He Used AI to Close 20% More Deals + Get Clients Excited for QBRs

[00:00:00] We've only done 20 deals with this so far, so this is very fresh, but we've won 13 of them and we've lost seven. So these are deals that have gone all the way through. That's 65% win rate. The moral of this story is you've got to insert AI where it makes the most sense to insert it. And for us, business-wise, for projects and those kind of initiatives, was at the very, very front end to basically put us in a position of very quickly determining, is this a go or no go?

[00:00:28] It doesn't give you 100% win rate, but comparing a 47% win rate, which is the old way, to a 65% win rate, and then you've got to think of all the opportunity costs I didn't incur chasing down those false leads. That's pretty good. Hey guys, Damien Stevens, host of MSP Mindset, founder and CEO of Servosity. Today I am blessed to be joined by Arlen Sorensen.

[00:00:57] Now, John used AI to tackle technology business reviews, and he had a really interesting problem. Zero percent of customers actually cared to read their report. He went from zero to 34%. More inspiring, he went from hours, five hours on average, to minutes per report with the assistance of AI.

[00:01:20] And then the real ROI came from his deal rate going from 44% to 65% win rate. If you want to know how John thinks about AI, some of the upsides and some of the risks, don't miss out on my conversation with John today. John, welcome to another episode of MSP Mindset with you. That's a rare opportunity that I get to interview somebody more than once, certainly more than twice. I'm really honored, Damien. Thank you.

[00:01:50] Yeah, I'm excited. Well, let's dig right in as we're talking about actually getting return from AI. And you have something super fascinating and it's pretty impactful. And I want to start with something that's still impactful, but maybe a little easier to understand and easier to pull off. So we were talking a little backstage about how you set up an app. I think you said maybe cost you 50 bucks a month. And it sounds like a lot of that's because you're hosting yourself in Azure.

[00:02:17] Tell me why and what that does. Like, why did you decide to do that yourself? Sure. I mean, one of the things that I want to set, you know, a little bit of table stakes here. I mean, or set the table on.

[00:02:31] We've been loving ChatGPT and Claude and all these generative tools, but we've been very cautious about allowing them unfettered access to our ticketing system, Slack, emails, QuickBooks, all these other places where there's a lot of data that we would love to have AI have the context for. But we've been uncomfortable using a publicly hosted one.

[00:02:59] Even with all the assurances they give, we feel a lot better about having a platform that is hosted by us that does API calls to use an LLM in a way that we can control, audit, and measure. And so with that as a kind of a backdrop, back in Q3, we started, Q3 of 2025, we started building this. And initially it lived on a laptop just to, you know, proof of concept. Like all good projects.

[00:03:29] Many development projects are like that. Yeah. And so once it looked like it was really working and we had something that you could query and get results. To be honest, the results weren't that great, but they were, it was clearly doing its core job. We then went to like kind of version two of it, which is really starting to figure out how to get vector stores working and get the data that we wanted to be more likely to show up in the consideration for our queries.

[00:03:58] And then once, then the third iteration was moving it off the laptop, getting it up into Azure, getting it hosted, which is, you know, it's its own little new problem now. And then the fourth version is what we're working on now. So we've changed from one very, very cheap LLM model. I think it was the ChatGPT 4.0 Mini. And now we're using 5.4 and it's much, much better.

[00:04:28] But what really made it better was we understood, because we went through all these little versions, what we needed to pay attention to. And the main thing we needed to pay attention to was the instruction set. We needed to give it a lot of really good contextual information about what we want and what a good looking output is.

[00:04:48] And so when you're using, you know, Claude or Gemini or ChatGPT on, you know, in their web portals or mobile app, they've already provided some core instruction set. When you build it yourself, you've really got to provide a lot more. But the benefit is you get to give it a lot of specificity. So for instance, and I know I'm going on a little bit here, but I think this is really interesting. We had a problem that, yeah.

[00:05:15] So in an ITIL format, a problem is a bigger issue. Lots of little tickets spawning, things going on. But it wasn't identified as a problem ticket yet in our ticketing system. But you could, a human could spot it. This is, there's a bad trend forming here. What was great was we could then go to our tool and say, here's a ticket. Show me all the other tickets that look kind of like this. And we were able to get those results back.

[00:05:40] And it found stuff that even when we were, we had a few humans trying to like crawl and find the tickets on their own. It found tickets that we didn't find. And they were tickets that we should have been able to find, but we couldn't for whatever reason. So we were seeing some real use cases. I could go on and on, but having our own system was great. So this thing that you built, help me understand, right? This like, is this something you go to a web portal? Does it live in your chat? Yeah. You know, Teams or Slack?

[00:06:10] And like, what is it? What does it do for you? So it looks and feels just like ChatGPT. You go to a web portal. So it's got its own URL. And what you do is you pick which kind of assistant you want. So we've designed different instruction sets depending on your role. So we've designed three separate instruction sets. One of them is you're going in as a support center technician, right? And so it's going to lean in really hard on technical details.

[00:06:39] It's going to assume that you're troubleshooting an issue, researching something hard, and going to give you loads and loads of tech tips that are pulling from our documentation system and past ticket history and notable Slack channels where troubleshooting tips are discussed and escalations are resolved. There's another one that focuses more on account management. So it leans much more into the kind of outputs that give you something that you could honestly deliver verbally

[00:07:09] or in written form to a customer. So it's much more translating whatever problems or whatever questions you might have into something that's more customer facing. And for whatever reason, I'm blanking on what the third is, billing. So the other one is related to almost anything that happens. It's going to talk about whether this has a cost associated with it, customer facing our costs, cost history, et cetera.

[00:07:38] So those are, in our mind, the three main vantage points that we typically want to query our data, but across multiple systems. So I think everybody wants some version of that. How did you decide what to start with? And like, did you just hook it up to ticketing first and later add billing or PSA? Or like, you know, do you, did you decide we need to connect it to every system we have to have value? That's a great question.

[00:08:07] So I want to make sure everyone knows I've got an abundance of riches in my MSP. I've got an integration team with three super smart coding developer folks who meet with me every couple of weeks and meet with our COO much more regularly than that. And so we had one of our meetings back in, you know, I guess it was July or August of last year. We were like, okay, we need to respect our acceptable use policies.

[00:08:36] We need to make sure that we put guardrails and governance in place. And the way we could really see ourselves achieving that with any AI system that has access to our data was we needed to build it internally. So that was number one. Number two, we sat down and said, well, what are our business goals? How would we use this thing? Give us some, I mean, we just kind of talked through like common use cases.

[00:09:04] We could see ourselves needing to go beyond just a normal search bar within our ticketing system or within, I mean, QuickBooks actually has a really good search bar. I find it very easy to find things in QuickBooks using it. So it was really about, well, what's, what is going to go beyond what I could just query? Because I mean, I can go to a customer in QuickBooks and say, show me all the billing activity I've had with them. That's not, that's, I don't need AI for that.

[00:09:34] What I need AI for is like, when was the last time we build them for this thing? Because I know we stopped billing them, but I don't remember when. And it can surface that very quickly. So those, we started going through those kinds of questions like, okay, yeah, that, that makes sense. That would be something where AI would do a really good job sifting through the pile of data and surfacing what it could find.

[00:09:58] So did you come up with like a, a funny, fun name to call this magical service? You know, it's really funny. We, we have not. It is, the URL is also atrocious. It is, I think, I think it's over 150 characters long and it's like, you know, it's like an Azure asset. It's, it's really ugly. It doesn't even end in netfriends.com. I think it's really bad.

