WEBVTT NOTE This file was generated by Descript 00:00:00.277 --> 00:00:02.967 Todd Kane: Welcome back to another episode of the Evolved Radio podcast. 00:00:02.967 --> 00:00:06.657 Today we're talking about something every MSP is running into right now. 00:00:06.947 --> 00:00:10.807 How do you let your team actually use AI without turning your ticket data 00:00:10.827 --> 00:00:12.357 into someone else's training set? 00:00:12.807 --> 00:00:15.377 My guest today is Callen Sapien. 00:00:15.377 --> 00:00:19.327 Callen is currently building Synthrio, an AI infrastructure platform built 00:00:19.327 --> 00:00:21.757 specifically for MSPs and SMBs. 00:00:22.127 --> 00:00:25.487 Before that, he was chief strategy officer at MSP-Bots. 00:00:26.247 --> 00:00:28.227 Here's how I'm kind of framing this one, Callen. 00:00:28.447 --> 00:00:32.287 As we talked about, we spent a decade dragging clients to the cloud, then we 00:00:32.287 --> 00:00:37.517 spent a decade dragging them to security, and now our clients are dragging us into 00:00:37.747 --> 00:00:39.347 AI, whether or not we're ready for it. 00:00:39.767 --> 00:00:43.797 And I think we're missing a lot of opportunities, both on the client end 00:00:43.797 --> 00:00:46.017 as well as internally at the MSPs. 00:00:46.417 --> 00:00:50.547 So today, Callen and I are digging into how you actually give your whole 00:00:50.547 --> 00:00:55.267 team AI superpowers without blowing up your PII exposure in the process. 00:00:55.347 --> 00:00:56.367 Callen, welcome to the show 00:00:57.363 --> 00:00:58.773 Callen Sapien CEO Synthreo: Thank you very much, uh, for 00:00:58.783 --> 00:01:00.083 the warm welcome and having me. 00:01:00.103 --> 00:01:05.555 It, you, uh, opened up on everything that I do every day long, so 00:01:05.855 --> 00:01:06.195 Todd Kane: Perfect. 00:01:06.655 --> 00:01:07.115 Great guest. 00:01:08.139 --> 00:01:08.919 Callen Sapien CEO Synthreo: Yeah, there we go. 00:01:08.959 --> 00:01:09.729 We're aligned 00:01:09.815 --> 00:01:09.955 Todd Kane: Yeah. 00:01:10.575 --> 00:01:14.395 So, uh, I'll, I'll maybe reframe this again, like we talked about, um, and 00:01:14.415 --> 00:01:18.035 then I'll, I'll kind of set you up to, to, to, to give us your perspective 00:01:18.095 --> 00:01:22.035 on this 'cause I, I think you're particularly well-suited for this pr- 00:01:22.335 --> 00:01:26.855 this, having your experience with MSP Bots, integrations with PSA, now building 00:01:26.915 --> 00:01:31.475 a very, very AI-forward platform, so good exposure on kind of both ends of these, 00:01:31.655 --> 00:01:33.755 these, uh, the, these, uh, this topic. 00:01:34.275 --> 00:01:40.395 And the part that concerns me is I see sort of this multi-layered approach 00:01:40.455 --> 00:01:43.835 for people, especially service managers that I interact with a lot in the 00:01:43.875 --> 00:01:49.195 work that I do, and some of them are really leveraging AI in great ways. 00:01:49.475 --> 00:01:53.395 They're, uh… In particular, I've seen, uh, a lot of people dumping ticket 00:01:53.455 --> 00:01:58.745 data into a model and having it do some analysis and, uh, spotting some trends and 00:01:58.755 --> 00:02:02.335 giving it some insights, and that stuff is golden because doing a service review 00:02:02.355 --> 00:02:07.085 manually is time-consuming and requires a lot of sort of mental bandwidth and 00:02:07.115 --> 00:02:08.615 energy to kind of connect those dots. 00:02:09.135 --> 00:02:12.515 So being able to offload that to a model is super, super valuable. 00:02:12.655 --> 00:02:18.875 But rightfully, people are a bit nervous about where they give that data to, right? 00:02:18.955 --> 00:02:22.185 If you're gonna export all of your tickets, just dump it into a model and 00:02:22.235 --> 00:02:26.755 say, "Give me some insights," you gotta be pretty particular about how you're 00:02:26.855 --> 00:02:30.635 actually servicing that model, what information potentially goes in there. 00:02:30.715 --> 00:02:33.955 Simple things like the fact that, you know, I think you and I, you and I talked 00:02:34.015 --> 00:02:36.835 about this, is like a lot of people don't even recognize, like even if you 00:02:36.915 --> 00:02:41.895 have a paid model, there is a setting in your options to turn off whether 00:02:41.935 --> 00:02:43.615 or not it goes into training data. 00:02:43.635 --> 00:02:46.345 So just because you're paying doesn't necessarily mean that 00:02:46.355 --> 00:02:49.475 you're clean to give it kind of as much information as you want. 00:02:49.515 --> 00:02:53.075 So I recognize why people are cautious about this, but I see this as sort of 00:02:53.115 --> 00:02:57.755 a tale of two MSPs, the people that are leveraging AI in order to be able to 00:02:57.775 --> 00:03:02.135 do these things and hopefully doing it well, versus the people that are not 00:03:02.195 --> 00:03:06.945 able to take advantage of this because they're somewhat justifiably fearful about 00:03:06.975 --> 00:03:08.935 the best way to take, to go about it. 00:03:09.555 --> 00:03:13.225 So what's your, what's your take on how we go, we go about this in 00:03:13.275 --> 00:03:17.355 a safe fashion in order to maximize AI's potential inside the MSP? 00:03:18.611 --> 00:03:22.521 Callen Sapien CEO Synthreo: I, I, I-- That's a great, uh, overview and, and I 00:03:22.531 --> 00:03:25.171 break it down into really three areas. 00:03:25.551 --> 00:03:27.671 There's the technological, right? 00:03:27.681 --> 00:03:29.321 Did I uncheck that box? 00:03:29.691 --> 00:03:33.821 Uh, I have ZDR enabled, zero data retention, right? 00:03:33.821 --> 00:03:36.451 Uh, uh, we love to throw out acronyms in our space and 00:03:36.555 --> 00:03:37.365 Todd Kane: It's a new one for me. 00:03:37.365 --> 00:03:37.715 I like it 00:03:38.191 --> 00:03:40.391 Callen Sapien CEO Synthreo: and, and zero data retention is actually 00:03:40.391 --> 00:03:41.681 probably the most critical piece. 00:03:41.681 --> 00:03:45.971 I'll circle back to it because I don't wanna, like, do the ADHD thing where 00:03:45.971 --> 00:03:49.281 I, I, I go off and I say I'm gonna talk about three things and then, then 00:03:49.281 --> 00:03:50.811 jump down, which I almost did just now. 00:03:51.041 --> 00:03:53.391 So you've got your technical, which includes things like 00:03:53.391 --> 00:03:54.911 zero data retention, ZDR. 00:03:55.101 --> 00:03:59.911 That includes the, the, the actual checking of the settings, uh, and the-- 00:03:59.951 --> 00:04:03.091 and then the secure, and we've been talking about this for a long time, 00:04:03.091 --> 00:04:07.781 whether it's cloud security or even going back to the aughts where it's the right 00:04:07.811 --> 00:04:09.921 permission levels, uh, set up, right? 00:04:09.921 --> 00:04:11.501 The right identity and access policy. 00:04:11.821 --> 00:04:13.861 Uh, so there's the technological side. 00:04:14.211 --> 00:04:19.321 Then something that we, we overlook a lot is the contractual side. 00:04:19.441 --> 00:04:26.381 We, we have gotten used to these, like, giant SaaS agreements that we 00:04:26.381 --> 00:04:30.621 just, we just click, "Yep, I read it. Yep, I read it." And most courts have 00:04:30.621 --> 00:04:34.921 said, you know, it's not really a, a-- truly something that we would expect. 00:04:34.921 --> 00:04:39.031 It does limit liability, but, but you can't enforce every clause, right? 00:04:39.061 --> 00:04:43.951 Uh, but the AI, uh, terms are actually not as long as most SaaS terms, 00:04:43.981 --> 00:04:45.561 and they have some hooks in there. 00:04:45.851 --> 00:04:48.671 Uh, and then the last piece is cultural, right? 00:04:48.701 --> 00:04:53.671 Um, y- y-- we, we are-- You, you mentioned the, the push, push, now pull. 00:04:54.241 --> 00:04:59.261 We cannot suddenly, three years later, come in and say, "I've been handling 00:04:59.261 --> 00:05:00.671 your security for a really long time. 00:05:00.671 --> 00:05:01.621 I ignored AI. 00:05:01.621 --> 00:05:02.581 I just told you to ignore AI. 00:05:02.701 --> 00:05:03.991 I know you didn't ignore AI. 00:05:04.841 --> 00:05:08.301 Now I'm gonna shut everything down, and you can't use it," because 00:05:08.301 --> 00:05:09.381 people have been getting value. 00:05:09.381 --> 00:05:13.291 Even if that value has been a picture of my fridge and see what I'm gonna make 00:05:13.351 --> 00:05:17.201 today or what's in my fridge, they're getting value, and they haven't had the 00:05:17.201 --> 00:05:21.031 pain as much as they have in other areas that they haven't had the governance. 00:05:21.321 --> 00:05:25.031 And so one other thing that's kind of interesting on that last part is 00:05:25.939 --> 00:05:31.529 I was on with an MSP in Canada and, uh, the, the Canadian AI Privacy 00:05:31.549 --> 00:05:33.669 Security, uh, Act failed, right? 00:05:33.669 --> 00:05:34.519 It did not go through. 00:05:34.879 --> 00:05:39.239 But that doesn't mean the FIPS Act that, that exists around data, 00:05:39.499 --> 00:05:43.259 data retention and all of those things can't be enforced, right? 00:05:43.559 --> 00:05:49.499 GDPR counts, FIPS counts whether or not it was done by AI, SaaS or a human. 00:05:49.869 --> 00:05:53.339 And we, we-- when we're, when we-- we do need to bring that in as a 00:05:53.339 --> 00:05:56.359 governance piece, but, but those are the three kind of buckets. 