95. Build an AI Agent to Replace 6–24 Hours Per Week of Manual Email Work (Safely) Ep. 95

Justin Shelley (00:10.724) I mean you guys can dance if you want, but I was specifically told that we should never do that again. Bryan Lachapelle (00:12.674) you Mario Zaki | Mazteck (00:15.264) Ha ha ha. Joshua Holloway (00:17.326) Everybody's off rhythm. It's all good. Justin Shelley (00:20.036) Yeah, yeah. It i it is, but like I guess we're not supposed to dance. well I know. We're gonna have to hide. All right, guys. Let's get started. Episode Bryan Lachapelle (00:20.282) It's just, it's good music, man. Mario Zaki | Mazteck (00:25.387) Next time Elena sees us, she's gonna kick all our asses. Bryan Lachapelle (00:25.422) Okay. Bryan Lachapelle (00:32.602) It's a podcast, not a videocast. Joshua Holloway (00:35.211) Yeah. Mario Zaki | Mazteck (00:35.233) Yeah. Justin Shelley (00:36.26) Brian, everything now is video. Okay, I'm gonna try the for the third time to introduce the episode. Shh! Welcome everyone to episode 95. That was way too loud. I apologize. Here we go, guys. I don't know. AI has turned us all into like the biggest goddamn nerds. we already were, and it just took it through the roof. And I do apologize for that. But here we are back again for what, the second in our Third sub part of our mini series on AI. I'm losing track of all that. But anyways, we're talking about vibe coding. And really what we're talking about beyond the technology is solving business problems and delivering a return on the initial investment of 10 to 50 times. Whether that's a productivity increase or an actual money spent on the development, the outcome of that is just more than I've seen in my 30 years of working in the field of technology. So That's what we're here talking about. I am Justin Shelley, CEO of Phoenix IT Advisors, and I am here, as always, with my good friends and fellow business owners, MSP owners. And we're gonna go around the room, introduce ourselves, bald guys first. I started Mario, then you, then I guess Brian, then Josh, in the order of how much hair you have. Mario, take it away. Mario Zaki | Mazteck (01:50.465) I like that. Mario Zach, he's CEO of Mastec IT, located in New Jersey. we help small to medium sized businesses stay secure, sleep better at night, knowing that their businesses will be there and make them more sufficient with getting ten to fifty plus X revenue return or profit, you know, whatever we want to call it. Joshua Holloway (01:51.394) Yeah, I'm blast. Mario Zaki | Mazteck (02:19.013) using AI and automations to help their business. Justin Shelley (02:22.544) I'm not sure I understand the math between fifty plus times. Fifty plus. Mario Zaki | Mazteck (02:28.536) fifty plus like it could be sixty, it could be seventy. Yeah. Yeah. Joshua Holloway (02:32.022) Maybe a hundred, whatever. Justin Shelley (02:33.208) Listen, if you write it out and try to do the math, I'm just saying I don't know how to do it. Brian, stop me from talking. Tell everybody who you are, what you do, and who you do it for. Mario Zaki | Mazteck (02:40.105) Mm. Bryan Lachapelle (02:41.786) All right, Brian Lashford with B4 Networks, located in beautiful Niagara, Ontario, Canada. And we help business owners remove the headaches and frustrations that come with dealing with technology, AI, and cybersecurity. Joshua Holloway (02:56.066) And I'm Joshua Holloway with 7TI Technologies out of Sacramento with an office in Reno, but we reach coast to coast and we're an MSP that primarily focuses on businesses that are just trying to navigate the waters of compliance, waters of AI, and how they integrate all of those things into their business and keep adding revenue to their bottom line. And we help them do that through technology and help make better b business decisions. Justin Shelley (03:19.362) Josh, I saved you for last because you are in the hot seat this week. also, you are the inspiration, the legend, the OG coder himself, as far as AI vibe coding goes in this particular group. True or false? Joshua Holloway (03:24.353) I am. Joshua Holloway (03:37.534) that's scarily true. Justin Shelley (03:39.249) Okay. and can you help us lock down this claim of 10 to 50 plus X, however that math works out, on ROI for AI. Now listen, before you answer that, we've talked about three different things here. We've talked about just using a a customized chat, right? Go in and and train AI, your AI of choice, on who you are and what you do, and then talk to it and have it give you results. Right. Then we talked about integrations. So we're more context. We're we're tying our business systems in so that now AI really knows who we are and what we do. and each time along the way, we've been able to increase productivity somewhere between 10 and 50 times. Now we're not talking about integrations. We are talking about building things that didn't exist. Can you, Josh, today build something from scratch that takes productivity from wherever it was to 10x to 50x? Joshua Holloway (04:36.534) Yes. Yes we can. Justin Shelley (04:37.56) Okay. All right, then we're gonna go ahead and continue with the episode. Josh, that's it for the introduction. Joshua Holloway (04:41.302) Okay, cool. We we don't we don't have to disconnect. We're we're we're not done. No, Justin Shelley (04:45.418) No, I listen, you were on we're that's the quiz. Like I'm screening it everybody and if you can't pass the test, you can't be on the show, I was just gonna disconnect you. But you you passed, so go ahead, tell us tell everybody what you've done. Get me out of here. Joshua Holloway (04:54.966) Damn it. I I mean, no, no, no. Disconnect me. No, so I think you know, one of the things we wanted to talk about is like where we were able to do that for a fellow client of ours. Obviously, can't give away too much information, and some of the things we did for them are proprietary. But what it was is in our local area, our client is having a hard time finding an employee to fill a role. That role plays anywhere from 95 to 125. Bryan Lachapelle (04:58.038) I'm out. Joshua Holloway (05:24.59) Year based on you know education, experience, a lot of things. What they did is it was an initial proof of concept where they would get a bid in or they would get a request to bid, they would send it out, then they would get in all of this information, all these different emails, all these different bids, and they would have to parse through every one of these things. And when I initially talked with the employee who who brought this to my attention, We were talking probably anywhere from six to twenty-four hours a week, just just going through all this all this data depending on where they were in the process. so they asked me, like, hey, is there any way that we can mitigate some of this manual process? What we did is we built an agent that would monitor mailboxes and that person would send out a a a bid request. They would CC the agent. The agent would