[00:10:24] So, so no, like it, we've been so focused on just making sure it worked. We haven't focused on any of the marketing machine. It also, it's really funny. You can tell you're the first one to use it in any given week because it has to like wake up the Azure. Azure unit. So like, if you're the first one in, you're like, ah, dang it. You know, and you have to go get coffee and come back. I mean, we were trying to run this thing on the cheap because we actually were afraid that we would get the bill and be like $30,000.

[00:10:52] You know, so we were like really freaking out about allowing it free reign. So, you know, as much as I feel like we're. So you've been running this for months. What is that? What is that? This, this thing that can find all these things for you typically. So far it costs about 150 bucks. And since October. Yeah. So seven, eight months. It's, it's, it just doesn't feel right. It's so wrong how cheap it is.

[00:11:19] Um, now, now the workload that it's taken to build it is, is non-trivial. Um, sure. And, and we're about to do something that I think is going to increase costs because we're, we're about to build a fourth agent, which is, uh, we call them assistant. Actually. So we're going to build a, uh, reporting assistant. That I think is going to crank the dial up 10 X or more on cost.

[00:11:46] Granted, I'm starting from a tiny place, but, um, we're going to be able to. Right now it does not take new inputs that we drag and drop into it. It only is querying and pulling stuff that behind the scenes roost, which is our automation platform is pulling and gathering, uh, from all our different systems and kind of keeping in a local store.

[00:12:07] Um, so once we allow it drag and dropping, uh, random PDF files, other assets into it, I think it's going to increase our token usage way, way up. And then I also think we're probably going to start asking the report agent to produce more rich content. Cause right now it's just spitting out, you know, some text, um, that I think is going to increase our token, our token usage as well. Gotcha. Okay.

[00:12:36] Well, what I love about that is everybody has that problem, you know, that where is this and who had this and when was this build? And we thought we could find the tickets that we can't and all of those. Um, and so many people, the examples are like the best examples. And we're going to talk about it. You've got some really amazing ones, but I love that we could start with one that anybody really could afford, um, to, uh, run. And as long as you put the time in to figure out how to do that.

[00:13:02] And, um, I'd like to add one thing that is kind of a pro tip. If you do plan on going down this route and you feel like you don't have what I have, which is, you know, a team of software developer coder type folks who can do something like this is, I know vibe coding and all that stuff is, uh, maybe not necessarily, you know, smiled upon.

[00:13:26] And then, uh, you know, it's starting to gain a little bit more acceptance, but you absolutely can iteratively build this through talking with AI, especially cloud code and just say, okay, how do I approach this? I am certain our team use cloud code a ton in building this out. So, um, I, I do think it's more accessible to everybody than, than they realize. Yeah. Yeah.

[00:13:52] And that you'll spend more once you go start using it, but that's 20, 25 bucks a month to start with. Right. So that's, it's not a, it's not crazy expensive. Um, so, uh, I want to, I want to get into the real juicy one. I, from my perspective, right. Which, um, I think there's a few ways to frame this, but, uh, I don't know what I would call the, we were talking about this VCIO unlock and you were saying that previous tools, especially around reporting had the worst outcomes.

[00:14:22] Um, nobody really read the reports. Nobody wanted to meet with you. Nobody's engaging on the things you guys highlighted. Um, and anybody that's listening probably can relate. Oh yeah. Right. We've, we've put those out and done that. So what changed? What was the aha? So let me just say that we've tried three different VCIO tools. I'm going to not name names.

[00:14:49] All of them, you know, made a really good attempt to try to crystallize all the things, right. To make it into a pretty report, make it into a slide deck or some sort of presentation assistant thing. The thing that we saw was time and time again, every single customer did not really want whatever these tools produced.

[00:15:12] Um, what happened is, is that when he started showing people the, the material, it's overwhelming. It's too much to take in. So everybody has a certain amount of absorption rate. And if I'm at like, if I went to like an investor conference and they just start going on and on about all these different places to put my money. I know it matters. I want to pay attention, but I just hit my saturation point. And then I'm just like, when is this going to end? And I think we put our customers through that.

[00:15:42] And, you know, almost anybody who would, we could corral into a meeting. Oh my gosh, we, to get them in the second meeting was nearly impossible. I mean, they would dodge and duck every single way. Um, and then we would send out reports on a regular basis and they're coming from our PSA system. They, they looked good. They were, I really emphasized like one pagers. Let's make them short and, and not try to overwhelm people and just crickets.

[00:16:11] They would go out. I would ask people, did you see him? And they're like, yeah. And I'm like, oh, what'd you think? And you could tell they were like, crap, he was not supposed to follow up on that. And, um, and so what we saw was we put a ton of time and energy into creating concise reports. They weren't moving the needle at all. No one, if they looked at them, they certainly didn't remember. And they weren't notable. We couldn't get anybody into meetings. And when we did, they weren't meetings I would, I would want to sit through either.

[00:16:40] So, um, I don't care how many jokes we cracked and, and, you know, casual conversation. So anyway, long story short, nothing's working, but I'm sitting on all this wonderful business intelligence informing data. Why do these businesses not have the time to absorb this? Right. And, and, but it's on me to solve it. Right. It's, it's my problem. So here comes AI on the scene.

[00:17:05] Now on one level, nobody wants to read something that they know AI generated on one level. Right. Um, however, if the AI can give them the headline that makes them go, oh, wait, wait, that's interesting. And then the headline is followed by right there surface at the top and in the right kind of color schema. So it's visually like arresting and it draws your attention. Okay.

[00:17:33] The AI just got me what I was kind of, it piques my curiosity and satisfied it. Boom, boom. And 10 seconds or less. Now you've got somebody who is energized by the information and wants to respond. And so I promised you when we last talked that I would like run the numbers and say, okay, do I have the receipts that prove it? So, um, so what we've seen is that when we generate these new reports and we've been generating

[00:18:02] them since February of 2026. So I want to, I want to make sure I don't have like a year's worth of data on this yet. Um, but what we saw was we, we picked a subset of our customers that we thought need a TBR. So these are our biggest customers. These are the ones with the most complex data and the ones who we know want to engage, but have not engaged yet. So using that subset of customers, there's 38 we identified.

[00:18:31] 13 of them responded within minutes of getting this, this AI report unprompted by us. We just sent it hoping for the best. So that's a 34% response rate when I was getting zero before. And what's amazing. One of the responses we got four pages, that guy wrote four pages of just, and it wasn't complaints.

[00:19:01] It was like, this was interesting. What does this mean? I need more information about this because I, if this is true, I need to do X, which is exactly what we want, right? A TBR is supposed to move the needle, supposed to make people want to go deeper. Um, we were seeing people say things like, this is awesome. I'm going to show this to our CFO. Um, and, and we had people asking for meetings with us. Can we talk about this? Are you free later today?

[00:19:31] I mean, we were like, wow. Okay. So it exceeded our expectations beyond like, we were hoping that it would at least be like, thanks. But like, instead we were getting some incredible traction and it, and it persists. So it wasn't just on first blush and they're like, oh, this is novel and new. Here you go. The next month we send it, we get similar reaction. Um, so this is really, really great.

[00:19:57] And what makes it even better is we, we were looking at like, okay, let's go back a year. How much time did we spend on the old method? Not only did we spend, you know, over a thousand bucks a month on the tooling for TBRs that nobody wanted to attend. We also spent somewhere between 500 and a thousand hours on it because it typically took like four to six hours to prep. That's the worst, right?