00:05:56.369 --> 00:05:59.339 Going back to the zero data retention, because I do think that's the most 00:05:59.339 --> 00:06:03.939 important one, second is the contractual, but the, the, the zero data retention 00:06:03.939 --> 00:06:08.709 is critical because you can uncheck that box that says or check that box 00:06:08.709 --> 00:06:13.499 that says, "Do not train on my data." There is no box with any vendor right 00:06:13.499 --> 00:06:19.229 now that says, "Do not retain my data." And they propose in their contractual 00:06:19.229 --> 00:06:23.489 side that they have legitimate reasons for it, if you're r- abusing it, if they 00:06:23.489 --> 00:06:27.099 need to see if a model's misbehaving, if they have these other things. 00:06:27.399 --> 00:06:33.109 But every single model provider asks for or demands as part of the 00:06:33.109 --> 00:06:37.129 use of the product, the ability to keep data for as little as seven 00:06:37.129 --> 00:06:41.349 years, which is a long time, uh, and as much as, uh, indefinitely. 00:06:41.719 --> 00:06:47.739 When we look at like Google having shifted from don't be evil for the first 11 years 00:06:47.969 --> 00:06:50.709 and, and listening to, "I'm not gonna train on your data, I'm not gonna do 00:06:50.709 --> 00:06:54.219 these things," and still work within the, the confines, we know that if somebody 00:06:54.219 --> 00:06:58.249 has access and they keep access, it may not always be used for good things. 00:06:58.249 --> 00:07:01.739 I don't wanna be like a scary, Anthropic is using your data to destroy the 00:07:01.739 --> 00:07:06.899 world, but, but if they retain a copy of it, you know, things can change. 00:07:06.899 --> 00:07:09.019 But I, I-- Before I jump into the contractual, I'd love any 00:07:09.019 --> 00:07:10.349 thought that you have there. 00:07:10.663 --> 00:07:14.413 Todd Kane: Yeah, I guess like the biggest issue with that is gonna be breach, right? 00:07:14.453 --> 00:07:18.753 Like your data is retained somewhere and it's not necessarily-- 'cause originally 00:07:18.753 --> 00:07:22.983 like when the, the, when GPT was first sort of exploding and there was like 00:07:22.983 --> 00:07:26.473 these stories of like, uh, these guys at Samsung that loaded up a bunch of 00:07:26.473 --> 00:07:32.343 information in-into GPT and, you know, uh, nearly eliminated one of their 00:07:32.363 --> 00:07:37.203 patents because they'd, uh, in ess-in essence made the information public 00:07:37.503 --> 00:07:39.153 into one of the, one of the models. 00:07:39.163 --> 00:07:44.083 Or like your data kind of resurfacing as information for other people if 00:07:44.083 --> 00:07:45.213 it's a part of the training data. 00:07:45.443 --> 00:07:49.543 And I think a l- some of that is definitely, uh, not as much of a 00:07:49.543 --> 00:07:53.163 concern as it originally was, and it goes to sort of the f- like the first, 00:07:53.433 --> 00:07:59.353 uh, I would say edict of AI safety is don't use an unpaid model, right? 00:07:59.353 --> 00:08:01.863 Like the free model is there for training data, right? 00:08:01.903 --> 00:08:05.053 So if you're gonna give your data to anything, make sure it's a paid 00:08:05.053 --> 00:08:07.443 model at the very least and you're turning off some of this data. 00:08:07.713 --> 00:08:10.253 But to your point, like it's still there, right? 00:08:10.253 --> 00:08:15.273 Like there's some websites even for myself where I will tell it not to save my credit 00:08:15.273 --> 00:08:19.483 card data 'cause I'm like, "Yeah, it's not really that big of a company if it's not 00:08:19.493 --> 00:08:22.243 using Shopify or Amazon as a backend cart. 00:08:22.263 --> 00:08:24.503 Like do I really trust it?" Same idea. 00:08:24.793 --> 00:08:28.023 Like you're, you're essent-uh, essentially kind of giving this, uh, 00:08:28.053 --> 00:08:30.183 this information over for free forever. 00:08:30.273 --> 00:08:32.723 Uh, so yes, there's a risk of data breach. 00:08:33.103 --> 00:08:36.553 Some people think like, um, "I'm a small player, you know, my 00:08:36.553 --> 00:08:40.063 information, you know, would it… w-why would they possibly use my 00:08:40.063 --> 00:08:41.543 information for something," right? 00:08:41.573 --> 00:08:42.493 And it's not really that. 00:08:42.493 --> 00:08:49.743 It's like, okay, data exfil, they dumped a petabyte of information from, uh, AI model 00:08:49.773 --> 00:08:52.003 X. Your information's in there, right? 00:08:52.003 --> 00:08:56.303 So I think that's sort of the, the biggest risk is, is just data exfil from some 00:08:56.483 --> 00:09:00.583 scary event in the future and, you know, could be even sort of strangely benign. 00:09:00.593 --> 00:09:05.423 Like we had these conversations recently with Alex Dao on, on security where, uh, 00:09:05.473 --> 00:09:09.313 training models broke into other companies in order to gain access into stuff, and 00:09:09.313 --> 00:09:13.373 that's exactly how this could go down is like, you know, uh, Fable Six, uh, 00:09:13.383 --> 00:09:16.813 launches an attack because it needs more training data and starts breaking into 00:09:16.813 --> 00:09:19.983 all the other AI companies and s- and pilfers all of its information, right? 00:09:20.313 --> 00:09:22.263 Like, uh, you never know how this stuff's gonna come about. 00:09:22.263 --> 00:09:25.193 So I think that, that's a good point is just understand sort of 00:09:25.193 --> 00:09:28.363 where the data is resident and how it's being utilized right now. 00:09:29.739 --> 00:09:31.659 Callen Sapien CEO Synthreo: That's, that is a, uh… And that, and that 00:09:31.659 --> 00:09:37.319 even in-- So there's a large problem of people being incredibly empowered. 00:09:37.319 --> 00:09:41.049 Uh, uh, we, we, we have the ability to kinda host your own MCPs, right? 00:09:41.049 --> 00:09:45.109 And, and that may seem somewhat superfluous because there's a lot of 00:09:45.109 --> 00:09:49.769 MCPs out there, but the reason we have it is because of the data residency and 00:09:49.769 --> 00:09:50.779 the other pieces you're talking about. 00:09:50.779 --> 00:09:56.339 I've got a, a partner that needs data residency, uh, from regulatory reasons. 00:09:56.619 --> 00:10:02.299 Uh, they can use the official app in the store that exposes financial information 00:10:02.299 --> 00:10:06.929 and these other things, the official MCP for a marketplace, uh, through Claude. 00:10:07.789 --> 00:10:12.779 But Claude won't guarantee that that data doesn't on US servers or it's 00:10:12.829 --> 00:10:14.259 saved there and these other things. 00:10:14.569 --> 00:10:17.759 People need to understand what that flow of data is, and it's actually a tremendous 00:10:17.759 --> 00:10:20.529 amount of value that an MSP can provide. 00:10:20.629 --> 00:10:23.699 Uh, it's a huge value add, giving this visibility and 00:10:23.699 --> 00:10:24.949 building the services around it. 00:10:24.989 --> 00:10:30.019 And at the risk of sounding anti-security again, and I'm not anti-security, I've 00:10:30.349 --> 00:10:34.549 been a CISO, but at the risk of it, uh, the thing that I would like to actually 00:10:34.549 --> 00:10:40.769 point out related to this is you can't sell and protect AI the same way that we 00:10:40.769 --> 00:10:45.809 did endpoints and other pieces because it doesn't operate or act the same way. 00:10:45.999 --> 00:10:51.699 And in the last five years, six years, done a lot in security, but 00:10:51.789 --> 00:10:55.679 a lot of MSPs that thought they were gonna be MSSPs did not become MSSPs. 00:10:55.729 --> 00:11:00.569 They bought a SOC, they bought an MDR, they bought an outsourced 00:11:00.999 --> 00:11:03.769 helper to deliver these things. 00:11:03.769 --> 00:11:06.469 So they've outsourced the services side of this instead of standing it 00:11:06.479 --> 00:11:07.879 up because it costs a lot to stand up. 00:11:09.259 --> 00:11:13.669 the, the challenge with trying to sell AI that way is the 00:11:13.699 --> 00:11:16.059 management is where the margin is. 00:11:16.289 --> 00:11:20.439 The, the, the, the driving of the output, the guidance that, that's 00:11:20.439 --> 00:11:24.779 required to actually impact the P&L is where the management is and the 00:11:24.779 --> 00:11:27.909 guidance of, and that, that actually shifts into the contractual side. 00:11:28.359 --> 00:11:33.029 In the US, a lot of the court cases are up in the air still on whether 00:11:33.029 --> 00:11:38.139 something like attorney-client privilege, if someone's using AI, counts, right? 00:11:38.139 --> 00:11:38.889 Because for the same 00:11:39.019 --> 00:11:39.359 Todd Kane: Yeah. 00:11:39.649 --> 00:11:40.539 Callen Sapien CEO Synthreo: destroying your patent. 00:11:40.629 --> 00:11:40.969 Todd Kane: Mhm 00:11:41.129 --> 00:11:44.109 Callen Sapien CEO Synthreo: And the one thing through the cases that are 00:11:44.109 --> 00:11:48.409 pro and for and against and this and that, the one commonality through all 00:11:48.409 --> 00:11:52.779 of them when we look at what's becoming, uh, stare decisis, the, the let the, 00:11:52.819 --> 00:11:58.839 let the decision stand, is who had the governance and the stewardship 00:11:58.849 --> 00:12:03.189 of the data contractually, and was there anyone that could take this data 00:12:03.219 --> 00:12:08.109 for a legitimate or an illegitimate reason and shift it, uh, and use it? 00:12:08.149 --> 00:12:10.759 And, and that's where the contractual side comes in. 00:12:11.019 --> 00:12:15.379 Pretty much every main, main provider that's out there that we're going to use 00:12:15.379 --> 00:12:17.429 this data to, to, to, to learn something. 