AI agents are quickly becoming part of everyday business operations, but many organizations are deploying them without the governance controls they need for human employees. In this episode, Brian Kelly, Founder of Automated Edge, explores the risks of giving AI broad access to business systems, data, and credentials without adequate oversight.
With more than two decades in the MSP space, Brian explains why unrestricted AI access creates security, privacy, and accountability challenges most organizations still haven't addressed. The conversation examines practical approaches to AI governance, including private AI environments, orchestration layers, permission boundaries, and skills-based access controls that keep business intelligence under organizational control.
If you're helping businesses adopt AI, this discussion offers a timely look at how to balance innovation with governance, security, and trust.
[00:00:06] Welcome to MSP 1337. I'm your host Chris Johnson, a show dedicated to cybersecurity challenges solutions, a journey together, not alone. Kevin, I'm joined this week by Brian Kelly of Automated Edge. Brian, welcome to the show. Thank you very much for having me.
[00:00:31] So, topics, because we like to use AI as often as possible in any of the topics we do for this podcast. Governance has come up. I had an episode where I had someone on that their self title is the fractional head of AI. I thought that was an interesting way to approach it.
[00:00:53] You kind of gave me some examples of some of the things that you're doing at Automated Edge. And so maybe, to get started before, we'll figure out the title of this as we go. But before we do that, if you just kind of give a little bit of background to kind of where you've been, you know, we talked before the show started, you've got the MSP roots. But maybe give some oversight of where you've been and why you landed on what you're doing today with Automated Edge.
[00:01:19] Sure. Well, yeah, as you mentioned, I most certainly do have the MSP roots. I've been in MSP for a very long time now, since about 2004 or so, something like that. And I've worked pretty much every position that you have, that you can work within an MSP, all the way from tier one, all the way up through the whole thing.
[00:01:37] But I've always leaned more towards the technical side. And my job has always evolved a lot of how are we delivering this from a technical standpoint? How are we scaling it? How are we monitoring it? How are we securing it? You know, sort of the thing that we're delivering for the clients has always been a big part of what I do and, you know, putting that together and maintaining its integrity and packaging it.
[00:02:04] So what I did, you know, towards the beginning of last year was I said, okay, I want to build a service offering and be an MSP, but specifically around AI service offerings.
[00:02:20] Like, figure out what does that look like for an MSP? You know, how do you deliver that? How do you put that deal together? How do you secure it? How do you, you know, do all this fun stuff that MSPs are supposed to do?
[00:02:34] So that's really where automated edge originated because I just saw a need for that, particularly in the SMB space of, you know, there's a lot of, you know, you might call it snake oil salesmen out there promising a lot of AI stuff, very little consideration given to security, I think in a lot of it. You know, so really jumping in on that lane is kind of where we are like trying to deliver a good solution.
[00:03:03] That's something like the other things your MSP put together for you, you know, like some plan, you know, some structure.
[00:03:10] So, so I think it's interesting, you know, a lot of MSPs right now are grappling with, in fact, there's a lot of research behind this, that MSPs are obviously primed for their, their role as the IT service provider to function as a, a catalyst to help businesses leverage things like AI to do really good things. And of course, there's the funnel factor that we see quite a bit of, Hey, just turn it on.
[00:03:38] We, we, we know we're going to be behind and we're going to save like a bazillion dollars because we can then lay people off. And of course, you know, even, even Zuckerberg has shared with like, I don't think that's our reality. And it's definitely not our current reality. And even businesses that are laying people off to solve for that FTE overhead that maybe they don't need in certain areas is turning into now having to hire back either same reason.
[00:04:04] Or a completely different FTE resources to solve the, now that I've done all these things with AI, I now need people to manage the AI. And so you mentioned that your primary focus obviously is in the SMB space. And I think that's really interesting for me. I think about the world of voice over IP going back to the early 2000s when we first saw digital and later VoIP options where the SMB can now stand up a phone system or for almost nothing.