[00:20:23] You put all this energy into bringing all these bits together, critically reviewing it, making sure it's organized and, and tells the narrative and focuses on the things you want. And then you have a no show or you have someone who shows up, but they're like, actually, can we just talk about these two tickets that went sideways? And I would rather dig into that. And, and you can say, no, I'm going to push forward with this information. Or you're going to be like, yeah, I don't have your attention because you just want to talk about the tickets. And then they use the whole time talking about the tickets.

[00:20:52] So, so now we're having people want to talk about what we just sent them. The conversation's on our terms and it's taking us all those reports in totality, sending them out for the last three months. It's taken us around 60 hours total. And that's including tinkering and tweaking and trying stuff out. I mean, it's just, it's amazing. So I know I went through all those numbers really fast, but, um, I can unpack any or revisit anything there that you want me to.

[00:21:22] So I don't, um, I know, you know, one was across a year and one was across a few months, but do you know, you used to spend four to six hours to prepare one. What does it take? Do you know what it takes now? Yes. It takes somewhere between 10 and 15 minutes of CLAWD time. Meaning we put our, we, we, we have a, a repository where all the data goes. We just drag all that stuff over into CLAWD.

[00:21:50] Now this is CLAWD for teams. This is our, you know, internal, it's not training the LLM and all that stuff. So we, we drag and drop all that information over and we copy in our instruction set that has also been honed to give us the output that we want. Cause you have to give it the constraints and the goal and the, all the specifics. And then we walk away and it takes 10 to 30 minutes.

[00:22:17] We have three different people generating these reports so that we can generate them quickly, three different user accounts. So that way we also spread out the token consumption rate. Cause we do hit the wall per user. That's helped a lot too, which also shows that what we're doing, it scales person to person. It's not just relying on the memory of one, one person's AI. So that after that 30 minutes, let's say it's worst case.

[00:22:46] It's a big client, lots of data. After the 30 minutes is up, the output is so consistent. We almost never need to make any changes within the AI. What happens is, is we export both the email. Cause it has a, here's the email that's going to introduce the report that, that gives the one to two highlights. And then we have the attachment that we put on the email. And, and we put all that together. Cause, cause in the process of copying and pasting that email in and adding the attachment,

[00:23:15] we're doing a quality control review. And you can see, as this goes on for a few more cycles, it's now time to get one of those assistants queued up and make sure that these instruction sets and everything else works with the open AI LLM instead of the anthropic LLM, which is another hurdle we'll have to look at. It may, it may be that we have to become multimodal, but we'll get to that. We'll cross that bridge when we, when we get to it.

[00:23:43] But the thing is we're saving so much time right now. We're still in that honeymoon period of not needing to optimize yet. Cause we're like, wow, this is already going so great. We've saved so much time. You know, honestly, all the tinkering has been about, um, trying to avoid having to pay anthropic extra fees because, cause, uh, you know, and, and we've only had to pay an extra a hundred bucks over these last, uh, three months. Um, and are you on like the hundred or $200 a month plan?

[00:24:12] We're on the, we're on the, is it 25 or 30 per user plan? Yeah. Um, and so as a company, we've only had to pay an extra a hundred bucks for the additional processing because it's wild how, how far 2 million tokens goes. Yeah. Yeah. Yeah. That's huge. Right. So, so here's the thing you were spending. I don't know what the math was $20 a month before on the tool you guys use internally. And I think that's what most people should do. Cut your teeth internally.

[00:24:42] You can QC it yourself. You're not necessarily risking sending out something that's wrong. Let's face it. Basically wrong. If you don't catch that. And, um, but if I understand what you're saying, you're talking about $25 a month per user. Maybe at the max, if that's all you did was, it'd be like $75 a month. And, um, and $100 a usage across all this time. Right. So you're, you're really not spending even $100 a month.

[00:25:10] Uh, if I understand you and you said something like it takes clawed on the high end 30 minutes, but that's, you can walk away or you can do other things while that's kind of cooking. So, um, what is, how long does it take in human time? Cause there is a QC there, right? You don't want to sense that. Do you have any ideas that like hours or is that?

[00:25:34] It's about 20 hours a month is in total of human time spent staging and doing quality control and delivering. Wow. Now that's only for 38 clients. So when we expand it, we do expect there to be some linear element to that.

[00:25:56] Um, so in other words, if we send this out, um, to 80 clients, I actually do think it's going to be closer to 40, not closer to 30 hours. That's the linear thing. Cause I do think that everyone who's working on this, they've all been involved in customer success and account management and communicating with customers. They know the drill, uh, they're, and they're working fast.

[00:26:20] So I don't think there's any additional scaling that can happen unless we make this happen with more automation. So, uh, and that's, and, and honestly, I'm not putting time and attention into that just yet because we're only three months in and we may find that there's, um, some really novel ideas that are going to come from these meetings that we're now having with our customers and these conversations we're having with our customers.

[00:26:47] And so we want to leave space to continue to iterate a bit. And that's another thing too. Don't rush to package this stuff. Stay in this kind of loosey goosey, you know, human in the loop. Not just because it's good for governance, but it allows you to continue to be very plastic and flexible while you're still optimizing so that it works for, you know, as broad an audience as possible. Yeah. That makes a ton of sense. Yeah.

[00:27:14] We have a, we have a few teams, but software development team. And of course they're used, you know, we, as you mentioned earlier in our pre-show, we do backup at DR so that everything has to be packaged very, very carefully. And in very specific ways. Um, and so there's some differences in culture versus some internal packaging that, you know, will be human in loop versus the, the opposite side of that is like, you know, software, this mission critical. Um, right. So very, very different.

[00:27:45] I want to kind of dig on this a little deeper. This is amazing, right? First of all, you got somebody to read anything nowadays from an email, right? It's hard. Um, 34% is really, really amazing. I want to stay with you and see if that holds as you continue to scale that. But I want to, I want to call out a point where I think you said something like, um, what do we call the resentment economy? Right. You, and we were talking before, because this is the raw truth that I love that you share, John.

[00:28:15] The traditional proposal takes so much time, right? Once I've got four to six hours in it, once I send it over and then you just don't read it or don't care or say nothing. It's hard if they ask you for an estimate, right? Or something like that, um, to not, well, just to want to respond to run it, right? Do I want to invest a few more hours doing that estimate or quote or whatever?

[00:28:41] Um, what kind of difference have you seen? If any, like just, I don't expect numbers from a human perspective. Yeah, so to unpack the resentment economy, um, speaking to anyone else who's in our industry, um, who get amorphous requests from customers or pretty, pretty straight up requests. Um, or we spot something that really has to be done, has to be changed.

[00:29:09] Um, we see a lot of customers kind of flippantly sending us down this pathway to have to build out and scope something. That four to six hours you mentioned, I mean, heck, we see some scoping engagements that go on for weeks that involve, you know, a dozen techs and, and on-site visits and all this stuff. And, you know, those are usually the projects that don't get approved for reasons.

[00:29:39] And, and no one tells you what the reasons are. You just, they're like, no, or worse. They're like, you know, I really appreciate what you put together. It really helped us get this competing quote for less. Like what? Ah, so, um, so what we've to, what we have found is that we should never scope any project until a customer has seen a rough estimate and a rough, you know, couple bullet points.

[00:30:08] This is what you wanted, or this is what we think you need. And it's going to cost this much. Do you want us to formally scope this? That's a, forget AI. That's just good business advice. Um, we, um, I was looking at like how many deals have we put together? Um, the number of deals. And so we think of any project as a deal, the number of deals that we've put together.