00:12:17.439 --> 00:12:20.959 Not train the model, but we're gonna learn something, and we can access 00:12:21.019 --> 00:12:22.569 it over the course of using it. 00:12:23.689 --> 00:12:26.479 And so that, that to me is a, a big area that we can get into. 00:12:29.053 --> 00:12:29.083 Todd Kane: Okay. 00:12:29.543 --> 00:12:31.493 Um, yeah, I think the, the governance piece, like that's, that's I, 00:12:31.493 --> 00:12:32.363 I think really interesting. 00:12:32.363 --> 00:12:35.343 I'll, I'll, uh, governance, I'm gonna write this down just so, 'cause we 00:12:35.343 --> 00:12:38.703 got-- we're gonna open up tons of ADHD threads here as we go, I'm sure. 00:12:39.993 --> 00:12:40.263 All right. 00:12:40.973 --> 00:12:46.123 Um, the, uh, I guess like the function around sort of the util-utilization for 00:12:46.193 --> 00:12:49.683 PSA data, like that, that to me, like I said, is like I, I wanna focus on this 00:12:49.683 --> 00:12:53.133 because I feel like there's an incredible amount of value for it, but it's such 00:12:53.133 --> 00:12:56.063 a fraught issue for doing it correctly. 00:12:56.443 --> 00:12:59.673 Um, so obviously, okay, you know, we're not using, um, 00:12:59.693 --> 00:13:01.323 a train, uh, uh, free model. 00:13:01.603 --> 00:13:04.663 Maybe we're, uh, handing it over to a paid model. 00:13:04.813 --> 00:13:08.943 Uh, I think understanding, uh, where the contractual I think lays into this is 00:13:08.943 --> 00:13:13.023 like being clear with the client that maybe you're using their data, right? 00:13:13.043 --> 00:13:17.143 Like if you're supporting a law firm and they have client information, 00:13:17.193 --> 00:13:20.253 uh, potentially in some of those tickets or the connections that 00:13:20.253 --> 00:13:24.293 you create to them, that creates a bit of a, a slippery slope as well. 00:13:24.593 --> 00:13:29.363 And maybe that's an interesting one as well, is how do we think about the PSAs 00:13:29.393 --> 00:13:34.983 that have AI capabilities built into them or other tools that are connected, right? 00:13:34.983 --> 00:13:39.263 Like that is, you know, you're not even necessarily giving something to a 00:13:39.263 --> 00:13:43.603 model, but the model is passing through, or sorry, the data is passing through 00:13:43.633 --> 00:13:47.993 another model in some fashion you don't necessarily have direct visibility to. 00:13:48.283 --> 00:13:51.803 So like, and I have to imagine this is incredibly common with most of 00:13:51.803 --> 00:13:55.503 the platforms creating some type of AI capability, whether or not 00:13:55.503 --> 00:13:58.603 it's a third-party module or just built into the system, right? 00:13:58.603 --> 00:14:00.103 So how do you think about that piece? 00:14:01.291 --> 00:14:02.691 Callen Sapien CEO Synthreo: Yeah, I think it's, it's, it's, it's 00:14:02.691 --> 00:14:07.141 a completely under-addressed problem that exists right now. 00:14:07.191 --> 00:14:11.971 Uh, it, it-- The, the, the, at least the use of AI within traditional SaaS 00:14:11.981 --> 00:14:16.811 products in a completely unmanaged way. 00:14:16.851 --> 00:14:19.521 Monday.com, for example, added a feature that basically 00:14:19.521 --> 00:14:21.371 put Lovable or Replit in it. 00:14:21.401 --> 00:14:26.021 You can build apps, build MCPs, connect things into your Monday as if you 00:14:26.021 --> 00:14:27.971 were building, building front ends. 00:14:28.241 --> 00:14:31.981 And that's not… That's-- It's very, very challenging to be 00:14:31.981 --> 00:14:35.061 able to get visibility into what people are doing in, in that area. 00:14:35.101 --> 00:14:39.641 And so I think that, uh, it, it's still kinda coming out in the wash. 00:14:39.651 --> 00:14:41.451 We're working on some ways of building visibility. 00:14:41.451 --> 00:14:45.261 There's some really cool apps out there doing that, but it's, it's, it's, it's, 00:14:45.271 --> 00:14:49.911 it's very heavyweight right now, and it, and it's across every single app. 00:14:49.921 --> 00:14:55.311 Uh, when you look at a standard MSP, they use 17 tools to deliver the, the output. 00:14:55.351 --> 00:14:58.041 Their customers use about 23 to 28 tools. 00:14:58.501 --> 00:15:01.801 All of them have AI in them now processing it, and most of them 00:15:01.801 --> 00:15:04.801 have updated their terms to make it an opt-out, not an opt-in. 00:15:04.841 --> 00:15:07.671 You're, you're, you're not… It's not like calling in used to be. 00:15:07.671 --> 00:15:11.911 I think it eventually will be, but, but it, but the data is being used, maybe 00:15:11.911 --> 00:15:15.121 not to improve things, but it's being pushed through a model that, that, that 00:15:15.121 --> 00:15:16.371 we don't know what it's gonna do with it. 00:15:16.641 --> 00:15:19.291 When we get back to kinda your core question, which is, you know, how 00:15:19.291 --> 00:15:22.651 are these things being used, and how are we able to do it in a, in a core 00:15:22.651 --> 00:15:24.851 way or in a, uh, in a strong way? 00:15:25.381 --> 00:15:29.161 One thing that, that I think we talk about a lot, and it's shifted with 00:15:29.481 --> 00:15:33.491 AI, people have asked me why I think that AI is so good at coding, right? 00:15:33.491 --> 00:15:37.231 This was a hard task that required a lot of brainpower and 00:15:37.231 --> 00:15:38.671 a lot of reasoning that exists. 00:15:38.841 --> 00:15:42.171 And the whole reason that it's great at coding is because open source software 00:15:42.171 --> 00:15:48.441 exists and, uh, Bitbucket and, uh, and, uh, and Stackbucket and Stackjack and all 00:15:48.441 --> 00:15:52.381 these, these things where I could go look at good code examples, bad code examples, 00:15:52.381 --> 00:15:53.811 and they train the models off of that. 00:15:54.151 --> 00:15:58.641 I think right now we're trying to protect the wrong things with our data, even the 00:15:58.641 --> 00:16:00.201 PIIs and the other, the other pieces. 00:16:00.201 --> 00:16:01.721 We have to do that because of compliancy. 00:16:02.061 --> 00:16:08.033 But the data Your MSP's data is not very different than this MSP's data. 00:16:08.043 --> 00:16:10.543 That lawyer's data is not very different from that lawyer's data. 00:16:10.883 --> 00:16:13.493 But what is different is your processes and the how. 00:16:13.873 --> 00:16:18.833 You're seeing more and more, uh, these attacks almost, uh, they-- I, I 00:16:18.833 --> 00:16:22.123 would actually say they're almost to a level of attack where, uh, Claude 00:16:22.443 --> 00:16:23.753 Design's a perfect example of this. 00:16:24.113 --> 00:16:27.133 Claude went and had a Figma integration. 00:16:27.483 --> 00:16:32.523 Claude learned how Figma works, and then they launched Claude Design. 00:16:32.933 --> 00:16:39.183 If someone doesn't put any PIIs in there, doesn't put any data, doesn't put anything 00:16:39.183 --> 00:16:42.633 in there around what the ticket actually contains, but they, they actually describe 00:16:42.633 --> 00:16:50.663 their process to the AI, now I've given the AI what makes me a 28% EBITDA MSP 00:16:50.843 --> 00:16:55.023 instead of a flat or an, a 5% EBITDA MSP. 00:16:55.183 --> 00:16:59.973 If I put that data on how the, the, the process of how this lawyer treats 00:16:59.973 --> 00:17:03.363 clients and, and runs through their process, now I've given away my secret 00:17:03.363 --> 00:17:07.593 sauce, and anybody who uses that model that it's improved itself on gets that. 00:17:07.593 --> 00:17:10.173 And I think that's one of the things that we really do have to, we have 00:17:10.173 --> 00:17:14.853 to talk through, and, uh, how do we protect the knowledge and the how? 00:17:15.033 --> 00:17:19.263 Because IP is, is, is shrinking as a, as a capability. 00:17:19.483 --> 00:17:22.673 Um, and that's something that I, that I think that is very interesting, 00:17:22.683 --> 00:17:23.793 and how do we protect that? 00:17:23.793 --> 00:17:25.883 And that actually goes back to that ZDR, right? 00:17:25.883 --> 00:17:28.473 If, if the model doesn't remember anything about what happened, 00:17:28.473 --> 00:17:29.553 then, then it can't learn. 00:17:30.219 --> 00:17:30.479 Todd Kane: Yep. 00:17:30.789 --> 00:17:31.059 Yeah. 00:17:31.459 --> 00:17:34.319 I, I almost feel like that one's unavoidable, and I almost feel 00:17:34.319 --> 00:17:38.959 like, uh, also, uh, like one of my favorite expressions that I've said 00:17:38.959 --> 00:17:42.819 forever since I've been consulting is, "Knowledge is easy, execution is 00:17:42.819 --> 00:17:46.929 hard." Uh, I don't know that there's a lot of proprietary information in the 00:17:46.929 --> 00:17:49.009 future that we can actually protect. 00:17:49.079 --> 00:17:52.289 Uh, uh, outside of things like the Coca-Cola recipe, right? 00:17:52.289 --> 00:17:56.469 Like sure, like there's some s- specifically proprietary information. 00:17:56.469 --> 00:18:00.599 But I think it's, it's a good example of what is really strange about the MSP 00:18:00.609 --> 00:18:04.899 industry that I call it a lot is we all fundamentally have the same business 00:18:04.899 --> 00:18:09.389 model and noob- no two businesses run even remotely similarly, right? 00:18:09.399 --> 00:18:13.459 Like, a- and this is despite the fact that like there's a huge incentive and 00:18:13.459 --> 00:18:17.919 a huge model for commoditization and forcing people to, to work the same way. 00:18:17.919 --> 00:18:23.069 But quite frankly, most organizations are a reflection of their owner, right? 00:18:23.109 --> 00:18:27.349 And anyone who has never noticed this, just like do a bit of a tour and think 00:18:27.349 --> 00:18:30.019 about the places that you've worked, the companies that you've looked at. 00:18:30.269 --> 00:18:33.629 They are always a reflection of the owner in this weird, weird way. 00:18:33.979 --> 00:18:36.729 So I think that's really only the, the only differentiation. 