[00:04:34] And in some cases, truly no dollars with open source solutions that are out there to basically say, you know, press one for support and two for sales and whatever other options you wanted to have billing, et cetera. But I can still be a one person shop. And that gave the, that enabled the ability for the SMB to truly have a platform to compete directly with the large mid-market enterprise companies that had, you know, auto attendance and platforms
[00:05:02] to support that. And I don't think we have really seen that since, like since the early 2000s, we have not really seen where there's been a technology that enables the one person to, you know, not very many employees required to pull off something that the end user who's consuming these services would have no clue that it's not a, um, a larger company. I mean, websites,
[00:05:28] some of those things could factor in as well to help with that idea that this is a bigger company than it might actually be. But to your point and the focus that you're, the category that you're focusing on, I argue that today more than ever, we're seeing that come again. And it comes in two, two, two categories. And I'd love to get your insights on both. The first one is the ability to really destroy the industrial revolution model. You know, if you go back in time, we had the,
[00:05:55] the CEO and then you had the C-suite or maybe the leadership team and the departments and they cascade down and you only allow for so many people to manage so many people. I think it's the rule of seven. Uh, I think there's some sort of capacity there and you get above like 12 people and management and now you're spawning a new team, et cetera. And what has happened is on, on that side to the employee side of things, it has flattened the need for as many people because you now have the ability
[00:06:25] to leverage AI to get the communication pieces to the people that need it without having as many staff people. Right. And, and that introduces obviously it's its own problems. Um, but it's, but it's an interesting concept to look at. Like now you have people managing what the agents are doing, but it's the same concept as before of like the information still has to get to those making the human in the, in the loop. And the other side, and the side that's more fun
[00:06:52] is what enables the, cause I think for the big enterprise, this is an easy model, right? This is easy. They're letting people go. They're solving it with AI. And then you go look at the SMB space and it's like, well, wait, they didn't have the teams of seven and then a new department spawned or a new, you know, hired somebody else to solve for this. Cause we've gotten too, too broad and not deep enough. SME has never been there. So they weren't hiring to solve for those things because
[00:07:17] that's not how those businesses were built. And so AI has a whole different perspective for them. So in my, in my humble opinion, mid-market enterprise is the ability to reduce staffing, to flatten an, and, uh, an overly broad and deep, uh, communication system. And on the SMB side to enable again, the competitive mailers that started with phone systems. So I just love your
[00:07:44] insights on that. And then we can shift into what the real value is, which I think is the excitement around products and services that are now possible because we have tools like AI. Sure. Sure. No, I, I definitely think that in many ways it is, it's an equalizer, uh, you know, and it's giving people the kind of bandwidth and capabilities that, you know, maybe they didn't
[00:08:10] have the resources to acquire before. Uh, and I really think that people aren't even seeing how far it's going to go yet. Like for example, uh, you know, we can circle back around to this if we want to talk any tech stuff, but sort of my go-to is a, an actual server, uh, or VM or whatever it is. And a database that lives inside of the perimeter of whatever the network is that we're already
[00:08:38] supporting. But to apply that to myself, I've already done that to myself. And now as a result, I have my own sort of front end to that and I operate almost 100% of it. Like I literally spend almost all of my day just orchestrating agents. Sometimes they exist for five minutes. Sometimes they're a permanent agent that hangs around for a while and handles my website or, you know, uh,
[00:09:03] looking at the bank account or wherever it is, but I've, it's, it's completely changed how I work. Uh, and I, but you know, going back to what you're talking about, about sort of that team structure and that sort of thing, I find that I'm still doing that pretty often, just not with people. Uh, like I, you know, I'm just sort of, I'll be like, you know what? I really need this to be taken off for by play. And now that I've already sort of fenced everything in and gotten everything
[00:09:31] connected securely. Well, now I can just take a few minutes and create an agent and prime it for whatever that job is and turn it loose. Uh, uh, you know, so that obviously in the past would have taken either. I need to learn that myself and find the time to do it, or I need to hire somebody and go through that process. And now I'm able to just action that stuff in an afternoon, you know?
[00:09:56] And, uh, so I think that really that that's a huge game changer for, for anybody that can, that could harness it. But to speak to that, I think the problem that we're being presented with right now or businesses, you know, SMBs, that kind of thing is yes, you can have these fun AI capabilities, but the cost is that you have to hook your entire business up to our stuff.
[00:10:21] Ours meaning chat, GBT, Claude, Dots, the meta, whatever. Basically, if you'll let us have complete unfettered access to all of your business intelligence, then yeah, you can use all of this stuff. And I think that that is not good. It's going to be a huge problem. And that right there is kind of, it doesn't have to be that way. We can instead make for your clients a private system
[00:10:46] that sort of operates the same way, but they control it. It's not in the hands of, you know, whoever, you know, you wouldn't do that with anything else. I mean, you wouldn't do that with literally anything else at all with your business. You know, you wouldn't, you know, hook your email up to some stuff that is just, what is it? You know, you wouldn't do that with your finances.