[00:30:35] Oh gosh, I don't even have this in like clean years, but it looks like roughly, we're talking about roughly 400 quotes every single year. So it's a lot. So if you, you, you know, not every single one's four to six hours, but it's probably on average about four to six hours. So that's a lot of labor. And, and when you lose all that, when, when someone says no or not now or something like that, it does hurt if you put a lot of time and attention into it.

[00:31:04] So, um, so looking back at, I asked, I asked Neil, which my partner who really oversees everything sales to like do some analysis for me. So he looked at 688 completed deals, deals that we took all the way to the end and we either won it or we lost it definitively. Um, and so when we, he looked through all this, he found that when we were just doing a normal

[00:31:34] scope with no estimate whatsoever, that, that covers about 264 of those deals. We won 116 of them and we lost 148. So that meant we, we had a win rate of about 44%. It's not terrible. You can argue that's within industry norms or maybe even better than industry norms, but

[00:31:59] I'm looking at those 148 and the lost revenue, the opportunity cost of having techs go deep into a solution. And then, like you said, the resentment that you can feel because I can guarantee you several of those losses were at the same customer who sent us down multiple rabbit holes, uh, maybe not even realizing they were doing it. Yeah. Yeah. Yeah. So here comes, uh, the idea of like, maybe AI can help here.

[00:32:26] So the first idea that we had was let's at least use AI to help us scope faster, right? We're doing a lot of similar things where there's recyclable components, reusable components. Let's try to always do scoping and run it through AI and get better at articulating the scope and increasing the speed, trying to get same day scoping. So we're turning it around fast. Still takes a lot of work.

[00:32:50] Um, as much as AI can assist, you still have to go out and assess the site. You've got to dig into their data and, and make determinations of what's possible. Learn about the newest and latest thing that Microsoft has created that may or may not help and all that stuff. And that did help. We, we had, uh, 310 deals that we processed through using this AI assisted review of those.

[00:33:18] We won 146 and we lost 164. As you can tell, we're still below water and, uh, 47%. So just adding AI into the mix, the way I described it here on it. If I was sitting in a, like a armchair theorizing, I would have expected more than a 3% shift. It was just a little. Okay. Bummer.

[00:33:43] So then, um, we, we really went in and said, okay, let's add this estimate step in. And when we had added the estimate step in, what it allowed us to do was, you know, not just same day, but I mean, crazy quick turnaround. We've had the meeting or they've gotten the report and it's, it's time to just flip them and say, well, this is what it's going to cost to do X. That we've only done 20 deals with this so far.

[00:34:13] So this is very fresh, but we've won 13 of them and we've lost seven. So these are deals that have gone all the way through that's 65% win rate. So, so we added AI initially and we leaned on that for many, many deals, but it was still in the construct of a heavy handed labor intensive scoping process.

[00:34:35] So the moral of this story is you got to insert AI where it makes the most sense to insert it. And for us, business wise for projects and, and those kinds of initiatives was at the very, very front end to, to basically put us in a position of very quickly determining, is this a, is this a go or no go?

[00:34:57] Um, it doesn't give you a hundred percent win rate, but comparing a 44% win rate, uh, or 47% win rate, which is the old way to a 65% win rate. And then you've got to think of all the opportunity costs. I didn't incur chasing, chasing down those, uh, those false leads. That's pretty good. That's huge. Right. I don't know what, what that costs you in terms of AI services, but 20, 20, as far as I know, just a couple licenses.

[00:35:26] Um, yeah, it's not nothing, no overages related to this as far as I know. Yeah, that's, that's crazy. Um, yeah, that, so that's what I'm getting. Again, we're in the space of 50 or a hundred dollars a month type of investments. Um, and, uh, and those you can afford to lose sometimes. And, uh, and, and, and, and, and waste that, I think. Damien, one, one thing, so I was telling the story, it started dawning on me. There's a data, uh, data element I don't have.

[00:35:56] And that is, I'm only tracking, cause I only asked Nealish for this, the number of deals that we proposed. So I don't know how many deals we deflected. I don't think that's tracked. I, and that, this is also something about AI too. Um, when you're innovating with your processes, you do need to have a KPI or success metric that makes you go, did this work or not? So I've got the success metric in terms of a win rate,

[00:36:25] but I've got a new metric that came up. A deflection rate. I did not, we're not tracking the deflection rate. Holy crap. I need to write that down. I love it. I love it when brainstorming, you know, it comes up right as we're talking about it. Um, the other thing I wanted to say is like, to me, this sounds a little bit, it's different, but it reminds me of, um, you know, cloud computing in terms of some of the early adopters where the cloud is great. Put everything in the cloud, just go for it.

[00:36:55] And I think the wrong way to do it most of the time was just take the legacy workloads and shove them in the cloud. Right. In most cases to have the right ROI and actually be more modern and better security footprint. And just to get a good ROI ended up being some level of replatforming. Um, so there's something that struck me about what you said, which is, you know, basically throwing it AI at whatever processes you have may not really help. Um, but rethinking them to,

[00:37:25] because deflection wasn't really a concept and not only that, but even the, um, that quick turnaround on the quote wasn't, wasn't a concept. That parallel is really, really apt because you're right. When cloud came on the scene, was it 2012, 2013 when it was really, really hot by 2016, you pretty much had to be in it or, or you were really, it felt like you were behind. You're right. We, what we did is we thought, Oh, it's kind of like VMware, but on someone else's hardware.

[00:37:54] So we just, we would forklift the entire virtual server up there or rebuild the entire virtual server up in the cloud versus using cloud native tools. You're totally right. And the, and the estimate, we never did estimates before. And so what we did is we took our old process and we just, we just added a little AI to it, which didn't materially change much. But once we used AI differently to think differently and to think of it, I don't think we called it, Hey, let's try to deflect projects,

[00:38:23] but that's ultimately what it was. We're so some, some of, I think everybody in my position needs to realize that you may not have it all figured out. That's okay. You can look back and get clarity in hindsight, almost like I just did live here on, on this. We're like, Oh yeah, it was all about the deflections. Cause we didn't want to work six hours that would make us feel resentful. Uh, we just didn't have a term for it. Yeah. Well, the other part is we always want the,

[00:38:53] you know, I won 20% plus more deals and that's amazing. But part of my question is like, would it have been valuable if you, you didn't go down, if you'd gone up from 44 to 47 and then that would stopped, would it have still been worth the 75 bucks a month? You might've been spending for some licenses just due to the, your win rate didn't drop and you saved a significant amount of money. The answer is yes.

[00:39:22] But keep in mind, you start down the journey on a hypothesis that my win rate's going to go up. So the fear that I would have is that if it didn't go up, then it, it would sound like I'm like tap dancing around the real problem. It was like, well, it didn't work out like I planned, but you know, so it was kind of nice to, to have the number I was staring at move in the direction I wanted it to go and move in a significant way. Um, but you're right. Uh, some of this, if,

[00:39:51] if we only had a 40% win rate, but we're working on a tiny fraction of the deals and, uh, that I probably could create a narrative around that. I got to say though, in theory, you want to know that when someone has approved the spirit and the budget for the thing, once you define it, they are also still ready to buy. you know, if, so I do think it was correct to stare at the win rate. I get where you're going,

[00:40:22] Damien, but I, I do think the benefit that I'm having here of thinking about all the things that we've deflected and the, the low cost of the, uh, of using AI considering how many, you know, dollars it, it, it costs for the per license that there is value in that, but you do want both. You want, you want to win on, on the metric you're staring at and on, on the savings from other, other parts of your business. Yeah. Well, speaking of figuring out like the, when you said it that way,

[00:40:51] I just wanted the platform cloud platform, replatform analogy hit me. It's not that I had this, uh, ready for this, but it just hit me. And my point is too many of us tried to stick the exact app in there and it became more expensive, um, or something like that. And so, you know, we, I remember talking a lot of mispeak. He said, I don't know why I can't get people to move the cloud. I'm like, yeah, but when you boil it down to it's the same, but it costs more, right? That's not a compelling, uh, argument.