00:18:36.949 --> 00:18:40.939 So I don't know that like the information of, uh, how I do something 00:18:41.089 --> 00:18:45.399 is, is necessarily as sort of a secret sauce as maybe, maybe it 00:18:45.419 --> 00:18:46.729 would be, especially in the future. 00:18:46.759 --> 00:18:50.269 Because knowledge will just be so systemic and available, right? 00:18:50.509 --> 00:18:50.799 Yeah. 00:18:51.159 --> 00:18:51.439 So, 00:18:52.239 --> 00:18:52.457 um- 00:18:52.757 --> 00:18:55.017 Callen Sapien CEO Synthreo: I really like, I'm not a huge Sam Altman 00:18:55.017 --> 00:18:58.037 fan, but I do think he, he's been right on a lot of things, already 00:18:58.037 --> 00:18:59.797 built a trillion-dollar company. 00:19:00.077 --> 00:19:03.027 Uh, but, uh, and a trillion-dollar non-for-profit. 00:19:03.097 --> 00:19:04.327 Uh, but, uh, 00:19:04.627 --> 00:19:06.187 Todd Kane: If you're getting technical about it, yeah. 00:19:06.617 --> 00:19:09.207 Callen Sapien CEO Synthreo: you're getting technical, uh, he, he said 00:19:09.207 --> 00:19:13.977 that AI, and it's, it's turning into hands as well, but he said AI is an, 00:19:14.017 --> 00:19:16.587 is a utility, but it's intelligence. 00:19:16.607 --> 00:19:21.007 It, you, instead of electricity, instead of, instead of gas or, or 00:19:21.007 --> 00:19:22.187 these other things, it's intelligence. 00:19:22.187 --> 00:19:25.397 And you turn on the spigot, and companies that can spend a lot like 00:19:25.397 --> 00:19:28.117 they would on electricity can spend a lot on intelligence, and companies that 00:19:28.117 --> 00:19:29.677 can't will get a little intelligence. 00:19:29.677 --> 00:19:33.967 And I think it, it actually bodes to your point of the fact that it, it's 00:19:33.977 --> 00:19:36.987 going to be ubiquitous, and there's, there's a, a, a time that we can 00:19:36.987 --> 00:19:38.877 protect it, but it is gonna go away. 00:19:38.877 --> 00:19:41.127 It's just something to be aware of as we go through. 00:19:41.127 --> 00:19:45.057 When we actually talk about the practical application of using AI 00:19:45.057 --> 00:19:48.037 in ticket data, I think that there's still a lot of focus on personal 00:19:48.037 --> 00:19:52.767 productivity, and we haven't really s- got into the infrastructure phase. 00:19:53.017 --> 00:19:57.377 That's where we're seeing the, the labor, the FTE, and the P&L impacts is 00:19:57.377 --> 00:20:02.057 when we start to get into the actual infrastructure phase of installing 00:20:02.067 --> 00:20:06.877 an agent, of installing, uh, the data layer, the ontology layer, if you will, 00:20:07.077 --> 00:20:12.547 uh, within an o- an organization so that we can delegate and dispatch to, 00:20:12.637 --> 00:20:14.947 to the, to the agent to get it done. 00:20:14.957 --> 00:20:19.217 I actually think that human in the loop is becoming a more and more antiquated 00:20:19.217 --> 00:20:21.487 term, and it's almost a Luddite type term. 00:20:21.747 --> 00:20:25.537 It, it, it… Human in the loop, humans make mistakes as much or more. 00:20:25.567 --> 00:20:29.197 Uh, there's, there's a cool company in our space, uh, that got me turned 00:20:29.197 --> 00:20:33.707 on to, uh, AI, uh, before it was cool. 00:20:33.737 --> 00:20:35.647 I was still on the machine learning and data as well. 00:20:35.647 --> 00:20:39.297 I was at, at MSP-Bots, and we were trying to figure out AI and machine learning. 00:20:39.497 --> 00:20:43.297 But, uh, Mark Elaev, the fou- one of the founders with Matt over at, at, at 00:20:43.297 --> 00:20:47.847 Thread, uh, called me in March of 2022, and he's like, "We've got this thing that 00:20:47.847 --> 00:20:53.107 we got into the beta of called OpenAI, and we're, we're using it to, to, to guess 00:20:53.157 --> 00:20:58.297 and accurately put in time entry against a ticket." And I was like, "Oh, cool. 00:20:58.297 --> 00:21:02.227 How accurate are you?" Like, "84 or 85% accuracy." I was like, "That's really 00:21:02.227 --> 00:21:09.125 high." And people didn't adopt it because it was 84 or 85% accuracy attacks, 00:21:09.525 --> 00:21:11.325 uh, remember these tickets had zero, 00:21:11.563 --> 00:21:12.393 Todd Kane: Yeah, exactly. 00:21:12.635 --> 00:21:13.135 Callen Sapien CEO Synthreo: entry 00:21:13.193 --> 00:21:14.693 Todd Kane: an 84% improvement, yeah. 00:21:15.465 --> 00:21:17.055 Callen Sapien CEO Synthreo: It's, yeah, and, and the ones 00:21:17.055 --> 00:21:19.665 that did are 37% accurate. 00:21:20.147 --> 00:21:20.517 Todd Kane: Wow 00:21:20.525 --> 00:21:23.815 Callen Sapien CEO Synthreo: you know, the, the, the level of which I, I get 00:21:23.815 --> 00:21:27.645 into debate a lot and people argue with me on, and I understand accountability 00:21:27.645 --> 00:21:32.995 is the hard part, but, you know, with like self-driving cars, they are 100% 00:21:33.035 --> 00:21:36.275 more safe than human-driven cars. 00:21:36.475 --> 00:21:38.885 But that isn't good enough for a lot of people because 00:21:39.125 --> 00:21:40.305 can't hold the car accountable. 00:21:40.305 --> 00:21:42.935 But you can hold the company and the machine and all these other, 00:21:42.935 --> 00:21:44.395 just like if a seatbelt fails. 00:21:44.745 --> 00:21:47.755 Um, I think the, the, the answer should be it should be better than 00:21:47.755 --> 00:21:50.985 the human in a measurable way, and that's what I think when we're doing 00:21:50.985 --> 00:21:54.399 these things we should be looking at within the, within the instances 00:21:54.699 --> 00:21:54.939 Todd Kane: Yeah. 00:21:55.497 --> 00:21:57.557 Tired of fighting the MSP fires alone? 00:21:57.737 --> 00:22:00.657 The Ops Leader Pro group connects service delivery professionals who 00:22:00.697 --> 00:22:02.297 understand your daily challenges. 00:22:02.557 --> 00:22:06.217 From KPIs and workflows to career planning and team management, Ops 00:22:06.217 --> 00:22:08.087 Leader Pro has systems for you to use. 00:22:08.517 --> 00:22:12.717 Join operations leaders from successful MSPs who are sharing real solutions for 00:22:12.717 --> 00:22:16.897 managing client expectations, optimizing service delivery, and making your service 00:22:16.897 --> 00:22:19.237 delivery team as effective as possible. 00:22:19.567 --> 00:22:23.217 Ops Leader Pro, 'cause your service desk deserves more than just survival mode. 00:22:23.587 --> 00:22:25.847 Visit opsleader.co. 00:22:25.887 --> 00:22:31.067 That's O-P-S leader.co to apply to join the public community. 00:22:31.133 --> 00:22:31.393 Todd Kane: All right. 00:22:31.413 --> 00:22:37.003 So you mentioned, um, uh, sort of like the where the information is resident, 00:22:37.063 --> 00:22:40.423 and this I-- this is another layer that I think is really fascinating about 00:22:40.423 --> 00:22:44.603 this is, okay, so yes, you could use a cloud model for some of these things. 00:22:44.943 --> 00:22:50.693 Um, you could use a hosted like Azure container model for, for these things. 00:22:50.911 --> 00:22:51.151 Callen Sapien CEO Synthreo: Mhm. 00:22:51.451 --> 00:22:52.933 Todd Kane: your guys' approach is different. 00:22:52.933 --> 00:22:55.883 It's sort of like, uh, maybe a hybrid between the two, if 00:22:55.883 --> 00:22:57.003 I understand that correctly. 00:22:57.083 --> 00:23:00.713 Yeah, can sort of private containers, uh, uh, type approach. 00:23:01.093 --> 00:23:04.403 Um, the other one that I was thinking is like, well, maybe are we just 00:23:04.403 --> 00:23:08.843 gonna start doing a lot more, uh, ho- self-hosted and private models, right? 00:23:08.873 --> 00:23:11.883 And, uh, like that kind of makes some sense because I feel like, like 00:23:11.893 --> 00:23:14.743 despite the fact that the frontier keeps on pushing forward, like the 00:23:14.743 --> 00:23:19.463 frontier models from Anthropic, OpenAI, uh, you know, Kimi, they're, they're 00:23:19.463 --> 00:23:23.723 advancing and their capabilities are, are, uh, advancing incredibly. 00:23:23.763 --> 00:23:27.563 But they're also like some of, some of this is diminishing returns on where it 00:23:27.563 --> 00:23:31.893 is right now is perfectly fit for all the things that I continue to need, right? 00:23:32.123 --> 00:23:36.423 And I am like this close to moving to a private model, right? 00:23:36.423 --> 00:23:39.933 Probably the next, uh, the next Mac upgrade that I get, 00:23:39.953 --> 00:23:41.053 I'll go, I'll be fine, right? 00:23:41.053 --> 00:23:44.823 Like I won't necessarily need a frontier model for most of the things that I do. 00:23:45.333 --> 00:23:51.323 But, um, so I was thinking to myself, so like maybe most of these companies should 00:23:51.323 --> 00:23:55.143 just maybe look at self-hosted models and, and for, for some of this information. 00:23:55.173 --> 00:23:58.413 Won't be as fast, but you know, if you're just doing some data analysis and let 00:23:58.413 --> 00:24:00.103 it churn for an hour, then great, okay. 00:24:00.133 --> 00:24:01.543 It spits out something relevant. 00:24:01.993 --> 00:24:04.903 I was having a conversation with, uh, uh, John Dobbin. 00:24:04.923 --> 00:24:07.883 I'll give him a shout-out for, for the, the insight on this. 00:24:07.931 --> 00:24:08.451 Callen Sapien CEO Synthreo: Yeah 00:24:08.613 --> 00:24:10.023 Todd Kane: he scared the crap out of me on this. 00:24:10.023 --> 00:24:14.933 He's like, "So here's the thing. Most of the self-hosted models, like 00:24:15.163 --> 00:24:18.513 the open source models, there's, there's very little guardrails and 00:24:18.513 --> 00:24:20.253 containers built around them," right? 