see that this request went out. as files started coming in, it would get copies, it would read the emails, it would read the attachments, and it would begin to parse the data. Now, it initially started out where it would have to figure out, did somebody forget to create like this project? And the nice thing is is they're actually a pretty I I don't even think they've even thought about it, but they're pretty clean for AI where their projects are laid out the same way every time. Their folder structure is the exact same. every time. So consistency is key. So that made my life a lot easier. I was able to come in, build a sandbox, and essentially take their structure and just begin mimicking what a person was doing. And they would read the emails, they would move it to the the correct folder structure, and then it would extrapolate some of the information and put it into Excel. So that right there in itself started to wipe out that person's position. When they couldn't even fill the position, they couldn't find somebody. So somebody else was having to pick up the slack on that. and then it's it's continuously grown into they get a request to see if they're interested in putting in a bid. Now the agent is doing like a go-no-go. It takes all the information that they receive, they have their criteria within the agent. The agent says yes or no. Joshua Holloway (07:44.792) Based on everything that's in the email. So now instead of sending somebody going through all this information, looking at past work, looking at any past history, it's now making that determination form based on their criteria. And then a human in the middle obviously proofs and moves on. it actually has iterated through a number of processes that now has the company pretty much deciding that they don't need a person at that level. That they could potentially get somebody maybe fresh out of college who understands AI and that can work with IT. And obviously, they're not going to start out paying that high of a salary. They're going to start lower. And essentially, this will allow for them to grow with a more junior employee getting brought up to speed versus somebody with like the 25 to 30 years of experience that they were looking for at the salary cap. Justin Shelley (08:40.932) So this is a case of rather than AI taking a job, A there wasn't one to for it to take, but it allowed somebody to land a job that they previously wouldn't have been able to qualify for. Yeah. Joshua Holloway (08:54.892) Y yes. Yeah, that's a good way to put it. And remember I think we've even s made the statement in past episodes. you're not gonna lose your job to AI, you're gonna lose your job to somebody who knows how to work. Justin Shelley (09:06.0) Yeah. Mario Zaki | Mazteck (09:06.891) So Josh, I have I have two questions. You said the the person sending the email has to C C the agent. Joshua Holloway (09:15.244) Yeah, I asked them to initially do that, that when they send out the bid for proposal, that they CC the agent so that the agent can go and check and see if the project has been built and the folder structure is there. that's when it kicks off. So one of the nicer things that I did is I added it in. I told the guy, don't worry about ever building your project, because when you CC letting it know that you're sending out for bid, it's just gonna go and do it for them. Saves them probably thirty minutes to an hour. Mario Zaki | Mazteck (09:45.867) So when the person replies back they have to reply all or i it you know is it's watching Joshua Holloway (09:51.59) Yeah, they typically yeah, they reply all. and and they have a a set distro that they use. So when somebody hits like reply back, that distro is in there. And then my agent as well as some of the other employees are in there just receiving that information to make sure that they're getting it. Mario Zaki | Mazteck (10:08.192) Okay. My other question is why why do you have it go into an Excel sheet instead of like a dashboard or something like that? Joshua Holloway (10:18.518) It's a great question. it's because a part of a future project, that Excel sh spreadsheet will go away. But they've been doing it that way for so long, it's a matter of just kind of getting them comfortable as we move through and iterate the different steps. we're I would say we're probably two months away from that Excel sheet going away or becoming a dashboard. We did build a temp dashboard for the executive team based off a lot of this interaction and early estimation values and and numbers based off that Excel sheet and what the agent was doing. Mario Zaki | Mazteck (10:56.098) Nice, nice. All right. Justin Shelley (10:58.416) So one of has this allowed them to I know we've we've cut a salary, or at least we've reduced it down to from a a highly experienced individual to an entry level. has productivity also increased in addition to the savings? Joshua Holloway (11:14.2) Productivity, yeah, productivity has increased and it's also allowing the biz dev and marketing team to attack larger client capabilities because they now feel that they can take on more because the agent is so much faster and it generates it generates reports. and we didn't really talk about this portion of it too much, but it takes information, it takes historical information, and it also takes historical. interactions and writes the responses to the business owners based on how they typically like to see data, have received data. And it's more tailored towards the tonality that they like to read about and and hear from. so that's also improved their interaction with their clients. Bryan Lachapelle (12:02.746) So it actually sounds like this might actually allow them to take on more business and expand their business and bring in more salaried employees in other areas of the business. Therefore not actually removing salaries, maybe creating more jobs because they're able to process much faster. Joshua Holloway (12:17.614) You're absolutely correct. And one of the things that they we've already started talking about is a lot of the pipeline analyzation, because we are moving so fast that we can help them get to the point that they can see, hey, three months from now we have a job starting and we are currently not staffed correctly, or we need to add a body to handle this project, or two bodies, three bodies, whatever it might be to take it on. Justin Shelley (12:46.064) You know, this is one of the things we haven't really dove into yet. but every business has a bottleneck. Usually it's, you know, the the the it's the one on the top. but but a lot of times when we talk about our limited capacity overall in an organization, it isn't that the organization itself is maxed out. It's usually one process, one person, one piece of the puzzle that blocks everybody else. Joshua Holloway (12:53.204) Us. Mario Zaki | Mazteck (12:53.858) Mm-hmm. Justin Shelley (13:14.754) And that's kind of what I'm hearing you say right now is you