[00:11:11] Whereas right now with these AI platforms, we're suddenly really willing to do it with everything. And it just doesn't make sense. Let's explore that for a minute. Cause you, there's a couple of things that you touched on that I think are all extremely valuable. Uh, the, you know, I think about, you mentioned temporary agents, which is really interesting. It's like, it's like hiring somebody to clean your house. You're not hiring them to clean your house and stand in the doorway when they're done so
[00:11:38] that, you know, in case you spill, they clean that up to, uh, you know, this temp, the temporary idea of like it, it goes away. And I was thinking about how easy, just from the commercialized side of the different labels that have been put on it. And I'm more familiar with like the co-pilot play and anthropic play than I am necessarily about some of the other, uh, ones that are out there, but like the researcher and the analyst and some of those that are like by default already there.
[00:12:03] And then you have like on the anthropic side, referring to things like is, um, you know, you can do things like you create an agent and then it's like, how are you keeping it programmed? What's the, how are you pre storing memory and some other things that need to be there all the time. And what, what you said that really. Picks my interest was you kind of brought it together. Maybe not even intentionally is thinking about if I hired an executive assistant, a human one, what is the probability that I'm going to be
[00:12:31] really keen on saying, and here's access to my inbox. And I don't mean like shared access in the sense, in the sense of like, so-and-so has responded on behalf of CEO, but the CEO has responded to an email, but the email wasn't written by the CEO, right? It was written by the assistant. And what's interesting is I hear people say this all the time. I would never give that level of access
[00:12:59] to this, this new human in my equation. And yet at the same, same breath are like, look at all my efficiencies because I gave all of my credentials to, to a non-human who has waiting more power and authority from the standpoint of what these credentials potentially can do that that AI agent is going to figure out, even though my human assistant probably isn't trying to
[00:13:24] push the boundaries to go, well, if I have credentials to log into their email, do these credentials also work to log into the bank account for the company? I mean, like I would argue that in most cases, it's probably not even crossing their mind is something that they can do because, well, that would probably set off some red flags and they no longer have a job. Yeah. And then, uh, it's, you know, if an AI decides to do all that stuff, then you have to ask yourself,
[00:13:51] well, I know, you know, do I have a system that's set up to where this stuff is logged and audited and I actually know what the AI went and did, or am I just going to have to discover it later? You know, it's just the, the cavalier attitude. Uh, you know what it reminds me of is like I said, I started in MSPs like, you know, 2003, 2004, something, something right around there. And, and back then, like you could go to a major metropolitan area like Chicago or something
[00:14:21] like that. And there would be like four or five MSPs, you know, that would be about all there was. And, and, uh, also at that time, it was not typical at all to have a security department or a security person on staff necessarily as an industry. Like it had not become as, I mean, now if you try to write an industry without any consideration for security, it's like, what are you doing? And, you know, there's whole specialized, you know, things and that, but, but at that time,
[00:14:50] it was just, you know, you put the stuff in, you do your best and, you know, respond to any issues that there might be, but the sort of, uh, standardization and expectation and regulation around it wasn't really there yet back then. And I've watched that kind of develop over time until now it's, you can't go without it. And I really feel like this is going to be no different. The, the, it'll move much faster than all of that stuff did, but what we're missing right now,
[00:15:19] is sort of that best practice kind of regulation. Like if you're an IT provider and you're putting in AI stuff, you know, here are some regulations and some best practices and some things that you really should probably follow when you're doing that, you know? David Morgan It's like you do. I mean, you said this without saying this, you didn't say the big word, but like governance in 2004 was
[00:15:45] simpler times, right? Operational efficiencies, productivity efficiencies. We hadn't yet come up with the categorization of cybersecurity because we didn't have to. David Morgan So that was in the South. Yeah. David Morgan You know, and, and even if it did, I think that at the time you didn't need it to exist because it wasn't the, the exploitations at that level were, were less concerning. You also didn't have two, 300 different places that you might
[00:16:11] be logging in for different reasons, whether it was SaaS based or on-prem based. I mean, if you just think about like a switch stack in any physical environment, small businesses had one login for the switch and firewall because it was two switches and a firewall. It wasn't the same as what you saw on the mid market in an enterprise. So at that time, I think going back that far, they were on the cutting edge of having to learn and adapt to what was becoming the threat landscape