[00:41:19] And so I do think the analogy holds in so much that so many of us are trying to do the exact same things we've always done with AI and maybe sometimes. Right. Right. But like, I think the art is saying, what could we do that wasn't possible before? How could we rethink our processes? And then we can invite this in and do something that is a much better outcome. Like your, but as a, as a humility story here,

[00:41:49] we, we did that when we worked on those 300 plus deals using AI, but it was like basically the same old way we were doing it before, but this time with AI and we didn't get much difference. We did that for a full year. And do you know why we stopped doing it after, after 12 months? It's because something changed with jet chat GPT and our results started getting worse. So it wasn't like, I don't want this to come across as, Oh, we've got all figured out. And we were so smart.

[00:42:19] No, we, we plugged away only getting a 3% gain, happily chugging along, enjoying that 3% gain until it actually broke on us. And we were like, Oh crap, we've got to do something different now. That's pretty typical. And I don't, so I don't want anyone to think like, Oh, this guy's such a pompous, you know, Oh, he's got all figured out. Like, no, we, we're, we did the same thing. We were, it probably took like, if you use your cloud metaphor,

[00:42:47] it probably took us moving hundreds of servers up into the cloud. And we were like, dang, these bills are really high. And then we're like, we should do something about it. So, yeah, I, I just want to make sure no one thinks that, you know, we went into this and like, within minutes of coming out of meeting, we had an 18 step plan that we followed to the letter and, you know, success and magic happened at the other side of that. I mean, it's, it's not that way. And, and everything is just about trying stuff out. And when something breaks, it might actually be a blessing,

[00:43:17] not a curse. Cause it makes, it forces you to go and rethink, well, how do I make it work again? So sometimes that's a, that's where the innovation happens. Right, right, right. Well, speaking of that, I want to talk about security is the permission slip. Um, and what I mean is you decided to kind of skip the vendor piece. Let me self host. Let me control more of this. Um, let me audit this. Um,

[00:43:46] and so I guess it's, you know, my question is, is, would you do that again? Is that the right thing? And, and how do you know where, where to draw that line? That's the right thing for us. So we are a SOC two type two audited firm have been for seven years. Um, we take information security very seriously and have since our origin,

[00:44:10] I was the one who eagerly wanted to be SOC two certified back in 2019 when we started the journey. And I sat there and I spent three weeks slogging through writing an information security policy that ended up being, I think it was, uh, 80 plus pages and it had 114 policy statements that were like the core underpinnings of it. Um, I mean, I'm,

[00:44:39] I take this stuff super seriously, probably too seriously. I'll be admit, I don't think anyone should look at me and say, Oh, that I want to be like that guy. I kind of went a little nuts. Um, and so what that meant is I'm probably on the more paranoid, side of governance. Now I'm very positive. I don't walk around, you know, with paranoia and, uh, you know, constantly plaguing me, but I've just, my,

[00:45:07] my walk through it services and it, um, um, has simply been fraught with security, uh, incidents, compliance needs. I've just encountered too many instances where someone did something cool and then got bitten hard because they, they were not careful. Uh, AI to me is like fire. We've just discovered fire and I'm excited about it.

[00:45:34] When I bring fire into my home, I realize it can burn it down. So I need to control it. And so building a fireplace carefully, only having the fireplace being there, having a very strong, like no one is allowed to bring the fire out of the fireplace is, is really important. That way you don't have someone going, it's so pretty when it burns the drapes down, you know, like no. Uh, so, so I've been really trying to keep that control because I'm, I have a,

[00:46:03] I think a healthy level of paranoia about it. And that paranoia is brought about by seeing AI move underneath you as these, uh, as the, the business needs of these, uh, businesses shift and change. They, they change their privacy policies. They change the product mix. Those things I can't control entirely, but there are things I can control. And, and, you know, I'm, I just,

[00:46:32] I just want to have as much control as I can reasonably get without, you know, living in a cabin in the woods and mumbling and muttering to myself. Yeah. No, that makes a ton of sense. Um, I know you, you wanted to ask, kind of turn the tables. You had a question. You wanted to ask for you, Damien. So, and I mean this in the best of ways. So if it comes across as, um, you know, probing and poking too hard, I don't mean it that way. Um, it's more of like,

[00:47:02] I'm speaking to you as, as I'm, as I want to speak to every vendor. Cause we, we have like a hundred vendors we work with. Right. And I know all of them are getting enormous pressure to innovate and use agentic AI or, or do something flashy with AI in your system. And what that means is two things. One, the obvious AI gets introduced to a platform, whether I want it or not, uh, often with an upcharge, um, that I get, I understand that. So I can,

[00:47:31] I can kind of see that coming, right? I can, I can, I can, and we can unpack that a bit more if you want. The second thing that I worry about a lot, isn't something that's being talked about enough. And that is every single vendor out there who wants to sell to an MSP is almost certainly doing some product development. They're, they're iterating their vibe coding on their own or, or whatnot. And my biggest concern, especially for you, Damon,

[00:47:59] cause I know you are very much leaning into agentic AI is my concern is, is how are you protecting me? How are you protecting the products that I've bought from you from being unintentionally infected, disrupted, altered in a way that actually you don't want either. Like you would be horrified if it happened. like, so I'll give you one example to like, you know, cause it's, you know, with backups.

[00:48:29] So imagine you, for whatever reason, you were designing something that was going to help me keep track of permission changes. Cause that is like a big bugaboo. Someone changes permissions, your backup tool can't see it anymore. It doesn't get backed up. And of course it's the confidential sensitive stuff. And then they're really mad cause it didn't get backed up. So let's say you design something that's trying using AI to try to keep tabs on that and, and address it live. So it fixes it. Unfortunately,

[00:48:59] you vibe coded it a little bit, and it actually opens up their permissions in a way that allows me to see other people's sensitive data and only their sensitive data for some reason. Like, Oh my gosh. You know, and it's a huge reputational hit for you. And this is a hypothetical. I mean, this has not happened, but I could see you with great intention, trying to solve a real problem you see, and you're, you're really excited about it. But unfortunately there is bleed through. How would, how do you, how do you assure me you can,

[00:49:27] you can lean on that example I gave you, or you can come up with your own, but I just love to hear from you. Like, what are your, how do you, how do you keep the guardrails from the core product that I've, that I've got from you? Even as a vendor, I encourage this, you know, ask me, ask your vendors, right? Because I think there's a new attack surface that we hadn't really had to think for. And nobody's really asked us about our software development lifecycle because they care. But if we were vibe coding, you would probably care.