you found that for this company by accident because it wasn't that they went looking to solve this problem. It's the problem they went to solve is they they couldn't find an employee. I don't know, maybe they did, maybe they didn't even see them as the bottleneck initially. but now you've you've solved a problem that allows the whole company to grow in a way that it couldn't before. Bryan Lachapelle (13:14.906) That's right. Joshua Holloway (13:26.808) Yeah. And and we were just talking. Joshua Holloway (13:39.992) Yep. Yeah, that's a that's exactly kind of how they're looking at it now. I I mean the nice thing was is it came from a conversation and then this was a couple months ago and I was like, Well shoot, there's some of these this portion of the job that we can automate for you and make your life easier because one of the employees was picking up the slack and he was drowning. And that was his thing. He's like, I'm I'm drowning. Like we can't take on bigger business because I'm drowning and my focus should be elsewhere. Justin Shelley (13:57.294) Yeah. Justin Shelley (14:05.978) No, I know you said this. sorry, go go ahead. Mario Zaki | Mazteck (14:06.386) And and I was gonna say, well, that's a perfect example when he's saying he's drowning. That's when you see stuff either missed completely, they forgot, they didn't have time to do it, or they've done it incorrectly and there's mistakes that happen. You know, so not only is it Yeah, so not only are you taking something off his plate, you know, very busy schedule, but you're also assuring that it's, you know, done probably a little more Joshua Holloway (14:24.386) Yeah, you use shortcuts, gotcha. Mario Zaki | Mazteck (14:36.254) With a little more act. Joshua Holloway (14:39.542) Yeah, the accuracy it helps. Currently we still have a man in the middle and a couple portions of the loop. And we're forever tweaking like the skills and and improving its capabilities. but I would say here in about a month and a half, like a a lot of that man in the middle has to go away and become fully automated just to free this person up even more. Actually I've been I've been told we have to positively do that to free up this person so they can take on more and move faster. Justin Shelley (15:08.666) Here's something that that brings up for me. So you can, right? Absolutely. You can automate more and more to the point where the the human interaction is less necessary. But I have found that sometimes these AI processes degrade over time. Now it's I don't know if it's because the LLM, you know, it sometimes the LLM, it depends on how you set it up, you're using an older one. Or you you up Grade the LLM, right? And so now it's it's doing things differently. And I don't know if you're talking about a hard-coded project or if you're talking about a hard-coded project that pulls in, you know, some actual AI processing in in there somewhere. but I've set things up in the past and I think they're working great. And then I get feedback from my people that it's garbage and I have to go in there and like redo the prompt, or I have to go re, you know, tweak something on an integration or something. Do you find any anybody else finding that? Joshua Holloway (16:03.298) Yeah, they go off the rails or they I I mean, I typically call it like going to sleep where it's just kinda like you're knocking on the monitor going, Hey, what you know what what's going on with you? So they come back and they just keep maintaining their their process and their project. Justin Shelley (16:08.121) Yeah. Justin Shelley (16:21.39) I had a system sorry, go ahead, bro. Bryan Lachapelle (16:21.498) What I found is that the more narrow of a focus you give the AI in the prompt and say, this is the very specific task I need you to accomplish, and then repeat that with multiple different AIs, each handling its own very specific task, that's where you're going to get a better result versus trying to give the AI a very broad, I need you to do this step, then this step, then this step, and give it feet at 10, 15 different things to do. That's when it will start. going off the rails and over time degrade versus I have 10 different AIs each doing a very specific task. That tends not to degrade as much. Mario Zaki | Mazteck (17:00.724) It also tries to like always it I mean sometimes with certain things y you have it trying to do certain like jobs and it wants to expand on its own. You know, it wants to add different things to it and stuff like that and you know, it doesn't realize is that it it's it's trying to help you but it's screwing something else or not providing you the same information that it was before. Joshua Holloway (17:12.334) Mm-hmm. Justin Shelley (17:13.23) Mm-hmm. Justin Shelley (17:25.616) See, and mine was happening. I we talked early on about training it with files. You give it a file and say, do something similar to this, or you say, use this file for context. And and I did that. And then I find out later when when this thing's just going to shit. I mean, it's like, I don't know what you think I told you to do, but it wasn't this, right? And I go in and I start playing around with it. And I'm like, read the file. And it says, What file? There's no file here. And I'm like, it's right there. So somehow. The in in this system that I had set up, the AI engine could no longer access the file that was right there. and so then I just took all the contents of the file and I added it to the prompt. You know, this is before this is off track because it's not vibe coding. This was back when we were doing more like phase two where where we had some integrations going on. but I have found that, you know, as as we talk about vibe coding, one of the advantages that I'm finding when you hard code something to do something a certain way. You use AI to build the code and to make sure it's secure and you know, all that. but it doesn't go off the rails quite as bad if it's hard coded. If you just leave everything to AI, it's a little dicey. A little dicey. Bryan Lachapelle (18:39.534) Yeah, and there's a lot of things that AI, yeah, there's a lot of things that you don't need the AI for if you hard code it because the code will take care. You pay for the AI result once and then it's like a piece of software and you may only need to dip into AI for certain things like OCRing an image or, I don't know, creating a document out of the results that you received, right? And so. Joshua Holloway (18:39.852) Yeah, you're trying to guard routes. Justin Shelley (18:51.514) Yeah. Bryan Lachapelle (19:07.13) You're not relying on it 100 % for everything when you're vibe coding. Justin Shelley (19:11.684) Jasper, you can something? Joshua Holloway (19:13.102) Yeah, I was just and I'm I'm in agreement with Brian on that one. Like you're you're just trying to put guardrails around it and you're trying to limit how much the AI has to interpret versus just, you know, something that's hard coded