[00:16:39] that was slow then, which is now very fast. The S and B market was dealing with more of like, we're not a target. And like in many cases, it was more, you were a target by accident because of an IP address and a block of IPs that they assumed existed for another company. And as IPs became, you know, scarce as they became the, you know, the becoming extinct almost and needing IPv6 to just continue to build this out. Uh, you've started to run into is like kind of like phone numbers,
[00:17:09] right? You could cancel a phone number and discover you canceled your phone service because it was the primary to the 30 phone numbers that you had. So it's interesting that the way you sort of described that, that we are, we're in a position now that is not really an AI problem. It is an AI amplification of an existing problem that has existed 20 plus years. And, and to your point, you keep sort of
[00:17:36] hinting at this, but like to, to the degree what you're doing, why you sort of stood this up with your MSP and thinking through this is like, well, because we have to govern this no differently than we would any other technology. Yeah. It's, it's honestly, if I, you know, I feel like if you're going, first off, I really, you know, just putting it out there, you know, hooking straight up to these services, I think is a bad idea. I feel like you should have, you know, a business to have their
[00:18:05] own orchestration server that, that it hooks up to everything and then you use it to decide what to release, you know, securely rather than just having that unfettered, you know, hooked up kind of thing. Uh, but I feel like, you know, if you, with that said, when you're going to put that system in it, you should look at it through the exact same lens as if you were putting in a file server or a mail
[00:18:31] server or a Hyperbeat server or whatever, you know, you, you obviously, uh, have to give consideration to, uh, have I closed all the holes? Am I backing up? Do I have disaster recovery? Do I have permissions set properly? You know, I feel like with this, it's, it's, it is, and will not be any different. People just don't realize it yet. You know, they just don't, it's just so new that they don't realize,
[00:18:58] you know, I can imagine probably S and B's and stuff probably don't even want to realize that this is another thing that they're going to have to have a provider do. And, you know, but it's, that doesn't change the reality of it. But I think that this conversation you and I are having right now will be very normal a year from now, you know, and, and yeah, for sure. It's, it's just, you know, it's, it's, I watched it all unfold over the past 20 years with all these other technologies. And though
[00:19:28] this one is bigger and will move much faster and, and also has much more risk. I mean, because the thing is, is that now if you're exposed, for example, it's, it's likely not just some dude on the other side of it with a keyboard trying to break into your stuff, it's going to be a whole network of AI agents that are running, you know, at superhuman speed and identifying any hole they possibly can. And I'm a firm believer that you're going to want to have a secure system and your own
[00:19:56] AI on the other side of that, that can say, whoa, hold up now. Like, who are you? You know, what are you doing here? I'm not supposed to talk to you about any of this, you know? Yeah. We see the, and we see the, I don't remember who I was listening to. They were talking about the AI slop. So it does these great things for us. It does it really, really fast, but it also does it in such a way that exposes the sloppy nature that it approaches solving problems. Like, so a task
[00:20:25] like print file on desktop has now exposed 12 processes and it's consuming all of those resources. It's like, well, it's still not built well, kind of like how we've been programming all of our lives to not build well and processing power solves for the poor programming by giving us, if you have enough processing power, it still gets to an acceptable timeline to end. So with that in mind, this is where
[00:20:51] I think, you know, you, you mentioned sort of the private space and obviously the media and the commercial players today are really hyping the importance of why they, you should use commercial products like theirs, as opposed to the open source, or even if they were commercial models that are able to be self-hosted and, and, you know, configured and modeled. That's where we are today,
[00:21:17] right? Like we are in a position right now where we're at a tipping point of, is it going to be regulated and depending on the country you're in, you might not be able to able to use any models that aren't publicly regulated or commercial regulation versus how I build something that is an AI model that runs on a Mac mini or a raspberry pi or whatever it might be. Cause I mean, we can see this
[00:21:42] value already, right? The ability to run a, an LLM, you know, limited parameters, it runs on a raspberry pie and it sits in a manufacturing, uh, to help navigate the side-by-side robotic arms that are putting together cars. Uh, that would be a token consumption that would make it non-sustainable or unsustainable if you were using a commercial model. It's like got you constantly burning tokens.
