Why SMBs Will Soon Spend More on AI Than Security — and What It Means for MSPs

Within two years, the average SMB will spend more on AI than on cybersecurity — and Jimmy Hatzell, Co-Founder and CEO of Hatz AI, says the MSPs that win this shift will look nothing like the per-seat help desks of the last decade.

In Episode 10 of How to MSP, host Andrew Moore sits down with Jimmy Hatzell, Co-Founder and CEO of Hatz AI, to unpack how fast AI is moving through the MSP channel and what it means for managed service provider growth, pricing, and the future of the vCIO role. Jimmy explains why the AI wave is moving roughly six times faster than the security wave did, how Model Context Protocol (MCP) servers from Halo PSA, ScalePad, Liongard, and N-able are changing what MSPs can actually deliver, and why intelligent model routing cut Hatz AI's costs by about 75 percent on common workflows. Andrew and Jimmy lay out a practical playbook — starting with shadow AI as the security conversation — and dig into how per-seat MRR breaks when clients hire fewer people and spend more on intelligence.

What you'll learn:

•       Why SMB AI spend is on pace to overtake SMB security spend inside two years

•       How to use existing DNS, RMM, and firewall data to surface shadow AI in any client environment

•       What Model Context Protocol (MCP) is, and why Halo PSA, ScalePad, Liongard, and N-able have all released official MCP servers

•       Why intelligent model routing can cut AI costs by roughly 75 percent without sacrificing output

•       How the per-seat MRR model breaks under AI — and what advisory and account-management revenue replaces it

Mentioned in this episode:

•       Behind the Cloud by Marc Benioff — Goodreads

•       The Grateful Dead on Spotify — open.spotify.com

•       Phish on Spotify — open.spotify.com

•       Hatz AI — hatz.ai

•       Halo PSA — halopsa.com

•       ScalePad — scalepad.com

•       Liongard — liongard.com

•       N-able — n-able.com

•       Cisco Umbrella — umbrella.cisco.com

•       ThreatLocker — threatlocker.com

•       ConnectWise — connectwise.com

•       Pax8 — pax8.com

•       Salesforce — salesforce.com

About the guest:

Jimmy Hatzell is the Co-Founder and CEO of Hatz AI, a secure AI platform for SMBs delivered through managed service providers. He previously led teams at SKOUT Cybersecurity through its acquisition by Barracuda and served as VP of Revenue at CyberQP.

Read the full show notes: https://www.ridgeviewadvisors.com/blog/smbs-spend-more-on-ai-than-security-msp

Key topics: MSP AI strategy, shadow AI, Model Context Protocol, MCP servers, MSP per-seat pricing, AI governance for SMB, MSP vCIO, AI model routing, AI cost management, managed service provider AI services

#MSP #ManagedServiceProvider #ShadowAI #AIGovernance #ModelContextProtocol #MSPGrowth #vCIO #SMBTech

About How to MSP

How to MSP is the weekly podcast for Managed Service Provider owners and operators who want to grow, scale, and exit on their terms. Host Andrew Moore — Founder of Ridgeview Advisors — interviews MSP operators, advisors, and investors on the topics that actually move the business: EOS implementation, valuation and exit strategy, sales and marketing, service delivery, M&A, financial benchmarks, and operational excellence.

Watch full episodes on the How to MSP YouTube channel.

How to MSP is produced by Ridgeview Advisors and hosted by Andrew Moore.

Transcript

Cold open: managing AI spend is the bigger opportunity

Andrew Moore 0:00

This is the How to MSP podcast. I'm your host, Andrew Moore.

Jimmy Hatzell 0:06

There's an initial revenue opportunity of rolling out AI tooling and access to models and whatever that is. And then there's a long-term, much larger revenue opportunity on being the go-to company that manages the intelligence in AI spend for an organization.

Welcome to How to MSP with Jimmy Hatzell

Jimmy Hatzell 0:25

Make sure they're spending in the right places, doing the right things, uh, and optimizing the right processes or re-architecting processes entirely. So the latter is the big opportunity, and not everyone sees that right now.

Andrew Moore 0:38

All right. Welcome to the How to MSP podcast. I am your host, Andrew Moore, and today we have a very special guest. This is Jimmy Hatzell with Hatz AI. And I'm excited to have him on the show. Thanks for joining me today. Sir, how are you?

Jimmy Hatzell 0:53

I'm doing great. Thanks for having me, man. This is this is awesome. Very excited to be on.

Andrew Moore 0:58

Nice. Uh so where does the podcast find you today? Where are you physically at? So everyone knows where you're where you're at.