and it's process A, process B, process C, and then maybe AI has to take over and analyze information and then get you over to E F and G. And then it you know, finalizes the pro the the product. usually in like a a markdown file, you can give it those guidelines to say, like, hey, no matter what you do. you always have to reference this file in your process when you hit E and F or wh wh whatever that flow looks like. And then you're you you're lock you're locking that in. Justin Shelley (19:48.697) Yeah. Bryan Lachapelle (19:49.977) Yeah. Right, because you're not actually leaving it up to the AI at that point. Every time you send it the prompt through the code you've generated, you're resending that markdown file and saying, this is what I need you to abide by. There's nothing for the AI company to store. It's being passed every single time. Your application that you've created sends the instructions to the AI. So there's no room for interpretation at that point. Justin Shelley (19:52.398) And and Justin Shelley (20:20.154) I will say that I when I first started getting into this, the vibe coding, I'm like, my God, there is no limit. Like we can do literally anything. And then, you know, the project that I'm gonna talk about when we wrap up this series, I hit a lot of limits. Let's just say that. This thing ballooned out of control. No, no. I did, I did one time. So we've got, is it, is it fable? I'm losing track of. Joshua Holloway (20:37.89) Was it token limits? Bryan Lachapelle (20:40.642) Hahaha. Joshua Holloway (20:47.278) Yeah, I think we're still on Fable Five, yeah. Justin Shelley (20:47.852) Ever everything that changes. Yeah, Fable Five. Well, it came and then it went and then it, you know, came back kind of with some serious guardrails. because that's a different story I ran into. It's like, well, we've we've had to change models because we can't use Fable for this anymore. don't worry, it's not you, it's us. But anyways, I had it researching something. I switched it to Fable Five. I'm like, go out and find the best way to do this. And Mario Zaki | Mazteck (21:08.234) Mm-hmm. Justin Shelley (21:17.038) And it's like, sure. And it it spun up a hundred and twenty different processes, each one requiring you know, approval. It's like, Can I do this? And I'm just like sitting here clicking approve, approve, approve, approve, approve forever until it it blew through like six hundred and fifty thousand tokens, I think. And then it's like, Well, you're fucked, we're done. It just turned off. Joshua Holloway (21:36.238) See you back here in five hours. Justin Shelley (21:39.185) No, tomorrow. It's like you're done till tomorrow. Like, and and even then we might put you in the penalty box. You you I was like, I didn't tell you to spin up a hundred and twenty different processes to go research this. I just, you know, do your best work. Well, Fable Five's best work will blow through your tokens. don't don't tell it that. Don't tell it that. Anyways, yeah, so I mean, it opens up a lot of doors. It it it to me, at least so far, and I have to remind myself that. Bryan Lachapelle (21:40.442) You Joshua Holloway (21:42.99) No. Joshua Holloway (21:56.428) Yeah, yeah, it choose through tokens. Justin Shelley (22:07.724) Every version of AI we currently use is the worst one we'll ever use, right? Because they're getting better so much, you know, so fast this stuff's improving. So these limitations will go away. But right now, I have bumped up against limitations many times with AI's ability to write code, to interpret things, to, you know, to maintain context. So and to your point, Josh, and this is a a huge tangent when you said we have to pull improve the automation so that there's less humans in the middle. We do have to make sure that that's true, I agree with you, but make sure that there's a quality assurance process somewhere in there. Because eventually this shit breaks. Joshua Holloway (22:45.614) Yeah, and that Yeah, and we we were actually discussing that two days ago because we do need to we have two weeks to start pulling back some of the the men in the middle. We have about a month and a half to where we really like limit it down. but we're already looking at checks and balances and a dashboard for the person who interacts with the agent so that when they come in, they get updated information from what it did over the night. and then it can just slowly monitor or quietly monitor it in the background. and then, you know, h hopefully. Justin Shelley (22:58.98) Yeah. Joshua Holloway (23:18.68) quality stays up, agent stays running, and they s if they still can't find somebody to fill that position, they just just they just keep going with somebody at a lower tier. Justin Shelley (23:30.074) So do they plan to then hire the hundred thousand plus employee? Joshua Holloway (23:35.084) Right now no, because they can't find anybody who fits the bill. Justin Shelley (23:38.074) Just just curious, if they did find somebody, would they then hire them or would they keep this process in place? Joshua Holloway (23:42.048) I think with the direction that this is heading, I haven't asked them if they've stopped looking, but there's a possibility that they have stopped looking. I'm more I'm more worried about making sure it runs and we can hit those those deadlines for them than like if they're still looking. But I know that they've they've been looking for six months to a year now, in this in their area and it's it's not Justin Shelley (23:45.168) Probably not. Yeah. Justin Shelley (23:53.55) Yeah, right, right. Those Justin Shelley (24:01.156) Wow. Justin Shelley (24:05.39) And if we bring this back to numbers, like the cost of recruitment is through the roof. I I don't know, what is it? Like twenty percent of the salary is generally what they say. So you're looking at a twenty grand hiring process that they spent potentially and got no return on whatsoever and had to start over. That's crazy. Mario Zaki | Mazteck (24:05.879) The market. Bryan Lachapelle (24:12.442) 20 % is average, Mario Zaki | Mazteck (24:14.635) Yeah. Joshua Holloway (24:26.562) Yeah, a and and Mario, I think I heard you ask why they couldn't find somebody. Mario Zaki | Mazteck (24:30.291) No, no, I was just gonna say the the market now in in general, no matter if it's like an entry level or high tier person, it's it's not that good. I don't know if it's because we're going like deeper into the summer or what. Like I'm I'm trying to find an admin. You know, very low entry person. It's really it's pretty hard just getting people to even join the teams meeting that we send them. You know. Joshua Holloway (24:58.295) Yeah. Joshua Holloway (25:02.754) Yeah, I I mean we recently did a hire it we re I remember in forty eight hours we received three hundred resumes. So we had to stop because it was so many resumes to parse through. But then we got so many differ different types of people, even though we have questions