[00:22:09] And as you expand, you're burning even more. You know, I'm glad you brought that up because I think that there's a piece that a lot of people miss there concerning token economics and how to keep that stuff under control. And that's another reason why I recommend kind of having your own system of orchestration. And it's a lot of what we do and that kind of stuff, because the fact of the matter
[00:22:32] is, is that not everything is a job for AI. Uh, for example, example, suppose you, you say, I need a report on this from AI and it's connected to your system. Well, you got two ways that could work. It could have all of the connections and be able to just freely run through all your systems until it finds what you're looking for. Or you've got your orchestration and database system in place to where
[00:22:58] it feels the request. And it says, you know what, this is the information the AI is going to need for this request. Let me get it for it. Uh, and now you've, you've both, you've reduced your risk footprint. You've reduced your token usage, you know, just by doing a little bit of, of pre-gaming on the front end, you know, with your own sort of system, you've, you've closed a lot of loops right there, you know?
[00:23:24] And, and right now, uh, and, and much more of a consumer kind of way, uh, like B2C kind of offering almost, uh, you know, all of these big frontier models are wanting you to just hook everything straight up to them, skip that step and just let it just come in and look around and figure out what it needs to do to answer your question. And I just, I think that's a big problem. Like that's just not
[00:23:52] very smart, you know? Uh, and, uh, you know, whether you look at it as, uh, you know, an orchestration server or call it an AI firewall, if you will, or whatever the case may be, uh, that, that's really where we're zeroing in is that when I think when it all comes out in the wash, we're going to realize why would this be even different than any other system we install as MSPs? Uh, you know, why would it not take a server? Why would it not take, you know, network security?
[00:24:21] Uh, uh, you know, on all those things. You, you brought up firewall and it's, it's, it's ironic. I think when we look at the models that we use and when we put in guardrails and harnesses and in, in my humble opinion, I think if they are inside the LLM, you probably have created your own, uh, doorway or, or easy, just kind of like using Swiss cheese to secure environment versus we still need web application
[00:24:51] firewalls. We still need hardware-based firewalls that sit outside of where it exists to ensure that when things start to get wonky, there's something else that's doing the integrity check. And, and it was, uh, it was just the other day we were talking about, you know, thinking about like the harnesses and guardrails we've been talking about is that also means that you now have endpoint detection and response or the XDR that now you have the AI XDR, right? Like where it's not just
[00:25:20] looking for malicious activity. It has to discern, is it actually malicious because it's AI and it might be doing a lot of things that we would expect from malicious, you know, bad threat actors or, or payloads that have been emptied onto the machine. When it does things like I ran this process using a PowerShell script and then I deleted it and then I went on to this next thing. And then it starts to realize it's doing things it doesn't need to do because it wasn't programmed to do it in such a tight and
[00:25:47] concise way that eliminates all that extra. Yeah. Absolutely. Well, I mean, and what you're referring to are skills, you know, like, you know, by creating skills, you're basically saying, you know, here's an area of concern and here's really what you need to do. You know, don't go looking around all over the place. Like here's kind of your instructions of where to go look for this sort of thing. And the way we do those, because you really, if you do it properly,
[00:26:16] you should have a whole Vipers nest of skills and rules and guard rails. So if you set up a whole like business and we put all of that stuff in a database, as opposed to right there in the prompt and in the, the, the thing, it basically, we have a database of all of that stuff. So when the session spins up and it says, okay, what am I working on? It runs over the database and it says, Hey,
[00:26:40] this is what dude asked me about. What skills do you have on that? And, and, and then it can feed that back to it and say, Hey, try this, this, and this. So any session never has 100% of anything. It's just at any given time, it's only given what it needs in that moment in the sandbox to do whatever it is that it was just asked to do. Uh, you just described a token creep, right? Because if
[00:27:08] you've got that in place, then it's not a running on and on and on session. That's just continuing to consume tokens. You can start a new conversation and start over, but it already has the information because it can reference that database and it can pull that stuff in without having to go out and start that ticking mile all over again. Yeah. And the other benefit of the database is, uh, you know, you can have a much bigger memory footprint, past sessions, uh, you know, it could make it to
[00:27:36] where it's much smarter by making it dumber, you know, in a way. Well, yeah, yeah. Yeah. You see what I mean? Like, uh, just only load at any given time dynamically with what's needed for the job, you know? Uh, so you had, you end up with a bunch of different agents. Like I said, some are temporary, some you keep around for more permanent jobs, but none of them have the whole keys to the kingdom. None of them on their own can really do too much, you know?