00:24:20.253 --> 00:24:24.803 Like the frontier models have a lot of, uh, capabilities around prompt injection 00:24:24.853 --> 00:24:28.623 that they will not do things that are, that are gonna be nefarious, right? 00:24:28.623 --> 00:24:33.243 But like, uh, what, what the, the sort of the, the idea that he, that he suggested 00:24:33.243 --> 00:24:38.373 is, say you have like a website where you can have clients submit tickets, right? 00:24:38.433 --> 00:24:42.183 And someone goes in and submits a malicious tickets that gets processed 00:24:42.203 --> 00:24:46.353 by a local AI that doesn't have good prompt injection, uh, management 00:24:46.353 --> 00:24:49.613 on it, and now all of a sudden it starts exfilling data based on 00:24:50.033 --> 00:24:51.763 that, that, that ticket submission. 00:24:51.763 --> 00:24:55.553 I was like, "Oh, crap. Okay. Private hosted models may not 00:24:55.553 --> 00:24:56.913 be the best play here," right? 00:24:56.943 --> 00:24:59.713 So it kind of scared me of like, there's, there's so many risk 00:24:59.713 --> 00:25:03.123 and reward trade-offs between the different approaches here, right? 00:25:04.155 --> 00:25:04.625 Callen Sapien CEO Synthreo: Mm-hmm. 00:25:04.965 --> 00:25:06.645 Um, no, it's spot on. 00:25:06.665 --> 00:25:10.965 There's, there's, there's really two challenges currently with the… 00:25:11.005 --> 00:25:15.925 Outside of the fact that you gotta probably spend truly to, to, to get 00:25:15.935 --> 00:25:21.005 something really meaningful, you probably spend eight to $10,000 just 00:25:21.005 --> 00:25:24.845 with component costs right now to, to, to get something that, that can run. 00:25:24.885 --> 00:25:28.185 And then, and then you've got energy costs and the, the other 00:25:28.185 --> 00:25:30.505 pieces because GPUs are heat. 00:25:30.585 --> 00:25:33.705 Uh, we, we, we, we went through this with the mining era of 00:25:33.829 --> 00:25:34.099 Todd Kane: Right 00:25:34.399 --> 00:25:34.605 Callen Sapien CEO Synthreo: right? 00:25:34.665 --> 00:25:37.945 Uh, and, and so we, we know that, you know, there's probably gonna be a lot 00:25:37.945 --> 00:25:41.815 more solar panels popping up, uh, for, for people in a little while, or geothermal. 00:25:42.355 --> 00:25:45.475 Uh, but the other piece to the local models, I, I do agree with 00:25:45.475 --> 00:25:48.225 John that, that, that there are… But you can, you could put the 00:25:48.225 --> 00:25:49.485 same type of guardrails, right? 00:25:49.485 --> 00:25:53.555 They do have lower context windows, right, which is a problem, uh, because 00:25:53.555 --> 00:25:56.475 you fill up pretty quick and you don't, you don't know what you can attack. 00:25:56.725 --> 00:25:59.695 Uh, but I do think that is, this is absolutely w- going to be part 00:25:59.695 --> 00:26:00.825 of the solution in the future. 00:26:01.055 --> 00:26:04.025 Uh, but one of the big challenges with these models right now 00:26:04.025 --> 00:26:05.285 is the hallucination rate. 00:26:05.735 --> 00:26:11.655 Uh, we see with locally hosted models, uh, in the 40s, 50% hallucination rate 00:26:11.695 --> 00:26:13.355 over the course of a conversation. 00:26:13.595 --> 00:26:19.105 And so you, you really need a lot of human intervention, both from the security 00:26:19.105 --> 00:26:22.245 side that John was talking about, but also from the business outcome side. 00:26:22.245 --> 00:26:26.175 The, the models are more prone to hallucination because the math isn't 00:26:26.175 --> 00:26:30.863 as strong, and the context is worse, and the memories are not as powerful. 00:26:31.163 --> 00:26:31.773 Todd Kane: So that it, 00:26:32.045 --> 00:26:32.685 Callen Sapien CEO Synthreo: really where 00:26:32.793 --> 00:26:35.813 Todd Kane: the, the, uh, let me dig on this 'cause like that, that is a function 00:26:35.863 --> 00:26:40.473 of smaller context window by nature because of limited infrastructure and 00:26:40.483 --> 00:26:44.683 something to do with the models themselves or sort of the parameters that are built 00:26:44.683 --> 00:26:46.383 around it, or just the context window? 00:26:47.451 --> 00:26:49.091 Callen Sapien CEO Synthreo: It's, it's, it's a combination 00:26:49.091 --> 00:26:51.641 of, of kind of all the pieces. 00:26:51.881 --> 00:26:54.799 The smaller models with the lower parameters have 00:26:55.099 --> 00:26:55.389 Todd Kane: Yep. 00:26:56.081 --> 00:26:56.569 Callen Sapien CEO Synthreo: capacity, 00:26:56.869 --> 00:26:57.169 Todd Kane: Yep 00:26:57.261 --> 00:26:57.781 Callen Sapien CEO Synthreo: what you said. 00:26:58.011 --> 00:27:01.261 Uh, you know, to put another way, it's like, you know, using, 00:27:01.361 --> 00:27:02.561 uh, less of your brain, right? 00:27:02.561 --> 00:27:04.881 A lot of those models were designed to have, for example, 00:27:04.881 --> 00:27:08.611 like 32 billion parameters that they can take on at a given time. 00:27:08.871 --> 00:27:12.531 We neuter them down to seven to eight billion parameters. 00:27:12.791 --> 00:27:17.421 So you've got a prediction engine that is trying to predict something 00:27:17.521 --> 00:27:22.521 with a quarter of the context and the memory capabilities that it has. 00:27:22.731 --> 00:27:28.561 So it does work great for, for certain repetitive tasks, QA, processing of, 00:27:28.571 --> 00:27:31.381 of, of almost a deterministic behavior. 00:27:31.651 --> 00:27:35.121 Uh, but where it, where it really starts to, to fall down is when we, when 00:27:35.121 --> 00:27:39.801 we start getting into, uh, you know, adding in multi-step processes, right? 00:27:39.831 --> 00:27:44.481 Uh, multi-step is really the big part that, that gets hit, and that's where 00:27:44.641 --> 00:27:48.351 we can add in different workflows and loops and, and these other 00:27:48.351 --> 00:27:50.201 things, but it, it definitely… 00:27:50.211 --> 00:27:54.381 It's actually one of the main reasons why John's security breakdown happens 00:27:54.381 --> 00:27:59.411 as well, because how you overwhelm and, and overcome the guardrails largely 00:27:59.621 --> 00:28:06.311 with these models is either by injecting a, a, a, a, a mistruth and calling 00:28:06.311 --> 00:28:08.831 it on it, and then, and then now it's like, "Oh, I was wrong on this. I 00:28:08.831 --> 00:28:12.841 lied, so now I'm gonna relax my other things because, because this is here." 00:28:12.871 --> 00:28:15.081 Or then context bombing. 00:28:15.091 --> 00:28:21.541 So I'm gonna drop in 50% of my context window because now it's the guardrails 00:28:21.591 --> 00:28:26.441 were maybe 30% of every turn in the conversation for the first half of 00:28:26.441 --> 00:28:28.731 this conversation, now they're 18%. 00:28:28.831 --> 00:28:32.841 So as that math is, is, is expanding, the, those, those guardrails just 00:28:32.841 --> 00:28:34.701 have less weight in the conversation. 00:28:34.721 --> 00:28:38.281 There's still more weight than the, than the, than the attack, but they 00:28:38.281 --> 00:28:42.411 have less weight overall, and then it's just about consistency in attacking it. 00:28:42.711 --> 00:28:43.573 Todd Kane: Especially given this… 00:28:43.711 --> 00:28:43.851 Callen Sapien CEO Synthreo: bel- 00:28:44.163 --> 00:28:44.463 Todd Kane: Go ahead. 00:28:45.141 --> 00:28:45.891 Callen Sapien CEO Synthreo: No, no, go ahead 00:28:45.933 --> 00:28:49.553 Todd Kane: Well, I, I think it's, that's relevant specifically for sort 00:28:49.553 --> 00:28:53.693 of the use case that I'm advocating for, in, in particular is ticket analysis. 00:28:53.703 --> 00:28:54.343 So you, 00:28:54.525 --> 00:28:55.015 Callen Sapien CEO Synthreo: Yes 00:28:55.403 --> 00:29:00.553 Todd Kane: say, even s- three months of data, that's still a lot of data to dump 00:29:00.553 --> 00:29:06.293 into, uh, like, that's certainly not gonna fit in a 200,000K, uh, window, right? 00:29:06.693 --> 00:29:07.723 Or 200K window. 00:29:08.073 --> 00:29:12.903 Uh, e- even a million, I think, like, it might fill up a fair bit, but, you know, 00:29:12.913 --> 00:29:14.723 you gotta be sensitive to that, right? 00:29:15.437 --> 00:29:18.997 Callen Sapien CEO Synthreo: We, we, um, we're handling this in a, in a way where 00:29:18.997 --> 00:29:24.777 we're tr- what we're trying to do is, is not offload the, the model itself, 00:29:24.777 --> 00:29:28.937 but offload the compute, because that will reduce some of the, the, the cost. 00:29:29.157 --> 00:29:33.167 And so what we do is we try and keep the thinking and try and keep the 00:29:33.167 --> 00:29:36.537 actions local, because that's, that's more controllable and definable. 00:29:36.877 --> 00:29:38.697 Uh, and, and then batching. 00:29:38.757 --> 00:29:42.947 I, I think batching is highly underutilized by practitioners, 00:29:43.047 --> 00:29:45.537 and this is where the Ticketdata side can come in very well. 00:29:45.797 --> 00:29:49.947 If I can, if I can run through a recognition pattern, whether that's 00:29:49.957 --> 00:29:53.717 machine learning or whether that's a SLM or whether that's just a model, 00:29:53.897 --> 00:29:56.677 and say, and categorize these things, then I can send them through in 00:29:56.677 --> 00:29:59.077 batches, and then I can reconnect. 00:29:59.297 --> 00:30:02.097 Uh, Gary Vee just, uh, said something a couple days ago 00:30:02.097 --> 00:30:03.117 that I really, really liked. 00:30:03.127 --> 00:30:05.907 Um, and, uh, I've followed him for a long time. 00:30:05.907 --> 00:30:08.917 I didn't expect him to become an AI guru, but, you know, the guy's 00:30:08.927 --> 00:30:10.367 a guru in so many other areas. 