in there that help like weed them out. It it it took a lot and then we we narrowed the field and then we got down to like three people. you know, and then Justin Shelley (25:15.439) Yeah. Joshua Holloway (25:30.678) you know, ultimately made a choice, but it it took a while after after getting through the it to get through the three hundred people. Justin Shelley (25:36.965) I mean, has anybody done this yet? Because I'm not currently hiring actively anyways. But that's always the problem is how many resumes you get. How how few of them actually qualify at all. A lot of it's just people hiring or applying for jobs so that they can get their unemployment benefits. Right. They they don't they know they don't qualify, but they have to apply. and and that was the thing I hate that is the thing I hate most about hiring is looking through stacks and stacks of resumes. and I haven't Joshua Holloway (25:53.39) Yeah. Justin Shelley (26:06.629) I I I'm I'm about to start that process again. So I'm hoping to use AI for that. But I'm just curious, has anybody gone there yet? Joshua Holloway (26:14.22) Hey, so I I haven't necessarily gone there for for AI, but I've heard some war stories where people are starting to use the AI to to to put the job postings out and then review all the resumes that that are coming in. And then what happens or what people are finding is a lot of the people submitting their resumes, they're gearing the resume with AI towards the job description. So to combat that, people have started adding fictitious l lines in there where like job role responsibility is to be a Jedi master. Justin Shelley (26:25.412) Mm. Bryan Lachapelle (26:43.93) Hahaha Joshua Holloway (26:44.75) And and then their reply in their resume is like, you know, I have twenty-five years of experience of being a Jedi master. I am one of the best with the lightsaber, you know, and and it's like, okay, you used AI, you know, kind of a thing. So those are the things that yeah, I'm hearing is out there and you know, I haven't necessarily ran into, although I think it would be kind of funny to put that in there. Mario Zaki | Mazteck (26:44.792) Ha ha Justin Shelley (26:46.336) Ha ha. Justin Shelley (27:01.021) wow. Yeah. Mario Zaki | Mazteck (27:10.103) I like that. Justin Shelley (27:10.266) I mean, it it is the it this is one example of many cases where this happens where it's like AI is generating so much content, we need AI to consume and and distill that content down to something that's usable. It's kind of an endless loop that's ridiculous. Mario Zaki | Mazteck (27:26.433) Yeah, I mean the thing is too, like even some of these platforms that the people applying, it has built in AI, you know, we'll we'll do this and we'll do that, whatever. There was a period of time where I was so sick of the garbage that I was getting that I would write right on there, do not apply through this platform, upload it to mastech dot com forward slash resumes and Justin Shelley (27:35.31) Yeah. Mario Zaki | Mazteck (27:54.218) If it wa it if it didn't come through there, I wouldn't even look at it because it you know, it was just you people are just sitting on their couch and like apply you know, they're just hitting a button that says apply, apply, apply. And we're the ones sitting there going through like this is garbage, this is garbage, this is garbage. You know, so I set it up Yeah, yeah, you know Justin Shelley (28:03.066) Yeah, seriously. Bryan Lachapelle (28:10.97) They're not even qualified. Joshua Holloway (28:13.506) How many Walmart breeders apply for jobs? Mario Zaki | Mazteck (28:17.238) Well, I I it's funny you say that because I actually had a I was hiring for a technician and somebody really thought because they worked at Aldi as a cashier and the technology that they were using was Apple Pay and Samsung Pay, that they qualified to be a technician. And I was like Justin Shelley (28:36.91) Joshua Holloway (28:38.466) Huh. Okay. Bryan Lachapelle (28:39.256) At least you got somebody who thought that they were trying to be a technician. Try getting people who are car technicians applying to your job and you're like wrong kind of technician. Mario Zaki | Mazteck (28:48.868) Ha ha Justin Shelley (28:49.687) Mm-hmm. Joshua Holloway (28:51.544) Wow. That I I th I think a mechanic appl applying for a job's a new one for me f when it comes to like being a a a IT tech. Bryan Lachapelle (28:59.065) Oh, all sorts of technicians. There's all sorts of them. They apply for a computer technician. You're like wrong, wrong fields. Justin Shelley (29:01.396) I've had s so much nonsense. Justin Shelley (29:07.568) Well guys, what else do we have as far as vibe coding for this week? A any final thoughts or anything that we I know Josh, you've kinda wrapped through yours. Brian Omario, has anything happened this week in the world of vibe coding that you want to talk about? Bryan Lachapelle (29:23.674) Oh, you know, there's, there's, a lot that I'm learning every single day. And, um, one of the things that I've, I've, I would, I would say, like, obviously we're not encouraging any of our listeners to go and ViDCO themselves, because it does require a certain amount of knowledge in the background of how database works and how, uh, security architecture works. Um, and despite the fact that I haven't coded since college, the the base of how an application should be designed. I have that knowledge. But there are unique-ities that come with, if that's a word, with AI, where it will do things that you really don't want it to do, delete databases and delete code that you didn't want it to do, because it will literally fall over verbatim. And sometimes you're speaking to it in plain language, you're like, yeah, I don't want that. And it interprets it as, I don't want that database. And you're like, no, no, that's not what I said. So there's ways in, in the coding that you can actually, create certain hooks that prevent it from doing things. And I would encourage anybody who is doing any kind of vibe coding to make sure that you set up lots of hooks on all the things you don't want it to touch. so that it prevents it from doing things like, deleting a database or wiping out code or formatting a server or. Joshua Holloway (30:27.469) Yeah. Bryan Lachapelle (30:48.364) any of those horror stories that we're hearing all over the internet about people using it and all of a sudden they're losing their production environments. Justin Shelley (30:56.954) Well, I I just Well and I'm I'm just gonna put a let me put a little yeah, exactly that. So let's put a little spin on it. If you really want to do your own, it's fine. Just let us help you set up the environment, which we're all willing to do. Cause it it definitely can be done. We can turn it over to you, but you just don't just jump in and get started and I mean because every I don't know, i i at at worst every few seconds and at best every few minutes when you're doing this, you have