Jimmy Hatzell 1:04

Live from New York. No, we're actually not live, but it's not Saturday. It's it's at the time of this recording, it's Monday morning at like 9 a.m. So I don't know if you tell people that, but that's where we are. It is. No, it's fair. And and the and the funny thing is, is you've become such a busy person. They were like, we can fit you in Monday morning. And I was like, when? And they were like nine o'clock Eastern. I was like, oh, okay. So sounds great. I'll get up and and I'll do a podcast at 8 o'clock central. That sounds fantastic. Thank you for doing that, man. I'm gonna be honest, I looked at it too and I was like, why do we schedule this? Okay. You know what? Let's just scale off. I'm excited to talk to Ancha. Screw it. Just screw it. Um, so no, is it I'm I'm glad that you were you're on. So why don't you tell everybody a little bit about you and your journey and what you're doing today? Just kind of give us a little bit of background on um why the why the the the channel should should care about your opinions

Jimmy's path from cybersecurity to founding Hatz AI

Jimmy Hatzell 2:00

and and and where you came from to get here. Yeah, sure. So again, Jimmy Hatzell, CEO, co-founder of Hatz AI. I've been working in the MSP channel for the majority of my career, the background in cybersecurity, um, worked in IT, um, you know, and I've worked at a couple of cyber companies um in various roles. And a couple of years ago, I was looking at sort of where things were going with with AI, and you know, like there's really three things that I was pretty confident and bullish on back then, and this is like 2023. Um, the first is that there's gonna be more demand for AI tomorrow than there is today, and that'll be true for a very long time. The second is uh people will need more help with AI tomorrow than they do today, and that'll be true for a very long time. And the third is that MSPs could bridge that gap and monetize that, create value, et cetera. And so we set Hatz AI up to enable MSPs, put them in that business, you know, help them distribute, manage, sell everything around AI, which is a pretty open-ended business to start. And Andrew, you've been uh, you know, you've you've been a uh uh uh a confidant to me over the years, especially early on. I definitely called you with many ideas, and you can see how things have drifted over time. But it this the easiest way to think about it is is we think of how small businesses will use AI, what's valuable to them, and how can we productize that so MSPs can manage and sell that and provide value and services on top of that. So we have an open-ended business model. What that looks like today, we can get into. Um but that's what we do. And uh I guess that's why you should listen. I don't know, maybe turn it off. No, no, no, I love that. And it's great because me personally, and I remember when we first started having these conversations years ago about what you were doing, I almost had a very myopic view of like, this should be good for the MSP. Like, this is a product MSP should be buying. And I think you rightly saw that there was an opportunity to sell through to the MSP customers, which at the time I hadn't even gone that far. I was like, well, this is just something that we need to be doing to automate things within our own MSP. I almost saw it like a like a roost or a PIA type solution. I was like, oh, okay, so we're gonna use AI to do this, and you know, Jimmy and his team are gonna build something for that. And I was at first I was like, wait, how are you gonna sell this MSPs? And then I started to see what you guys were doing with it. I was like, oh, okay, like that's that's wise. Like you guys made a a really good. I don't know that I'm gonna call it a pivot. I don't know how how it landed in your roadmap, but I think it was a really interesting place for you guys to go, and obviously very prescient that you were able to get that product out uh when you did. I I think it it it really has borne um a change in in the industry as a whole. People are talking about it, which is great. It's where you want to be personally. Yeah, you know, it's it's it's been fun. And and to be honest, we've been a bit ahead of where things are for the most part, and in some cases behind, right? Like so there's this duality to it where we see where things are going and we make bets on that. Um, but like early on, right? Like I had a small engineering team. Now we have a huge engineering team, right? And we're able to ship quickly and keep up. And you know, new feature would come out on like ChatGPT or something, and the next day people would be like, Well, why don't you guys have that? And it's like, we're working on it, we're working on it. You know, now those cycles are a lot smaller, and and you know, we can like uh the new features have sort of like uh there's not as many now, I guess. Uh, or people are still trying to um intake, you know, how to use AI um in its uh maximum capabilities where it was at like six, nine months ago. Uh so we're we're always thinking of the future, we're always building ahead. Um, but we're meeting people where