[00:49:57] Right. And that probably wouldn't be what you'd want to see. So there's just different, I think, attack surfaces that could end up in there. And so from our perspective, we treat this like we treat security, which is, it comes back to security and it comes back to culture. So, you know, yes, we have guardrails and products and tooling that help with these things, but it's really about what are you doing? Do you know what you're doing? Do you know what you're saying yes to? Right. And, and I know this is, this is oversimplification,

[00:50:26] but it's just like everybody has the, did, was that really the email for me that you clicked on and, you know, granted approvals to or transferred the money. And, um, there's so many prompts. There's so many approvals now as there should be to kind of limit the blast radius with AI. It would be very easy to, to try to trick somebody doing that. The other thing is, um, look, depending on the moments I can be on the look, we've discovered fire. Now we can now, now cook food. And wow,

[00:50:56] isn't this so amazing? And other times I can be on the fire really does burn you. And we can talk about all the bad things, um, when it comes to AI. And what I mean by that is specifically, um, we've decided to treat it like an adversarial attacker. Um, and so I'll give you an example. I'll say something out loud that a lot of us probably aren't saying. You've given your team, we've given our team an access, whether it's chat TPT or a Claude subscription or a Gemini thing,

[00:51:26] your team has it. And if you don't think they do, they do. Oh yeah. Right. So there's some AI and our team does too. The thing is, um, what are you allowing them? You know, what policy is there and what do they know is expected to be done and not be done. And then the way that we're treating it is like an attacker because I trust the team that I have, right. They've built trust. They've earned trust over time. And an AI is not the same. And so here's the thing that's been counterintuitive,

[00:51:56] but it's helped us. Uh, hopefully you can trust Damien or I can trust one of my team members. You can't trust the agent I'm using, right. treat it like an adversary. Um, and so, uh, we have it just in the last month. We're a vendor. Everything we have, we built has an API, but not only of my team using APIs, we now have customers using cloud code and consuming our APIs and

[00:52:24] automatically deploying backups and doing wonderful things. Um, but that means you can shoot yourself in the foot with it because it's not even now. It's not just my team. Now it's people that are, maybe they know what they're doing and maybe they're more on the vibe coding spectrum. Um, and so I can't always prevent you from shooting yourself in the foot. But for example, we just increased the, the, um, like in our web portal, there's always been MFA. And for certain actions, there were, even with the API, there's always been MFA.

[00:52:53] We just rolled out and increased more MFA, which sounds in the day of moving quickly with agent to K I, the wrong thing, but we realize we need to, we need to have a way to prove it to him taking that action. And if it's somehow gotten access to your computer, whether you're an attacker or an AI, let's treat it the same. And we're going to need some out of band way to verify that you're a human. In our particular case, our MFA is actual Yubico keys. You have to physically touch and, uh, and those sorts of things.

[00:53:21] So I don't think we have it all figured out, but my point is we talk about this openly. And frankly, I come a couple times a month just to be direct and say, here's the way I've messed up as the CEO. Um, and when I say that it's not like, Oh my gosh, I let all of our customer data out. Not that, but like I tried this and then this did this thing that I did not expect. And here's an example where it's not that I was being arrogant. It's just, here's an example. If it blew up on me, it could blow up on you. Right.

[00:53:51] And so reminder, we're, we're trying to talk about both sides of this. There's an amazing upside and an amazing downside because sometimes it's like the brilliant person, the next person I was going to hire. And sometimes AI can be like that next brilliant person I was going to hire and do amazing things. I'm so excited. And then five minutes later, the same AI can be an adversary that felt like somebody hired to attack. And I don't even mean it's compromised. I just mean, you know, the wrong problem, the wrong access,

[00:54:21] the wrong access controls. And all of a sudden, you know, it goes away. I'll give you an external viewpoint on this. You know, OpenClaw has become the fastest growing software repo in history to give you bright perspective. There's more downloads or more stars than Linux over 20, you know, and it happened in the last few months, not 20 years. And we, we put it in a sandbox, we run, we experiment, we see what does this look like?

[00:54:51] But we don't, we haven't deployed that to our team because the value in it is giving access to lots of things. And the security track record with it is not great. And by the time you give it access to lots of things and you have really a troubled track record with history, you're, you're really in a, in a spot that kind of makes, I think like for us, it just, the cons are too high. Right.

[00:55:20] I see that we've experimented in the sandbox. We've, we've seen it. We've given it limited permissions. We've done this for that. And we're like, this is really cool. But to do that past a little trial sandbox, we're like, we, we can't, we're not sure we like, even with our APIs, that's going to do MFA before it does something harmful, like delete the backups. For example, that's a, that's a thing a human has to approve changing. Maybe the time, maybe not, but you know how, how often, you know, there's still issues like that. And so the reason I only bring up open clause, not to pick on, right.

[00:55:48] It's just that sometimes we jump in with the latest trend and we deploy a thing. And then, you know, we haven't given it the same kind of security or SOC or other governance audits, things that we would normally do with other software we haven't done. And my favorite example is it's done powerful things. My favorite example is there was a machine learning engineer for one of the AI labs. So somebody that really should know a lot about this running open clause,

[00:56:17] it was working fine on the personal and you can see it and you look at, look up on the internet. And then they were like, Oh, I'll clean up my inbox here. And then they attached to their work one and it had worked so well on all their personal. They were like, okay, we'll search for these things and then find them. And then once you do delete them and for whatever reason, it became search and delete all in the same, not search and then ask me. And then like this person's frantically saying, no, ask me, no, stop what you're doing. And the way that it was through a chat interface,

[00:56:46] it was hooked right into their Slack or teams or whatever. And so the way that interface worked, there was no interrupting. It finished its operations and then it asked you for the next. And so that was a funny story. I thought, because even this machine learning engineer, you know, ran across and unplugged the computer as the resolution to, you know, quit deleting my email. Um, so I suspect a lot of those will happen. And I'm not this like, um, doom and gloom guy that we should never embrace, you know, technology. We,

[00:57:15] we all use technology. Like my, tell my team, it's like, we're not going to be Amish, nothing against them, but like, we're going to have to embrace and we're going to choose to embrace. And we are probably more, um, experimental with agentic AI running things. A hundred percent of my team uses agentic AI and has the ability to create it. So in one way that puts us in a different spot, that's both good and could be higher risk for sure. But that's why we have to talk about those.

[00:57:43] We have a dedicated channel for talking about that and the risks and sharing them and what's going on, not just security, but just related to AI. And they'll, we, we try to share first person, like here's what Damien did, you know, that was dumb. And then, um, occasionally we'll share the, like the link, the one I was telling you about, the link we shared with the team of like a machine learning engineer that worked for one of the big labs, you know, they did this. So just understand, like,

[00:58:12] it's not a measure of intelligence. You know, that person was probably pretty intelligent. Um, and they didn't think they would, that would happen. So let's just keep reminding ourselves of like, there's so many ways. So, um, so we have a lot of guardrails. We've thoughts around it. Cultural is really the big, big piece, but there's a long winded way of saying like, this, this is the beginning. Like, I don't have the answers. I don't have all for sure. I appreciate you elaborating on that.