00:30:10.667 --> 00:30:15.117 But he calls it the 15/85 rule, and he, and he started it with his, 00:30:15.327 --> 00:30:16.767 with his, like, leadership team. 00:30:16.977 --> 00:30:19.567 He wants to be involved in the first 15% of the planning and 00:30:19.567 --> 00:30:24.367 strategy, let them go off and do 80%, and then back there for the 5%. 00:30:24.367 --> 00:30:28.477 And I think that when we apply it to AI, it's very similar. 00:30:28.737 --> 00:30:31.877 Let me be part of the planning and the strategy, go do your work, 00:30:32.087 --> 00:30:33.597 uh, and then, and then come back. 00:30:33.607 --> 00:30:37.707 And we could actually probably localize those much better than, than otherwise. 00:30:37.707 --> 00:30:39.797 I think Zofik was on, was on that path. 00:30:39.797 --> 00:30:43.347 I d- you know, I haven't talked as much to Lee since their acquisition, 00:30:43.347 --> 00:30:45.977 but, but I think that was really what he was trying to solve for was 00:30:45.977 --> 00:30:49.347 because we had people running these local models, and so he, he took that 00:30:49.587 --> 00:30:51.615 on th- uh, upon theirselves to do it. 00:30:51.915 --> 00:30:54.365 Todd Kane: It's funny 'cause like this morning actually in my group coaching 00:30:54.365 --> 00:30:58.075 session, I, I, I talk about this model a lot, and it's 10 / 80 / 10. 00:30:58.075 --> 00:30:59.295 So, you know, he's got the, 00:30:59.387 --> 00:30:59.717 Callen Sapien CEO Synthreo: Yeah. 00:31:00.017 --> 00:31:01.345 Todd Kane: the 50 and 80 and five. 00:31:01.655 --> 00:31:04.125 Uh, so yeah, my model is, is 10 / 80 / 10. 00:31:04.125 --> 00:31:07.695 I don't remember originally where I got this from, but it's, it's an industry, 00:31:07.795 --> 00:31:11.515 uh, sort of leverage productivity model, especially for delegation. 00:31:11.585 --> 00:31:16.235 This is where like I originally started using this in, uh, uh, avoiding 00:31:16.235 --> 00:31:18.005 executive swoop and poop, right? 00:31:18.005 --> 00:31:22.065 Because the, the executive would sort of get in there and mess around with stuff, 00:31:22.085 --> 00:31:25.415 and they didn't set up the parameters for what the delegation was or what 00:31:25.415 --> 00:31:28.315 they wanted, so they would make a mess and then sort of show up at random 00:31:28.315 --> 00:31:30.225 times and not pr-produce any value. 00:31:30.255 --> 00:31:34.135 So I would tell people like, "Okay, use the 10 / 80 / 10 model for delegation, 00:31:35.085 --> 00:31:38.955 10% l- front loading expectations, delivery, what you wanna protect 00:31:38.955 --> 00:31:42.565 against, then let them go and do 80% of the work and then 10% cleanup on the 00:31:42.565 --> 00:31:46.315 end." So exactly the same, same model, slightly different, different numbers. 00:31:46.315 --> 00:31:50.005 But where I think this is particularly useful is delegation. 00:31:50.045 --> 00:31:52.525 Like that's where I talk about this framework a lot, right? 00:31:52.565 --> 00:31:57.235 And what-- I think what you're pointing to is we need to start thinking more about 00:31:57.335 --> 00:32:02.895 agents as independent, sort of autonomous employees doing particular work. 00:32:03.205 --> 00:32:07.095 And it's weird because like it's not necessarily contiguous, but I think 00:32:07.095 --> 00:32:11.805 we still wanna think about it in a 10 / 80 / 10, uh, delegation exercise 00:32:11.815 --> 00:32:13.305 of like, "Here's the parameters. 00:32:13.305 --> 00:32:14.145 This is what I want. 00:32:14.145 --> 00:32:15.225 This is the expectations. 00:32:15.225 --> 00:32:16.955 Don't do this, do this," all of those things. 00:32:17.225 --> 00:32:20.735 Let it churn for 80% and then inspect it on the back end, right? 00:32:20.775 --> 00:32:25.945 And like I'm sure you've heard this term meat proxy, um, which is, uh, uh, people 00:32:25.945 --> 00:32:29.985 that just take whatever comes from AI and turns around and, and spits it out. 00:32:30.145 --> 00:32:32.885 Uh, and this is where like work slop is coming from, right? 00:32:33.145 --> 00:32:37.655 The, the whole idea of, uh, you know, I, I, I was gonna send an email. 00:32:37.855 --> 00:32:42.395 I had ChatG- ChatGPT punch it up to like a three-page or a three-paragraph 00:32:42.395 --> 00:32:44.625 summary with lots of information and data. 00:32:45.015 --> 00:32:49.105 Uh, I emailed that over to somebody else, and then that person uses their AI agent 00:32:49.115 --> 00:32:53.095 to digest it into like the f- the, the much smaller version of this, right? 00:32:53.095 --> 00:32:55.695 So it, it's just meat proxy to meat proxy, and we're kinda 00:32:55.705 --> 00:32:58.145 uselessly using AI in, in between. 00:32:58.515 --> 00:32:59.715 Uh, so now I'm just ranting. 00:32:59.735 --> 00:33:04.335 But the, I think the, the, the, the, the way of, uh, using delegation 00:33:04.885 --> 00:33:08.385 frameworks for AI and the similarities that we should be applying in 00:33:08.385 --> 00:33:13.385 general, uh, sort of collaboration and management is very on par, right? 00:33:15.033 --> 00:33:17.583 Callen Sapien CEO Synthreo: I, I, it, it actually uncovers a very human 00:33:17.583 --> 00:33:21.523 problem that we're dealing with right now, and, and, uh, it's-- there, 00:33:21.523 --> 00:33:22.793 there, and there's two pieces to it. 00:33:23.123 --> 00:33:28.993 We have looked at leadership as a… And I mean leadership and 00:33:28.993 --> 00:33:32.145 managerial, because you don't have to have reports to be a leader. 00:33:32.445 --> 00:33:32.655 Todd Kane: Mhm 00:33:32.793 --> 00:33:35.863 Callen Sapien CEO Synthreo: looked at that as there's a class of people that 00:33:35.863 --> 00:33:40.643 have earned that right, and we're going to invest in those people because we think 00:33:40.643 --> 00:33:42.343 that they would be good leaders of humans. 00:33:42.753 --> 00:33:48.353 Uh, in order to be a powerful AI user and actually use it in a way that is 00:33:48.353 --> 00:33:52.293 going to impact in a positive way your, your clients, yourself or, or your 00:33:52.303 --> 00:33:54.623 mission, you have to be a good delegator. 00:33:55.043 --> 00:34:00.343 And we've gatekept these things to the, to the, to the upper class of this. 00:34:00.363 --> 00:34:02.833 And including, we've told people like, "Well, I don't think 00:34:02.833 --> 00:34:03.963 you're cut out for leadership." 00:34:03.963 --> 00:34:05.813 Well, now you have to be a leader. 00:34:05.993 --> 00:34:09.933 And I think it's going to-- And we-- I've actually seen a rise in more like 00:34:09.943 --> 00:34:13.103 HR consultants going into leadership management, teaching these things. 00:34:13.343 --> 00:34:17.553 It's also, uh, been a challenge for people that have, uh, globalized 00:34:17.563 --> 00:34:22.503 talent because we ma- we, we globalize the talent and we, we, we offload, 00:34:23.403 --> 00:34:29.963 workflows and intellect to individuals to follow a process with no deviations. 00:34:29.983 --> 00:34:35.243 And, and so, you know, comp- countries like the Philippines, uh, certain areas 00:34:35.243 --> 00:34:42.173 of India, uh, Vietnam, Indonesia, uh, Pa- Pakistan, these areas have, have 00:34:42.183 --> 00:34:47.803 built, you know, farms and, and, and, and cities mul- you know, um, uh, 00:34:47.833 --> 00:34:52.663 of, of people that, that are going to follow these things, and if they don't 00:34:52.663 --> 00:34:54.513 follow it every step, they're fired. 00:34:54.863 --> 00:34:59.413 And we're asking them to adopt the management capabilities and the 00:34:59.413 --> 00:35:04.113 creativity that we've trained out of them for 45 years as we've outsourced. 00:35:04.133 --> 00:35:08.693 And it's, it, it-- and a lot of MSPs use globalized talent, and it's something 00:35:08.693 --> 00:35:13.283 that, that we're going to really have to address in a very short amount of time. 00:35:13.543 --> 00:35:16.773 Uh, otherwise, the token cost is not gonna be worth it, in my opinion. 00:35:17.147 --> 00:35:19.497 Todd Kane: Yeah, and that's the other issue that I think we're facing here. 00:35:19.537 --> 00:35:22.977 Maybe, uh, slightly off track, but, you know, tangen-tangentially 00:35:22.977 --> 00:35:27.927 related is, um, the over-indexing on sort of the capabilities of AI 00:35:27.927 --> 00:35:31.247 and where they can actually fit and, you know, the token cost for those. 00:35:31.247 --> 00:35:35.767 People are now going, uh, you know, uh, it was sort of the last six months, 00:35:35.797 --> 00:35:39.127 eight months was token maxing, token maxing, we're gonna do amazing things, 00:35:39.127 --> 00:35:43.747 20% layoffs, and then they're like, "Holy crap, why are our costs up 70%," right? 00:35:44.407 --> 00:35:45.557 And so they were token maxing. 00:35:45.567 --> 00:35:49.577 Now it costs more than the people that they had sort of doing jobs before, and 00:35:49.577 --> 00:35:52.867 maybe not achieving the same ends that they thought coming into it, right? 00:35:53.147 --> 00:35:56.427 So I think this will be an interesting trend to watch, but I, I think 00:35:56.717 --> 00:36:00.147 what you're, you're alluding to is more of sort of a business process 00:36:00.147 --> 00:36:06.327 thing of like, how do we, how do we compartmentalize and really understand 00:36:06.347 --> 00:36:11.967 the work that we're doing, and how do we intelligently hand it off to workers or 00:36:12.107 --> 00:36:15.657 digital workers in a way that actually produces the results that we want? 00:36:15.707 --> 00:36:19.427 I think that's gonna be a really fascinating exercise going forward, right? 