Why AI is moving six times faster than the security wave

Jimmy Hatzell 6:00

they're at too. Yeah. Well, and and so when I'm what I've I've been to a couple of conferences um just in the last six months. And um it what people are talking about today is at the MSP conference in the channel, they are not what they were talking about 12 months ago, right? So 12 months ago, you would go to a conference and there was people were saying, hey, maybe we should be thinking about AI, and then they were talking about security and you know, all the you know, governance and you know, automation and all the other stuff. And then it's, you know, this year it's been, hey, let's talk about the shift to AI and how it how it matters to the channel, how it matters to how we sell to our clients, what are our clients asking us? So let's talk a little bit about how fast this is happening and why it's happening so fast. Like what is happening outside of what you're seeing with your company and the models with like Anthropic and what we're seeing with OpenAI and with Google and all of them as they're building these models. Why is this happening so fast? And why is this different than anything that's happened in the past, like say virtualization or cloud compute or whatever that looks like? Like, what are you seeing? Yeah, I mean, like I have a very um unique view on this because I was there for security, right? Like I was meeting with MSPs 2016, like, hey, we should, you know, we're gonna offer SOC services, you know. And people are like, well, what do you mean? We have antivirus, we have, you know, webroot installed or whatever. Like, uh, you know, it's like, no, you need more than that. You know, you need to offer security as a service and sort of all of that, and watch that happen over like, say, a five, six year period of maturity. I've seen this go through that same period of maturity in a matter of maybe a year, right? So it's like six X faster. So like there's two sort of uh macro things that are happening. Um the cost of intelligence is going to zero. So what does that mean? Well, you know, you get these, you've got an iPhone, you get like the summary, the AI summary, right, on your phone, right? Where it's just a local model sort of doing that. It's like okay, right? It's like, hey, your group text, your family's going off, and like they're trying to plan Easter or whatever, you know. And then it'll like summarize it and tell you what's going on in your chat. Yeah, yeah, yeah. Yeah, but then the cost of the the top most frontier, highest capable intelligence is up and to the right through the roof, right? Where these models are are getting more expensive and more capable at the same time. So you have this weird duality where, like, you know, I need to go and and um analyze my email gets cheaper and cheaper and cheaper, right? As more open source models come out, as uh lower parameter models that are a lot more capable uh are released uh at a lower cost. But then the most expensive models uh are just getting the the most capable ones that can run for sometimes like an hour or two hours, uh sort of unmonitored, uh, are just crazy expensive. So what does that mean? Well, it means that we can now rethink a lot of business processes and rethink the way that we do them when you have cost of intelligence going, you know, down, approaching zero, right? So you can have some imagination. We didn't have this magic box that you could ask almost any question to uh a couple years ago and it answers pretty quickly.

Shadow AI: SMBs are already spending on unmanaged ChatGPT

Jimmy Hatzell 9:30

Uh but then at the same time, if people are using Frontier highest capable intelligence, the costs are gonna go crazy. So, like I'm telling you right now, we're a year maybe two out from the majority of SMBs spending more money on AI than security. So let's let's put a pin in that for a quick second. Because if it's happening that fast, though, when you're talking about people spending money, like is SMB spending money on it, do you mean that it's already injected into, you know, like let's just say I'm I'm making this up, like the version of QuickSbooks they're using, there's QuickBooks and then there's QuickBooks with AI, right? So they're spending a little bit extra on that. Are you talking about that kind of stuff? Are you talking about they're using some model within their business exclusively? They're using like a containerized version of Cloud, or they're using some sort of enterprise version of Copilot. Like what does that mean to you? Like when you say that they're spending money on this, what like what are they spending money on? So they're not using an enterprise version of anything. They're using personal Chat GPT accounts that are unmonitored, not issued by IT, and people are doing it. And everyone, not everyone, I shouldn't say everyone. That's not that's not realistic. But many people are just sticking their head in the sand, like, oh, well, you know, I don't think they're ready to use AI. And it's like, no, like people are using it. They're they're just doing it on their personal phone. Like, if everyone started using their Gmail to email everyone, their personal account, like everybody'd be freaking out, right? But then and they'd be like, we need to get, you know, approved, blah, blah, blah, blah, blah. Like, you know, and and yeah. And so, so like it's it's changing very quickly. So, so you're talking about like the systems of record, right? The, the QuickBooks of the world, your your PSAs of the world, right? Uh, your CRMs of the world. They're all getting AI added features. And yes, like people are spending more on that. People are doing the upgrades to get the extra license to get access to it, all that kind of stuff. But these systems of record are all becoming more and more queryable. So if you look at like where Salesforce is going, they're going with this Agent 360, where they're sort of like, hey, in the future, maybe we won't have a UI. We'll just have like the best connectors ever to people's data. That is going to trickle down through uh a lot of these systems of record. I'm sure lots of people will uh create great add-on AI features and they'll be useful. You know, in my business, right, we use Slack, we use Notion, they have AI features. Um, but where it gets really useful is when you can pull them in in an enterprise context management sort of uh situation into your AI agent of choice. So in in my case, right, I use I use Hatz, right? But it might be Claude, it might be GPT, you know, whatever. Um, but these connectors are getting better and better. And even in the past, like, like so