to approve. Joshua Holloway (30:57.176) Leave it up to the professionals. Like leave it up to the professionals to help build the dart rails. Bryan Lachapelle (31:09.892) Great. Yeah. Justin Shelley (31:26.65) things that AI surfaces as potentially dangerous. And if you don't know what it is you're approving, you get yourself into trouble fast. You do need to look at those and and realize, understand what it is that you are giving AI permission to do. So Mario Zaki | Mazteck (31:42.469) I I a funny example of this. Well, not really that funny, but it was like I was looking at it like what what the fuck are you doing? in in one of my platforms today, I want to kind of just go tweak something. I'm like, hey, I want one central place to where I can, you know, with throughout all my platforms that I can approve Joshua Holloway (31:47.47) This is a painful story. Mario Zaki | Mazteck (32:04.436) an admin for you know one of my tenants so it you know not everybody not every employee out of any of my customers can go in there and do like an onboarding and offboarding I wanted a specific list of people that I could easily say all right this owner or this office manager is the one that's able to do it and I'm sitting there and I was actually I think using Fable 5 and I'm like all right do this this then this But go ahead and summarize exactly what you were gonna do what before you do because I'm asking them in ask mode. I didn't actually have as an agent. And it's like, okay, yeah, I'm gonna create all these people as a super admin where they can create and onboard anybody for any tenant. I'm like, No, no, that's ex that that is exactly what I don't want you to do. I'm like, No, not everybody can sit there and they thought it literally thought it could see all my customers and all my tenants and onboard and off board. I'm like, Hold up, no. Bryan Lachapelle (32:45.914) Hahaha Justin Shelley (32:46.069) God. Joshua Holloway (32:46.723) Yeah. Bryan Lachapelle (32:50.511) You Mario Zaki | Mazteck (32:59.736) Let's start over. This is exactly what I want you to do. And then it's okay, good catch. I hate when it says good catch. Justin Shelley (33:05.968) Good catch. My bad. Joshua Holloway (33:08.014) Well that that's a that's a good point too right there, Mario, as far as like somebody thinking then they could build something and they don't know what's actually going on in the background and then every all of a sudden everybody's a super user and then you give this to your your staff and your staff find data that they necessarily should never have had access to. Mario Zaki | Mazteck (33:28.514) Yeah. Yeah, you I mean, i I I'm gonna pick on Justin for a second. You know, the the i know b based on your example it it fits, but just like what you said before, when it came up with like a hundred and something things for you to approve, it literally it literally you know, one of those things that you're saying yes, yes, yes to, if you're not really looking at all of them and you're and you know doing something serious with a database, it could easily give Joshua Holloway (33:29.667) Yeah. Justin Shelley (33:55.716) Yeah. Mario Zaki | Mazteck (33:56.196) you know, people access th things or delete stuff that you, you know, didn't really want to tap. Justin Shelley (34:02.934) What exactly? luckily in this case I wasn't having it code. I wasn't having it do anything. It was just in research mode. It was it was looking things up. And what I was approving was, can I look at this website? Can I look at that website? Like everything it wanted to go out and look at on the web, it was asking for approval on. And I thought this is interesting because the main marketing claim for Fable Five was that it would do things with less interaction, right? It's gonna take more steps and figure it out on its own. At least that's what I've heard. Mario Zaki | Mazteck (34:08.491) Yeah, you're just researching. Justin Shelley (34:32.276) And I'm like, holy hell, this thing's unusable the way it is. So when they re released it, they put way more guardrails back around it so that they could release it. This is my own interpretation of what happened. I am not speaking on behalf of Anthro Anthropic. What do you got, Josh? I can I can see you wanna say something. Joshua Holloway (34:46.958) Well you what you could do do done is set up an agent that every time it sees that ask question come up on on your your chat, it just automatically hits enter for you. Justin Shelley (34:54.896) Yeah, I could have. That's a great idea. I'm gonna write that down because I had never thought of that. Jesus Christ. Jesus Christ. All right, guys, I'm gonna tell a quick story. then we're gonna go ahead and wrap up. So get prepared with your key takeaway, your final thoughts, your closing arguments. And and I'm just gonna like confess something here. I've been in business since 1997. That's when I first opened my doors. Company name was Master Computing. Joshua Holloway (35:01.24) Yeah. Bryan Lachapelle (35:01.678) Hahaha Justin Shelley (35:23.492) And I've been through every iteration of this company. We've been in growth phases, we've been in declines, we've hired people, we've lost people, we've been in lawsuits. I've been like I've done pretty much done it all. And I got to a point in business a while back where I just and I I was talking to my business coach and it just kind of came up that it's like, you know what? Truth is, I've fallen out of love with my business. Now, when I first got into business, it was I wildly passionate about it. I in fact, My ex-wife, and maybe this is why. why she's an ex. on date night, I would want to go hang out at my office. Like, hey, can we just drive by the office real quick? I didn't need to do anything. I would just want to go sit in there and and look at this thing that I had built. You know, I I loved my business. And over time, that love faded to the point where I hated it. Like I just I loathed it. I didn't want to get up and work. I didn't and especially the technology part, right? I loved business. But I didn't really love technology anymore because it got stale, it got boring. And this is my confession, right? That now let's talk about the the revitalization. AI has breathed new life into me, into the world of technology like nothing has in a long time. You know, back when I loved things, I would on on days off, family gatherings or with friends, I would want to talk about computers, the new processor that's coming out, the new graphics card. You know, that stuff was just wild. I loved it. And then everything just got boring. We've been working with the I5, the I7, and okay, the I9. That's the newest thing we've had in the last decade. I don't know. Like everything got boring and stale. It it just lost all of its excitement for me to the point where I didn't think I liked technology anymore. I just liked business. And now, since I've started coding with AI, I'm I'm it's it's just rekindled