00:36:20.635 --> 00:36:24.005 Callen Sapien CEO Synthreo: Yeah, I mean, that's i- in the-- We're, we're, 00:36:24.175 --> 00:36:28.785 we're coming up on a year officially really in market, uh, in, in November. 00:36:29.235 --> 00:36:34.445 And in that year, we've had really two pretty dramatic shifts 00:36:34.455 --> 00:36:37.415 onto my company's identity and, and the problems that we solve. 00:36:37.725 --> 00:36:42.605 There, it's the same mission, but, uh, technology and the landscape has shifted 00:36:42.605 --> 00:36:47.765 so much that, that we've realized that we need to be, uh, the optim- the, like, 00:36:47.765 --> 00:36:54.255 the, the greatest optimizer of delivering those outcomes for the business class. 00:36:54.265 --> 00:36:58.795 So what I, how I would, how I would liken that is basically like Cursor for 00:36:58.815 --> 00:37:03.455 delivering results with AI, Uh, and, and Cursor does deliver results for AI. 00:37:03.455 --> 00:37:04.795 It's just specifically around coding. 00:37:05.135 --> 00:37:09.035 And so we've, we've built it, and it's to the point where I think you mentioned 00:37:09.035 --> 00:37:10.465 the diminishing… No, uh, I don't think. 00:37:10.475 --> 00:37:12.515 You did mention diminishing returns. 00:37:12.885 --> 00:37:18.505 And we can get out of a, a, an open weight model like Inkling similar 00:37:18.505 --> 00:37:23.995 to Fable 5.0 results, uh, with, with our Wingtip engine, right? 00:37:24.015 --> 00:37:28.885 And, and the cost of that is a 10th or, or a 40th, or if you're willing 00:37:28.885 --> 00:37:33.925 to use a GLM, maybe a 100th to get, to get 90, 95% of the output. 00:37:34.575 --> 00:37:38.655 The other piece is that infrastructure piece, and I think it's a, the, 00:37:38.685 --> 00:37:42.335 probably the biggest opportunity with AI, though, for the MSP 00:37:42.335 --> 00:37:48.925 market, is we gave up infrastructure when we, when we went to 365. 00:37:49.015 --> 00:37:50.655 You've been in the industry long enough. 00:37:51.135 --> 00:37:56.945 I-- fortunately, cybersecurity came quick enough, but I remember talks of 00:37:56.945 --> 00:38:04.225 layoffs, panic in 2016, 2017, as we moved from, "We own all the infrastructure. 00:38:04.235 --> 00:38:08.175 We're building data centers," to, "We're getting 20%, 10%, 8% of what 00:38:08.175 --> 00:38:11.815 Microsoft will give us, and, and now we've got to survive on this." 00:38:12.055 --> 00:38:13.995 Then security came in, it saved us. 00:38:14.065 --> 00:38:17.695 Uh, and but we're now back in the spot where we get to manage and 00:38:17.695 --> 00:38:21.085 maintain the infrastructure again, and we need the right people that 00:38:21.085 --> 00:38:22.675 can figure out how to do those tasks. 00:38:22.885 --> 00:38:25.055 But then we need to figure out, like, where does that live? 00:38:25.055 --> 00:38:26.075 What does this look like? 00:38:26.075 --> 00:38:27.165 And how does that scale? 00:38:27.445 --> 00:38:32.135 And it's a huge opportunity, though, that, that exists, uh, in, in the space 00:38:32.145 --> 00:38:37.535 because I think we can get back to the 60, 70% margins and have partners 00:38:37.535 --> 00:38:41.195 that are absolutely willing to pay it because they're getting so much value. 00:38:41.205 --> 00:38:44.495 But it requires exactly what you said, the right delegation, the 00:38:44.495 --> 00:38:47.645 right harnesses, and the right, uh, infrastructure to, to do that, 00:38:47.715 --> 00:38:49.385 Todd Kane: Yeah, and that's the biggest thing here is, is 00:38:49.685 --> 00:38:49.905 Callen Sapien CEO Synthreo: at. 00:38:49.925 --> 00:38:53.365 Todd Kane: yeah, there's always where there's complexity, there's margin, right? 00:38:53.405 --> 00:38:57.055 And we're now getting to the space where like y- you can't just sort 00:38:57.055 --> 00:38:59.305 of, "Oh, just go get ChatGPT," right? 00:38:59.305 --> 00:39:03.305 I think that there's sort of like layers of where MSPs are going wrong in their 00:39:03.495 --> 00:39:06.835 AI strategy, and I've been ranting about this as like that's why I sort of frame 00:39:06.835 --> 00:39:11.325 it as like pulling their clients to cloud, pulling their clients s- to security. 00:39:11.325 --> 00:39:12.105 Like, where are you guys on AI? 00:39:12.105 --> 00:39:13.835 You're not leading, right? 00:39:13.845 --> 00:39:16.645 Like there's so much shadow AI going on, it's insane. 00:39:16.965 --> 00:39:18.875 So first there's just the governance aspect of that I 00:39:18.885 --> 00:39:20.255 think is a massive opportunity. 00:39:20.635 --> 00:39:24.575 Um, but I think like the, the work that you guys are doing is a huge opportunity 00:39:24.605 --> 00:39:28.465 of actually stepping into, you know, the industry term Pax8 is, is sort 00:39:28.465 --> 00:39:34.325 of, uh, advocating for is, um, is, uh, managed intelligence provider, right? 00:39:34.355 --> 00:39:38.285 And right now, a lot of people are just like, "Here, let me set you up with 00:39:38.315 --> 00:39:43.505 Copilot." And it's like, well, okay, like that's easy, but like that's not 00:39:43.525 --> 00:39:45.105 gonna produce a ton of value for people. 00:39:45.145 --> 00:39:49.625 I honest- I kind of firmly believe this is why we're seeing so many damning 00:39:49.625 --> 00:39:53.425 stats in the industry of AI initiatives failing in enterprise businesses, 00:39:53.455 --> 00:39:55.005 'cause they're just like, "Here you go. 00:39:55.005 --> 00:39:55.655 Here's AI. 00:39:55.665 --> 00:39:56.845 You're gonna do great," right? 00:39:56.885 --> 00:39:59.865 Like, "Here you go, Tommy." And like nothing comes out of that 'cause they 00:39:59.865 --> 00:40:01.015 don't know what to do with it, right? 00:40:01.035 --> 00:40:02.025 Like, no one's trained them. 00:40:02.035 --> 00:40:04.315 They don't know how to use those models beyond just asking 00:40:04.315 --> 00:40:05.665 for recipes and stuff, right? 00:40:05.985 --> 00:40:08.335 So the, of course, that's not gonna produce anything. 00:40:08.615 --> 00:40:13.135 The second layer would be, you know, "Hey, let me do consulting with you. We'll 00:40:13.145 --> 00:40:17.135 s- we'll set up, uh, like so a Claude account or an enterprise GPT account 00:40:17.155 --> 00:40:21.665 and get everyone sort of rolled out on a maybe a arguably a better model," right? 00:40:21.705 --> 00:40:23.325 And maybe this is the way to go about it. 00:40:23.325 --> 00:40:26.935 And now it's just consulting with reselling kind of a SaaS product. 00:40:27.205 --> 00:40:30.555 But I think what you guys are doing is, is much more the hybrid of like, "I'm 00:40:30.555 --> 00:40:35.735 gonna give you a console that allows you to do like open weight routing on to 00:40:35.735 --> 00:40:40.915 different models and give you the privacy protection and, and the portability to 00:40:40.925 --> 00:40:44.115 multiple models." And I think this is massive 'cause of your point, like when 00:40:44.115 --> 00:40:48.075 we were, we were talking before, maybe you can d- you can dialogue on this story 00:40:48.075 --> 00:40:52.475 a little bit of like you had a client, like they were paying 50,000 a month 00:40:52.565 --> 00:40:58.465 in self-hosted Claude plus Azure, and you guys slashed that to like just over 00:40:58.465 --> 00:41:02.425 10,000 a month with, uh, with, uh, the, the platform that you guys are doing. 00:41:02.685 --> 00:41:05.615 So I think this is the value that you can bring to people is like, not only 00:41:05.615 --> 00:41:09.445 am I giving you the capabilities for a fraction of the cost, but I'm actually 00:41:09.455 --> 00:41:13.545 able to manage it for you rather than just sort of like, "Okay, hang on, let me 00:41:13.805 --> 00:41:18.495 try and log in as an admin to your, to y- to your, your, uh, your profile," right? 00:41:18.745 --> 00:41:20.845 Maybe you can expand on sort of how you guys are doing that. 00:41:22.139 --> 00:41:24.379 Callen Sapien CEO Synthreo: Yeah, I, I think, you know, the most 00:41:24.419 --> 00:41:31.349 forward-looking, that we end up encountering are dealing with actually 00:41:31.359 --> 00:41:34.029 some of the initial 365 problems, right? 00:41:34.029 --> 00:41:36.769 The federated access into all these portals. 00:41:36.769 --> 00:41:41.939 We-- There was a partner of ours that, uh, they were wor- they were using N8N, and 00:41:41.969 --> 00:41:44.679 they, they were around with my co-founder, and they're like, "You know, the, my first 00:41:44.679 --> 00:41:48.845 thing I do every day is log into 38 N8N portals and see if everything's okay." 00:41:49.145 --> 00:41:49.205 Todd Kane: Ugh 00:41:49.419 --> 00:41:51.799 Callen Sapien CEO Synthreo: It's like, that is, that is not… I mean, like 00:41:51.899 --> 00:41:55.019 none of us got into… I mean, maybe somebody got into it to do that type 00:41:55.113 --> 00:41:56.783 Todd Kane: Some, some automation nut, yeah 00:41:57.239 --> 00:41:57.699 Callen Sapien CEO Synthreo: Yeah, yeah. 00:41:57.699 --> 00:42:01.689 There, there are some you know, that, that exist in the world. 00:42:01.709 --> 00:42:04.849 Uh, but, uh, and I, I don't want to yuck anybody's yum. 00:42:05.079 --> 00:42:08.329 Uh, but, uh, but at the same time, that isn't sustainable, right? 00:42:08.359 --> 00:42:11.809 And, and you're gonna miss something, and something's gonna crash, and, 00:42:11.809 --> 00:42:13.479 and, and we're, you're gonna miss it. 00:42:13.519 --> 00:42:18.227 And, and so for us, we, we do look to, to, to bring that You said 00:42:18.227 --> 00:42:22.497 it, manageability, guidability, governance over, over the piece. 