[00:58:42] And that definitely gives me a lot of assurances. I mean, you know, our acceptable use policy around AI does not allow agentic mainly because we have so much automation here. I've seen people confuse agentic AI and automation. And I want to make sure that, that actually automation for a lot of the things that we do is the better solve versus allowing AI to be a decision maker and an actor. Um,

[00:59:12] that's just that for, I think for many MSPs, I think it's probably smart for us to draw those lines. Um, my biggest concern about agentic AI is that like, and you, you alluded to this earlier in our conversation that, you know, how do you know if you got an email from me, that it's an email from me. If I, if you know, I'm an agentic AI user and you know, I'm using it to optimize my inbox, you might feel like, well, I don't know. Did maybe,

[00:59:42] maybe I need to text John instead of email John that way. I know it's him. And then if I like, Oh, Oh, well let me use agentic AI and texting. Then I'm like, we're in this, like I'm destroying pathways of authenticity. And so I want to make sure that I have the most amount of authenticity and reliability of my name. Cause I do feel like we all have kind of a personal brand and, uh, and I want us all to feel like, you know, we are certain that the person on the other side is,

[01:00:11] is really there because between COVID and AI, we've got these two big forces that have driven a wedge between us as humans connecting, you know, as we used to do in, in person and whatnot. And, um, yeah, technology is an amazing enabler, but there's also a deepening thirst for us to get together. And for us to know that when we get together, we're wholly together. You're not like going, okay, here's John. I just can't wait to get back to my mobile game or something or, or back to, you know,

[01:00:40] Instagram or something. Uh, but rather we're really together and really connecting and making meaningful encounters. So, um, um, as long as agentic AI is doing things that genuinely take rote work, you know, the kind of drudgery that we all have to deal with, I think it's great, but we want to all check ourselves. If agentic AI enables us to simply retreat further and further into isolation, that's bad. Uh, that I think as humans,

[01:01:10] we need to say that's never a good outcome. We all employ, I mean, look, we're in technology. It's the, it's the team of introverts really. Right. So, you know, like there's the exceptions, but I feel like that's the rule. So we, we, we, we, I feel like that's cultural, you know, mindset and culture before the tools. Um, and one of our core rules is you own what you ship. And I don't just mean like that's a software term. Right. But what I mean by that is it's like you're, um, you're using this to generate these,

[01:01:40] uh, proposals or TBRs far faster, but you have that crucial QC gate. And that's the point. Like, I don't, I don't care how it pulled the things together. When you send it, you signed off on it. You like you as an individual, not just as the team. And that, that's our thing. And so we, we want to make sure you know that. And the other part of that is like, we never want to outsource the relationship. Like the preparation, the rote work, like you said, like all the TBR stuff, that kind of thing.

[01:02:10] Wonderful. Um, actually knowing what it said, actually signing off on it, making sure I read what it sent so that we can have a conversation about it. That, you know, I don't ever want to outsource that because I want to have a relationship. And so just because you could, like a lot of things in technology doesn't mean you should. Um, and then, you know, there, I don't think most of us, not just MSPs, but most people can define what an agent means or agent is, right? There's too many,

[01:02:39] there's no settled definition. And, um, and so our view is it's, um, the security has to be structural, structural, not instructional. I like that. I'm stealing that. Right. And, right. Well, I steal from you guys, right. And my point is there, there's policy that needs to enforce and we can't rely on a prompt and because they, whatever they want. So there's various ways to have structure that doesn't, whatever, it can do whatever it wants,

[01:03:09] but we need to have structure and non-instruction. And, um, and, and we've just learned the hard way. Like, and I will just, I wanted to say that because I feel like chat GPT is what we all knew as AI for forever and nothing against it. Right. They have great models. My point is we just got used to being kind of casual and chatty, but when you bring that into the agentic era, when you're giving that ability to create documents or compose emails or affect systems or whatever that is,

[01:03:37] you need to have structural guardrails around you because we're, I think it's lulled us into a little bit of a false sense of security. It's like, well, don't worry, I'll just correct it or I'll just tell it a second thing. And, um, and, and, you know, for the longest time, getting good at prompting was the answer. Now there's the structural guardrails. And my appeal to everybody listening is as an MSP industry, this is where we can shine. When you kind of missed out on SAS, uh,

[01:04:07] the MSP industry, we didn't really get a say when they bought the CRM, they didn't really involve us. And I think that's a shame because there was no governance and there was no security and, you know, that sort of thing. And look, this will be no different in terms of, there was a CRM company that was huge and there were CRM implementation consultants. There's going to be all versions of those with AI consulting shops and vibe coding shops and, you know, marketing that has AI and their platform is going to sell stuff. And I'm not saying you shouldn't, shouldn't use that,

[01:04:36] but I am saying you need somebody that's got a view of your, you know, your sock to type two, uh, you know, we need to think about this differently or you're this, or you're, you're in that industry. Let's figure out how this kind of folds in. Um, and then, you know, you mentioned it depending on the company, you've got to determine like the whole, can we use public models as long as they don't publicly train? Or if you,

[01:05:02] I don't know if you noticed some of the big ones default to training. So you have to go kind of be very careful to not have that on by default. And especially if you onboard new people, I've also noticed another one is if you do the whole, like, this was a good thing, um, like in Claude, it actually sends that back to Claude. You're, you're clicking to enable an escape valve of your, what you thought was a contained, you know, private encrypted conversation.

[01:05:32] So you're right. And those, that's what I'm, I was thinking of too, is like the ground is going to move under our feet quite a bit. And that's something we've got to be. Well, it's faster than ever. And that's the thing we have to acknowledge, right? I think that's, that's my question, right? Open to everybody is how, how do we deal with something that we can't, um, I don't think we can move at the speed of, uh, like I'm not picking on it, but a SOC two type two, right? That was nailed down a while ago.

[01:05:59] And I don't know how often that is updated, but the pace at which AI changes. And like you said, chat GPD used to work and didn't, I've had that happen in number of times for multiple AI platforms. So it's like, you know, now they train by default and they didn't use to, or, you know, who knows which way this is going. Um, and so it's, it's how do we, how do we have governance and security that really operates at the velocity of change? Um,

[01:06:28] and not just tell everybody the answer is no to everything. Come back in 12 months once we've sat down and thought about it. To kind of sum it up, you figured out that the way to get governance is to be open and have normalized conversations about it, create a place where those conversations occur and recognize that we are in a state of unknowns. And when you know that you should behave differently. When you're walking down a street that you've never been down before, you should,

[01:06:57] you should like be attentive to different things. The problem is, is that we can't ever assume that because we've walked down this road before, we know everything around us. This is always a new road. Every time you go into it, stay frosty, keep your antenna up. Yeah. This is definitely, if you have that mindset, you're probably going to be more likely to make the right choices, uh, while you're in that space.

[01:07:27] It is, like you said, it's an adversary. We are not in a friendly environment. It feels friendly. It's not. It does. And depending on the day when it does what you want, it feels great. It feels friendly. And you're like, man, look what I did. And I feel like a hero and I've brought this to my team and now they've, you know, gained benefits. And then you go, well, what's, what are we giving access to? Um, and it's not easy, right? That the thing that I try to do is lead by example. And I don't mean doing it right.

[01:07:57] I try to do that. But what I mean is like, I I'm showing up with some of the failures. I'm showing up with some of the, you know, wow, I got to pay attention to this. We, you know, my team will regularly share like, Oh, I logged in today and it popped up and said, you know, um, give us feedback. And we're like, no, no, if you actually click that link, it's going to send the whole thing off. Like you said. And so like these seemingly innocuous things can trip us up. And so I have showed up and said, I thought I was going to do this and it didn't work.

[01:08:25] And they're usually not security failures as much as like, you know, your project of, you know, if you had stopped at 43 to 47, you could have declared that success or failure, depending on what metrics you were looking for. Right. So the, those are the ones where we try to show up and say this, you know, the more experiments you do, the more you're going to come back with some that don't achieve what you want. And I think that share your experience with others and your experiments. Cause that, cause we're all, the thing is we're all trying to figure it out.

[01:08:55] And actually that's the point of this whole podcast is like sharing the experience, being vulnerable while you're doing it. So that there's a possibility that we all can grow together. And cause we all need to, this is, this is not a time I think for all of us to go, I've got some special sauce now and I'm going to keep it all to myself. That is not that that's a tempting mindset to be in because it feels like you've got a competitive edge.