that the love that I had for coding. I was a computer science major in college. And I'm I'm going back and I'm I'm remembering stuff. I'm learning new stuff to the point where I've even spun up some Linux boxes. I'm looking over here at my lab that I'm building. I've never had a lab. My technicians have always had a lab. They go home and they nerd out. I'm like, get a fucking life. But now I'm the nerd. I'm sitting here, I went and hauled in some old shit from the garage that an old server that I had archived years ago and just haven't thrown away yet. And I'm I'm spinning that up. Bryan Lachapelle (37:34.232) Yeah. Yeah Justin Shelley (37:48.465) by next week, I'll probably have a dozen computers sitting in here in my little home lab so that I can get back into building stuff and understanding stuff on a level that I haven't done in a long time. All of my focus has been on sales, marketing, finances, business growth, business development. but the love of technology has come back in a way that I had no idea was even still alive. so I'm I'm I'm I'm loving it. And of course, I've mentioned before, my work days went from, you know, a good solid eight hours to, I don't know, twenty. It's like it's all I do now. I have to like remember that I have a family, you know, that, I have to eat lunch today. Shit. so I mean, I I don't know. Yeah, this this is my journey, my story, and and I'm leaving a lot of stuff out. A lot of stuff has happened in my life that we're not talking about right now. but it is Bryan Lachapelle (38:25.102) Yeah. Joshua Holloway (38:25.774) That sounds about right. Bryan Lachapelle (38:34.788) I've never been so tired. Justin Shelley (38:45.858) It is one of the unexpected results of AI for me. And I do think that this can play into a lot of us in the business world, is that this it it breathes life into places that have gone dead or stale in business. It allows us to do things like, you know, this this employee that you're talking about, Josh where a company was frustrated. They had a bottleneck. They've got somebody doing a job that they're not supposed to do, they're not trained to do, they're not qualified to do, or whatever. I'm not, I'm not speaking for them, but that's kind of what I'm hearing. And then it it takes that. that pain point and turned it into a new opportunity that they didn't even know existed. Right. Now that's again my interpretation of your story. And I'm not trying to speak for them, but it it does kind of make me think through what I've what's happened to me as I've dove back into a into technology in a way that I haven't done in probably 15 years. So that's it. That's that's kind of what came up for me this week around the the world of AI, vibe coding, technology at large. Jesus Christ, I'm spinning up Linux boxes. I never, ever, ever thought I would be doing that. so guys, let's go around the room. Let's close this out. Any any key takeaways, any final thoughts, anything that you want the the listeners to know and understand about where we're at in this AI journey, where you're at in your AI journey. we're gonna go in reverse order today, the people with the most hair all the way around the room to the people with the least hair. Josh, that makes you first. Hell no, he's not. Joshua Holloway (40:10.318) All right, Mario's up. No. Well Justin Shelley (40:14.328) He's more polished than I am. I think he goes last. Yeah. Mario Zaki | Mazteck (40:16.878) Yeah. See the the glare? Joshua Holloway (40:21.058) You know, takeaways from this one is for any businesses that that are feeling that they're they're stuck or they don't know where to go or they have processes that people just seem to be band-aiding or anything like that, this is that time to sit back and think about where can an automation or an AI help them have those conversations with your IT company. If your IT company, because there are ones out there that are anti-AI, they they don't want to do it, they're they're warning everybody against it. Justin Shelley (40:45.658) They won't do it. Mm-hmm. Joshua Holloway (40:51.522) Reach out to us, reach out to somebody who embraces the AI, who has learned it, learned how to put the guardrails on it and analyze job roles, analyze positions, and see like what processes can be automated. Your businesses will thank you silently in the background. And I'm sure employees who feel less pressure on their back because certain things have now been automated that weren't before, you know, they can breathe a little bit more. Maybe they become more productive. Your bottom line, you know, you know, gets a lot better. And the biggest takeaway is don't be afraid to analyze all of those jobs, all of those job roles, including your own, and see where you can improve through automation and AI. Because a lot of those, a lot of those issues or gaps will pop up as you begin to look at your job. I mean, I I think about it for myself too. I receive 800 email messages a day. I have an AI now that looks for the top five I need to look at, and it just periodically pings me and saying, hey. These are the top five you should look at before you sit down and go through all eight hundred messages and clear them all out. So review your job roles. Justin Shelley (41:56.974) Okay. Brian. Bryan Lachapelle (41:59.407) All right. I've got two key takeaways. The first one is dream. The sky is really the limit right now. There's hardly anything in a business that we can't either help automate or help improve. And number two would be there's a lot of things that your employees, you yourself are doing that you hate doing. You absolutely despise doing them and we might be able to help there. So what I would say is there's a document in EOS that is called the Delegate and Elevate Sheet. And in that sheet, you have something like, you know, top left corner, things that you're great at and you love, top right hand corner, things that you're good at and you like. Bottom left would be things that you're good at, but you hate doing. And then the bottom right is things that you hate doing and you're terrible at. That bottom half of the sheet, if you were to fill that in with everything you do the entire month, Every time you start a new task, you write it in one of those four quadrants. That bottom half of the sheet, that's the stuff for AI. If we can find a way of automating that and or creating systems that will remove as much of that friction as possible, your business will skyrocket because that's the stuff you are avoiding every single day because you hate doing it. And what do we do with things we hate doing? We don't do them. We do the stuff we like doing. So let me and let the rest of the four, three folks here, automate that stuff for you. Let's get rid of it off your plate so that you can do the stuff that you love doing. And maybe you'll be in the same position that Justin is, you know, loving his business again. Justin Shelley (43:38.533) Well, it it's interesting if