00:42:22.507 --> 00:42:26.627 W- w- it's interesting 'cause we also had a, a recent, uh, partner 00:42:26.627 --> 00:42:28.667 that is a Claude network partner. 00:42:28.687 --> 00:42:31.097 They're, they're reselling… Not really reselling Claude, but they're 00:42:31.097 --> 00:42:32.467 reselling a service around Claude. 00:42:32.827 --> 00:42:35.387 And they had a, they had a partner that said, "You know, I'm, I'm, I'm 00:42:35.407 --> 00:42:37.737 over." You know, we started this conversation by saying, "Hey, a 00:42:37.737 --> 00:42:41.317 new model might have dropped," uh, and, and because it all sucks now. 00:42:41.597 --> 00:42:45.407 And, uh, and they, they said, "You know, I'm tired of this, this rat race. 00:42:45.717 --> 00:42:48.977 I need to go to something that's agnostic because if I pick a winner 00:42:48.977 --> 00:42:53.397 right now, I'm gonna be jumping every six months to try and get the most out 00:42:53.397 --> 00:42:56.907 of this stuff." And they said, "Okay, so what do we do with the last eight 00:42:56.907 --> 00:42:58.247 months that we've been in Claude? 00:42:58.647 --> 00:43:00.227 Do, how do I export this work? 00:43:00.237 --> 00:43:01.647 Where do I put my artifacts?" 00:43:01.817 --> 00:43:06.137 And, you know, the MSP's like, "I've, I don't know. We'll, we'll export your 00:43:06.137 --> 00:43:09.997 chat history that we can, and we'll, you know, see what we can do about your, your, 00:43:09.997 --> 00:43:14.557 uh, your projects." But the portability issue and the, the, the other pieces that 00:43:14.557 --> 00:43:16.887 people are just starting to understand. 00:43:17.187 --> 00:43:20.087 You know, I- we call it internally the Claude wall, uh, but it 00:43:20.087 --> 00:43:21.667 could be my chat tool wall. 00:43:21.837 --> 00:43:22.917 It could be Copilot. 00:43:22.927 --> 00:43:24.457 It could be ChatGPT. 00:43:24.467 --> 00:43:26.777 It could be, you know, uh, devs.ai. 00:43:26.777 --> 00:43:27.287 It could be HATS. 00:43:27.297 --> 00:43:27.977 It could be anybody. 00:43:28.007 --> 00:43:32.277 When we hit that personal productivity wall that you alluded to, it, you just, 00:43:32.287 --> 00:43:35.947 you can't get anything else out of it, and now your two options are, do I go the 00:43:35.947 --> 00:43:40.087 old way and build an app, and I manage that app and the infrastructure, or do I 00:43:40.087 --> 00:43:41.867 figure out how to use the modern stuff? 00:43:41.917 --> 00:43:44.107 And that's, that's where we come in and some of the others come 00:43:44.107 --> 00:43:48.677 in to deliver agents, uh, without having to host the infrastructure. 00:43:48.677 --> 00:43:52.107 W- build MCPs without having to spin something up. 00:43:52.117 --> 00:43:56.057 We had a partner the other day that said that one of their clients connected 00:43:56.057 --> 00:44:02.967 their banking software through our Excel app, uh, using a, a, an MCP 00:44:02.967 --> 00:44:06.757 they found on GitHub because Claude told them to, to, to go do that. 00:44:07.077 --> 00:44:09.927 And it had two stars and no comments, 00:44:10.279 --> 00:44:11.359 Todd Kane: This sounds scary as hell. 00:44:11.997 --> 00:44:12.757 Callen Sapien CEO Synthreo: banking through 00:44:12.959 --> 00:44:13.159 Todd Kane: Yeah 00:44:13.227 --> 00:44:17.067 Callen Sapien CEO Synthreo: And so, like, those are areas where governance makes 00:44:17.067 --> 00:44:19.337 sense, and they still need to do that. 00:44:19.497 --> 00:44:20.877 So build the MCP. 00:44:21.447 --> 00:44:21.917 Uh, offer it. 00:44:21.927 --> 00:44:24.467 Even if you don't use 3.0, that's, that's… For me, when we build 00:44:24.467 --> 00:44:26.217 an MCP, it's universal, right? 00:44:26.217 --> 00:44:30.777 If someone is, is dedicated, if they're a maxer and they're using Ultra and they're 00:44:30.787 --> 00:44:33.947 going, you know, in Groq, and they're, you know, they're taking their toket subs- 00:44:34.027 --> 00:44:38.697 token subsidies to the next level, don't necessarily want them running on my, my, 00:44:38.737 --> 00:44:43.267 my, on, on that area, but we still need to give MSPs a way of managing that. 00:44:43.277 --> 00:44:47.127 So that's when we, we build MCPs, governance policies, observability. 00:44:47.297 --> 00:44:51.947 And I think all of the things that we've, we managed before that I hope people 00:44:51.947 --> 00:44:56.247 don't take this pejoratively or, or, or as a wagging finger, but that we've 00:44:56.257 --> 00:45:01.087 kind of went away from over the last five or six years as we've resold a lot 00:45:01.087 --> 00:45:04.997 of SaaS and a lot of security products without wrapping services around it, 00:45:05.267 --> 00:45:08.257 uh, is, uh, is it needs to be relearned. 00:45:08.597 --> 00:45:14.207 And one thing that's a little scary to me is we've had so much M&A in 00:45:14.207 --> 00:45:22.329 the last 10 years that I wonder how many, like, experienced service 00:45:22.329 --> 00:45:23.969 delivery people exist out there. 00:45:23.969 --> 00:45:25.769 You mentioned John Da- Dobbin. 00:45:26.149 --> 00:45:28.899 I, like, that guy's worth his weight in gold. 00:45:29.109 --> 00:45:33.009 He has 24 years of delivering service. 00:45:33.129 --> 00:45:40.239 If I was an MSP, I would be grabbing onto him and saying, "Okay, should we sell AI, 00:45:40.269 --> 00:45:42.099 and how should we build that offering?" 00:45:42.299 --> 00:45:45.859 And find that service manager that is so passionate and had been in that thing, 00:45:46.089 --> 00:45:50.089 and have them reteach the entire company on how to deliver services at scale, 00:45:50.109 --> 00:45:52.009 and then base your AI policy on that. 00:45:52.289 --> 00:45:56.389 That's, it's get back to the basics, which sounds silly, but it's, 00:45:56.409 --> 00:45:59.727 it, it's, it, uh, the, the idioms are, are there for a reason, you 00:46:00.027 --> 00:46:01.617 Todd Kane: Yeah, there's so much truth in that. 00:46:01.647 --> 00:46:05.457 Like, I find everything in business and somewhat, uh, I guess a large degree in 00:46:05.457 --> 00:46:08.327 life is you, you getting back to basics. 00:46:08.387 --> 00:46:10.947 It's just we've gotten so far away from the simple things. 00:46:10.957 --> 00:46:14.957 Like, I always tie it back to, uh, you know, the John Wooden a- approach, 00:46:14.967 --> 00:46:16.917 very famous basketball coaches. 00:46:16.937 --> 00:46:18.967 We're gonna start with tying our shoes, and everyone's like, "What the hell 00:46:18.967 --> 00:46:21.327 are we doing?" It's like, eh, you know, tie your shoes correctly, you don't 00:46:21.327 --> 00:46:24.477 get blisters, and then you're not out of the game, and then a higher point 00:46:24.477 --> 00:46:26.497 scorer s- is in the game, we win, right? 00:46:26.497 --> 00:46:29.337 Like, all of this stuff from, starts from absolute basics. 00:46:29.347 --> 00:46:30.487 So I think there's a lot of 00:46:30.686 --> 00:46:31.087 Callen Sapien CEO Synthreo: Угу 00:46:31.387 --> 00:46:32.107 Todd Kane: truth in that. 00:46:32.177 --> 00:46:36.837 Uh, and I am right behind you on, uh, you know, let's get passionate about this. 00:46:36.837 --> 00:46:40.827 Let's start building an AI practice because, as I said, your 00:46:40.827 --> 00:46:44.157 clients are doing this without you, and they're doing it poorly. 00:46:44.217 --> 00:46:45.977 So they absolutely need your help. 00:46:45.987 --> 00:46:49.247 Whether or not they've explicitly asked you or not, they need your help. 00:46:49.747 --> 00:46:50.967 So no, this has been great. 00:46:51.067 --> 00:46:53.377 Uh, really appreciate you coming on, Callan. 00:46:53.767 --> 00:46:56.947 Um, if, uh, people wanna reach out to you and know a bit more about 00:46:56.967 --> 00:46:59.217 what you guys are building in the platform, where should they find you? 00:47:00.365 --> 00:47:05.185 Callen Sapien CEO Synthreo: Uh, either Synthrio, uh, .ai or you can email 00:47:05.185 --> 00:47:10.935 me directly at callans@synthrio.ai, and I'm happy to, happy to jump on. 00:47:10.935 --> 00:47:14.985 I've got a, I've got im- I've got-- My link is public, so I love this industry, 00:47:14.985 --> 00:47:18.415 and thank you so much for having me because this is a mission for us, right? 00:47:18.415 --> 00:47:23.535 We, we-- When we were building this company, we, we saw this problem 00:47:23.535 --> 00:47:26.875 coming, and we actually thought, "What if we were to build…" 00:47:26.875 --> 00:47:30.005 Because my co-founder built a successful MSP and sold it and exited. 00:47:30.345 --> 00:47:34.415 He's like, "What if we built an AI-native MSP and just went and sold?" 00:47:34.425 --> 00:47:39.205 And w- both of us realized how much that this community has given us. 00:47:39.205 --> 00:47:39.635 I don't care. 00:47:39.635 --> 00:47:42.405 When I say community, I mean all of it, ConnectWise, Autotask, 00:47:42.405 --> 00:47:44.805 Halo, all the, the individual ones. 00:47:45.075 --> 00:47:48.675 And we, we just want to help get back to those basics. 00:47:48.675 --> 00:47:51.595 And, and so yeah, please reach out to me, even if you're not a partner or 00:47:51.815 --> 00:47:53.245 interested in being a partner right now. 00:47:53.305 --> 00:47:54.625 W-w-we want to help 00:47:55.109 --> 00:47:55.309 Todd Kane: Cool. 00:47:55.369 --> 00:47:58.769 So I'll link to, uh, everything in the show notes, but, uh, it's been awesome. 00:47:58.799 --> 00:47:59.219 Thanks, Calum 00:48:00.451 --> 00:48:01.161 Callen Sapien CEO Synthreo: Thank you very much