go sort of in the same corporate direction as they sort of start to understand, how I bring these things to market in a way that, you know, what person A is doing and person B is doing isn't conflicting anymore. Like it's building on each other. You know, think we're going to see those flywheels start to happen ⁓ and then going to see the news, but you're going to see it in the financials. You're going to see it when they do their quarterly announcement. You're going to see.
Nick Melnychuk: Mm-hmm.
Brian: know, the productivity or you're or you're or you're not right. It didn't work, but it most likely it's going to go up once these individuals get aligned. Like we're just seeing individuals hit these individual flywheels of productivity where they're just pumping out ⁓ much more than they could ever do. Right. People having like AI log into JIRA and bring all their tickets together and start to format them in ways that are ready for coders in different ways that would have taken them hours and hours of going through all these submissions and AI prepping all that work so they can put it out the next day. Weeks of work being done overnight. And that doesn't change the rest of the company's delivery cycle yet, but as it hits the rest of the company, it will.
Nick Melnychuk: Well, this is amazing. It was a great conversation and thank you so much for sharing the best of the best insights from your day to day with us and with the viewers. I'm sure you're going to have a few questions coming from the audience, so ⁓ I'm to leave your contact. If folks want to reach out or they're to be interested in the case that you described with that explosive revenue. going to be a very interesting chat for you.
Brian: and I'm happy to ⁓ up and have conversations. And I'm glad that we had the opportunity to talk today. I really appreciate it.
Nick Melnychuk: It has been a pleasure. Thank you so much for coming on the podcast and thank you so much dear listeners and viewers for attending.