I had done that delegate and elevate, I know that form well. if I had done that a few years ago. All the stuff that I love doing now and that I'm good at would have been on the bottom. It's like stuff that I hated and avoided. So kind of flip that on its head. so keep your eyes open and and and be open minded. You never know. Brian, before I let you off the hook, you are on the hot seat next week. Do you have anything you would like to tease about what's coming up next week? No details. Bryan Lachapelle (43:43.757) You Bryan Lachapelle (44:06.988) I'm not gonna tease anything. Yeah, the big reveal will be next week. Yeah. Justin Shelley (44:08.164) Or you want to just keep it a secret? Big reveal next week, okay. Okay. All right. Fair enough. Mario, you're Joshua Holloway (44:15.682) You can't give us like what's the the the X amount you may have saved or productivity? Justin Shelley (44:19.152) He hasn't He hasn't built it yet. Just kidding. Bryan Lachapelle (44:22.07) I can, it's, it's, could be, an unlimited amount. Like, know, you guys are talking about one times to 50 times. I'm talking about one to 50 times times one to 50 times, right? Like we can exponentially increase our productivity across all departments. Yeah, exactly. Yeah. 50, 50 plus 50, 50 plus X. I'm saying 50 X times 50 X. Justin Shelley (44:35.994) So like Mario's math. Mario Zaki | Mazteck (44:38.928) Mm. Justin Shelley (44:39.866) Plus X. Plus X. I don't Joshua Holloway (44:40.898) Yeah, one time fifty plus. Mario Zaki | Mazteck (44:42.854) Posso. Justin Shelley (44:45.412) I don't know the mathematical I so you're doing you're doing exponents. Mario is doing some new plus X thing that I haven't heard of, but anyways. All right. Well looking forward to that one. And you know, a a double tease the week after that we're gonna finally reveal or unveil the the unhackmybusiness.com site. It's up right now, it's hosting all of these episodes and it has a lot of cool features. and I'm just gonna put a little plug in there, then it's your turn, Mario. Bryan Lachapelle (44:53.018) Ha Yeah. Justin Shelley (45:13.73) but go in there, pick an episode. It doesn't matter which one, right underneath the audio and the video players is the two forms that pop up. Or you can just send us some feedback if you want to just chit chat. And if you want some help, there's another form to book a consult and and we'll just kind of distribute those amongst us. little little shameless plug, but then we're gonna get into the logged in session and what goes on behind the scenes with that. when it's my turn in two weeks. Mario, your final thoughts or your key takeaways. Mario Zaki | Mazteck (45:42.482) Yeah, for me, you know, I I'm I'm sure you guys will agree. getting a new customer right now, getting, you know, growing the business, especially the last year and a half hasn't been as easy as it was. You know, you know, it it's it's a little harder, people a little more slower to to make some moves and stuff like that. So growing the company, making more money is is a little harder than it used to be. then You know, the next best thing is really seeing where you can save money, either with subscriptions or salaries or you know, it sometimes you you may want to hire somebody and stuff like that. Using AI can really, you know, to you know, you can you don't need to sometimes reinvent the wheel. You could kind of automate something or recreate something that you're per you know, purchasing now internally that will make, you know make it a little easier. If you're able to save, you know, a subscription to something that you're you're paying like maybe three thousand, four thousand dollars a month, you know, that turns into, you know, what, thirty-six thousand or forty-eight thousand dollars for the year. something like that could end up being almost equivalent to just signing a new customer. You know, i i if you can't bring it in, you know, save it from going out. And AI is the perfect tool to for to do that. you know, we've done several different platforms, some that have replaced others and some that have kind of been on its own and it has made, you know, things a lot easier and I've able to save some subscriptions to certain things that I kind of used occasionally but didn't really need. that you know it i it's been able to save us you know some money besides you know the money that we talked about in in salaries and time spent on a on doing certain projects. Justin Shelley (47:43.022) I'm gonna I'm gonna add a little bit to your math there. So if you, you know, we we're always trying to gain customers, right? That is the goal of most businesses to to grow. You put $3,000 on the top line revenue, you're lucky if you bring a few hundred bucks to the bottom line, right? But if you can cut $3,000 and in you actually save that money, that goes straight to the bottom. You th so you save $3,000, that is $3,000 right there on the bottom line. so you can't you cannot save your way to wealth, but you know, you you you have to grow. But where we can cut, you know, trim a little fat here and there, that makes such a dramatic difference in the world of business. So guys. Mario Zaki | Mazteck (48:19.547) Yeah, I mean we we've discussed it. I mean Josh, I know me and you have talked about it a couple of times where, you know, I I don't wanna name some platforms, but there are some platforms that you you've recreated that you're like, Yeah, I no longer need this, you know? Justin Shelley (48:33.134) Yeah. Yeah. Joshua Holloway (48:33.186) Yeah, absolutely. Yeah, absol absolutely. Bryan Lachapelle (48:33.54) Mm-hmm. Justin Shelley (48:38.094) And you can use AI for that. connect it to your financial, you know, Claude. Is it Claude or Chat GPT? I can't remember. It's always telling me, hey, let us analyze your finances for you. like carefully, you could have it look for your subscriptions because how many times do we sign up for something that we never use or we forget and it just comes out of our account? anyways. All right, guys, that's it for this week. We're gonna go ahead and wrap up. Thank you guys all for being here. Listening, audience, please go to unhackmybusiness.com. Mario Zaki | Mazteck (48:39.909) Mm-hmm. Joshua Holloway (48:50.766) That's scary. Mario Zaki | Mazteck (48:51.559) Mm. Justin Shelley (49:07.618) Listen to the episodes, look at the key takeaways, add them to your little roadmap. That's one of the features on there. You can just hit a little plus to say, hey, I want to do that. And it'll go on a roadmap that we're building for you. or fill out the form and connect with any or all of us and let us help you along your AI journey, keeping you safe and secure along the way. all right, guys, that's it. Thank you for being here, and we will see you next time on Unhacked. Take care. Mario Zaki | Mazteck (49:35.281) on her. Bryan Lachapelle (49:36.224) Unhacked. Justin Shelley (49:36.748) Unhacked. Joshua Holloway (49:36.844) A neck.