[00:28:06] So, so we, we've got a couple minutes left, a couple of things. One, I would love to hear your take on why you built out some of the things with automated edge. And then, uh, the, the last question is to think about why you're probably able to quote from memory, what you've got with automated edge, why people should check it out is, is there a book that you have read recently that you're like, everybody that listens to this show should read that book. Uh, is there a book that, uh, you know
[00:28:35] what? Honestly, I just, I haven't really had the opportunity to read too many books lately. I think I've been kind of, uh, my attention island has been ruined by all of this, this fun technology. So I swear, I think my patience is gone down. Like even if my AI agents start to talk, I act to me a little bit. I'm like, come on, man. You know, uh, Amy, you may should start reading you stories so they don't have to subscribe to the Kindle audio anymore. Well, yeah, that's, it's really been a more
[00:29:01] of a short form. Uh, uh, you know, honestly, the kind of stuff I've been into lately really wouldn't even be relevant to like, like don't get me started about, uh, like the Amazon jungle and uncontacted tribes and pyramids and stuff down there. I mean, we could hold you a whole nother podcast about that. Reading, so to speak has been a lot. It's really been more hands-on, like let's get them here figured out. Yeah, that's okay. So focusing back on the automated edge, tell us a little bit
[00:29:29] about that and then close out with where people can find you if they want to learn more about what you're up to you and automated edge. Sure. Automated edge is basically about providing that secure solution with all the fun stuff we've been talking about for the past 30 minutes. Uh, we do it for SMBs, but we also white-laid with four MSPs that, you know, have this sort of line of revenue potentially coming in, but don't really know how to service it. And it ends up taking a very familiar form.
[00:29:58] Uh, one thing we haven't discussed, we have a desktop agent that we install, uh, called Edgerton, uh, that goes to a private database for a client and it actually just watches and tells you exactly where AI would be appropriate to an AI automation. Here's where time's being wasted. Here are systems identified we should look up. So that does a lot of, I basically automated the assessment and, and typically it's, I'm trying to make it so familiar assessment. Typically there's a proposal
[00:30:27] right after that. Typically there's a project to put this stuff together. And then typically after that there's the MRR to keep it going. And I mean, how familiar does that sound? You know, like how many systems do MSPs put in that just that the lifespan goes exactly like that, you know? Uh, and really we're trying to do the exact same thing. Get in, assess what kind of AI system would actually help. Let's put a fence around everything and make it secure, make it to where you can actually start securely do this stuff.
[00:30:56] Well, I mean, what you just described, I'm not sure there's any MSP that doesn't need something like this to evaluate both client pricing and internally how they might capture, you know, better utilization of the human time. Yeah. On automated edge.ai, there are some demo assessments that, uh, as well, that come back from that agent that tell you, here you go. Like here's where people are wasting time. Here's the ROI. Here are
[00:31:22] systems that are disparate from one another. Uh, you know, and it's just an agent we put out there, let it, you know, run for a couple of weeks and, and it comes back and tells us almost everything we need to know in terms of where all the pieces on the table are. And then we can put, put together a proposal of here's how we put in a server, a database and time all together and securely, you know, and store all the keys in the band of fault, you know, and all that fun stuff, you know,
[00:31:49] and, and, and really, like I said, at that point, it really starts to take a very familiar, uh, uh, uh, form, you know, uh, like if you're installing a new server for anything, you know, or some new infrastructure. Yeah. Well, this has been great. I, I really enjoyed the insights. This is definitely a new area. I thought I had heard everything, uh, about AI, not really, but like, I've heard all of the
[00:32:15] the hard stories, right? The, you know, governance is important and harnessing and, and pacing in the frontier models. And now we have responsibilities, ITSPs, but at the end of the day, you know, we're probably not going to actually keep up with the LLMs evolution. We just can look at how can we do these things better. And so with that, I thought it'd be a great idea to just say, Hey, if you're interested in what Ryan Kelly's doing over at automated edge, go check it out, automated edge.ai.
[00:32:45] And with that, you've been listening to an episode of MSP 1337. Thanks and have a great week. Thank you very much.