[01:09:21] There are still so many customers that are not served by technology consultants, technology specialists, managed service providers, you name it. So what we need to do is recognize that we all have, we're all part of this community and all of us benefit when every other one that we call a peer is acting well and is kind of synced up to where we are roughly in maturity level and awareness. That's kind of like my passionate plea is like share,

[01:09:52] be transparent. You don't have to give away like that one little ingredient you think is your special little, little thing, but talk about it openly, talk about your failures and your, your middling successes as well, because that's where some, some sparks of ideas could come that can really get you to the thing that makes you proud that you built something. Just we need to keep sharing. you know, cause we did miss the SAS revolution, um, then the SAS explosion.

[01:10:20] And I think part of that was something you said to me earlier. We were waiting around for a vendor to solve it. Or for whatever reason, we, we, we hunkered down and did our own thing or maybe tried to solve it ourselves. We need to be open about this and share AI. We got to figure it out. We got to find a solution for our customers solution for ourselves. And that, that's dependent on us being really, really vulnerable with each other and saying, I don't have it figured out. I think I've got something here. What do you think? And,

[01:10:50] and then we're, then you're on the right plane where we're all can potentially grow together and, and, and honestly do what we're supposed to do with our customers, deliver value, make them feel like technology is harnessed and, and doing what they need and not, you know, not defeating them and holding them back. Yeah. Yeah. And I think that what you just said is the hard part because I, I heard of the,

[01:11:17] of the MSP that was now having like really junior people, like vibe coding apps and selling them to clients. And I can't imagine that, you know, somebody that knows something about software development, like that's, that's crazy. And I'm not saying they misled them. Maybe they said, here's this tier one guy sending you some really stuff and it works. But that is, that would be tough to maintain or have any security or anything around. And so you have on one hand like that. And on the other hand, I think you have,

[01:11:45] there's a lot of fear because it's like, there's not a vendor that's going to neatly wrap this up, at least not yet. And I think part of the question is like, how do we engage so we can be part of this and give them something, provide some value. We don't need to be bleeding edge where they have open claw, 10 seconds after it's out and no security or vibe coded software, you know, immediately. But we also shouldn't sit on the sidelines because they may pick the person

[01:12:13] that can deliver some of that and still provide some security. Yeah, they may, but I want to make sure that every MSP out there realizes that you still have users that need to be set up devices that need to be provisioned, deprovisioned. there's, and then there's just random stuff that breaks or a vulnerability. I mean, shoot, Microsoft just had that DMARC vulnerability, critical flaw. And guess what? They're not going to patch it. So get on the DMARC train or, or, you know, accept the fact that you're going to get some very,

[01:12:42] very scary phishing attacks that look really authentic. So, so, I mean, on one level, I know we need to, yes, you know, get into the AI game and, and not be, not be timid, but we also need to realize that it is important, but so is all this stuff we're doing day in and day out. Cause I do feel like there's a little bit of that, that phrase that I hear so much and it just, I hate it so much,

[01:13:12] which is like, you're already behind. That is, I would love to see that phrase go away forever. Cause it's so dismissive of the 5,000 decision points and things that you have to do to keep capital flowing, paychecks going, all the, all the things it takes to run a business. And it's like, you know, I, I would like to see someone not constantly make me feel smaller and ineffective and instead say, wow, you're doing a lot of great stuff.

[01:13:40] There's new things coming up and I'm going to try to help you. Maybe move forward a little bit on it, but you know, we're all, I don't know. I think we're awash and a little bit of fear mongering about, you know, how fast AI is going and you're going to be left behind. It's like, it's going to be okay. You know, it's your, you do need to pay attention to it. You do need to work on it, but not frantically, not as if, you know, you got to bet the farm on, on open claw. Like you just said, like slow down,

[01:14:10] you have time to process this. Yeah. Yeah. Look, I have heard of some MSPs. They're like, I'm running my MSP with it. I'm like, that's not going to end well. Like that might be cool for a little while, but, uh, kind of way of saying it. That's bold. Yeah. Well, I'd go even further and say something less nice. Um, but, uh, the, uh, we've tried to address the mindset first and then culturally. And like you said, we're community. And my point around the culture is, you know,

[01:14:39] I think a lot of us got into technology, not just the owners, but a lot of, a lot of our team, because we liked the things change. But now we find ourselves in a spot where things are really not possible to keep up with. You could keep up within an area, but like trying to say I'm keeping up with AI is probably like trying to see, I'm going to keep up with everything related to all facets of healthcare. They'd be on the edge of all of those things. Um, and I just know that. So part of what we're trying to say is that's okay.

[01:15:08] It's okay to not know everything and not try to wear yourself out and believe all the headlines that say that, you know, you no longer need developers or, you know, you know, all kinds of these things and, um, and realize it, like focus on where you're delivering, right. Figure out if there's a particular part where you can, but I love the way you've approached it, John, right? You didn't start off with this VCIO offering. You started off with fixing an internal problem that nobody ever saw.

[01:15:38] They saw kind of that you could find information more quickly, but like no customer ever saw the output of that, right? It was an internal system. You built it. You didn't put it this way. I'm saying this, but if it failed or if it was just down one day, you could do it the way you'd always done it. Right. So you just had to start and then you have the battle scars, you know, right. To figure out, well, why does this thing even do this? It's not a magical box. You start to learn a little bit more about it. So yeah, the, I don't,

[01:16:08] that's the other thing. I just want to give every listening, my team, every listening, like you're not behind. You shouldn't like everything technology. You shouldn't wait a decade before you engage. Right. You may, you know, if you waited a decade on cloud computing, you were, you were going to be behind. And certainly that could be the case here, but you just need to figure out where you're creating value. Double down on that. Right. I love that. One of my takeaways from this, John is we focused on AI,

[01:16:36] but one of them was just making sure that they really needed it before you spent all the time. That's right. A hundred percent. And I, I really just want to say thank you for the audience and the ability to talk about this. Cause every time, every time that I either listen to your podcast, we have a chance to talk. I always have takeaways. I mean, the insights I got on this just by like talking it through kind of also gets

[01:17:04] at what we're all supposed to be doing here, which is engaging with each other, having conversations, listening in on other conversations. This is one of those fly on the wall moments. And it's, it's just so important that we keep doing that. Cause this is how we also realize we're, we're all, we're all working on the same problems together. No one's got it all figured out. Some people may have some things figured out that you can borrow or steal. I'd be on how you want to frame it. But, but yeah, it's just,

[01:17:31] I just appreciate the hours and hours and hours you've put to, to put out great content that really has changed for the better, the whole MSP channel. So thank you, Damien. A little shout out to Zach too, for producing. You're awesome as well. Just thank you for all that you do. No, thank you. That's amazing. I have an amazing producer, Zach. So thank you for what you do, Zach. And to Khan, John,

[01:18:00] thank you for being on this episode of MSP mindset. Yeah. It was a pleasure. And if people could reach out to you, if they have questions, if they would like to engage like in the community aspect, what's, what's the best way to find you? Definitely on LinkedIn. I'm John Snyder on LinkedIn. You can reach out to me via email. Um, uh, J O H N S at netference.net, netference.com and.net actually. Um, but, and I'm at conferences. Um,

[01:18:28] if you hear this and you see me out and about, um, I think the next conference I'll be at is, Roost's flow conference in June. I mean, just pop up. I'm always eager to, to talk and, and engage. Awesome. Thank you so much, John. Thank you. Appreciate it.