MCP servers explained: Halo PSA, ScalePad, Liongard, N-able

Jimmy Hatzell 12:30

just to break things down for you, like in IT Nation in November, people, vendors kept coming up to me and saying, Hey, how can we build an integration with you? Can you set up a meeting with my team? Can you blah, blah, blah, blah, blah, blah, blah? And every single one I said, no. You need to build an HTTP compatible MCP server, and I can connect it. And then they're like, Well, what's that? You know? I was like, break that down for the audience. Like, explain, explain yourself. Yeah, right. So that that is like it's like an API, but it's a it's a standard protocol. It's instructions for how an AI can safely and securely interact with your data and your system, where you can just say, hey, I have this MCP server, it has authentication, users can log in, they can use an API key, whatever. Any AI system can connect to it, and we've defined a set of security controls, guardrails, et cetera, on that. So like that has been moving, and we made a big bet on that as a company over a year ago, really. Uh, and it's been through various iterations, right? Like the first versions of it, you had to run on local machines, and like that's not scalable, right? That's like not secure, right? Like injecting environment variable files, and and um so but this protocol has uh matured over time. And um, you know, fast forward four months later, you're seeing, right? Halo PSA has released an official MCP server, um, Scalepad has released an official MCP server, LionGuard has released an official MCP server, uh, Enable has a release official MCP server. And then there's dozens of third parties that all they do is take sort of these APIs for these uh vendors and then you know remake them into MCP uh tools, which is which is great, right? So this like sort of like connecting your AI to the data problem is a short-term problem that's getting work through. And you're seeing the same thing with SMB tools, but they actually, in many cases, are moving faster, right? Because uh a lot of the vendors that we know and love and rely on in MSP world have made lots of acquisitions, have very complicated API services, weren't necessarily set up to integrate with AI. So it's it takes a little bit longer and it's a little, you know, harder uh in many cases to get a PSA with you know a thousand API endpoints connected to AI versus like you know, Monday.com or like something that an SMB might use where it's like, hey, we're you know, we have all this stuff and on it. Um so like you're seeing that acceleration where we internally did 30 integrations in 31 days last August. And um, you know, that was a big push for us. It took a lot of engineering power. A month ago in one week, we added 26 integrations in one week. Now we didn't drip them all out immediately. We put a lot of them feature flagged and reached out to users to test them before we released them GA and we've been putting them out. But like we'll we'll be adding sort of 100, 200 connections, data connections a month um in like, you know, a couple of months from now or in a given month. Sort of, I guess what I'm getting at is that like connection to your third party data in a secure and relevant way is a short-term problem. And um when that happens, people start, you know, the possibilities start multiplying and um you can really, you know, like my like our business lives on on um Notion and and um you know, in the code base and in linear. You know, my AI talks to linear and our code base and notion all the time, right? Like that's where I point it. And I say, go update this, go manage this, blah, blah, blah, blah. So I think um when I say that people will be spending more on AI than security in the near future, I do mean in some cases, like these add-on solutions for um products, but uh in many cases, like pulling those uh products and data into the core um sort of interface that they interact with, uh, their AI or where their agents live, I guess I should say. No, it's that's it's incre incredibly helpful to go kind of deep dive into some of that because um I think it it it's there's new language that has like new vernacular that has to be brought out into the channel about what we're doing because it is a new technology. Um But I think that part of what I'm seeing is there's almost like this disconnect in the fact that this technology is all new and it's got all these different parameters and things that it needs to work with. And and what people are seeing as it's practically applied because of what you said, how fast it's coming out, but how familiar it feels in the way that you interact with it, because it uses like a like a common voice in a lot of ways, right? Like what people are used to are these chat functions where you can have this conversation with it so it doesn't feel um as scary,

Why no single AI model wins, and why routing matters

Jimmy Hatzell 18:00

right? As, hey, I need to take all of my data and move it into this cloud infrastructure. And you're like, oh, what is Azure? Like, how do my, you know, what is containerization? Like, how do I put my data here versus like, oh, I have a storage folder on this server locally in my office, right? Which I understand because I see it. Um so I think it's really helpful to start laying that out for folks because there's a lot that goes into being an MSP and having to know about all this stuff so that you can have a meaningful conversation, not only with your clients, but at some point your own internal team, because we're gonna have to support this as MSPs, right? Like we're gonna actually like have to answer questions about what an NC MCP server is and how do you connect to data and why you can't connect that to that yet, but like maybe one day it's coming, or how do we work around that? Um so I think it's super helpful. Um so I do want to jump into um, you know, the different types. You talked about the different models. You talked about your model, you talked about, you know, you know, there's Gemini, there's uh, you know, Claude, there's so um you know, what are you making suggestions? What are you suggesting to folks outside of um, you know, use use Hatz, right? Uh, you know, we're we're we're doing what we're doing. But when a company is trying to, an MSP is trying to have a conversation with their client about what models to use, where are you guys at with that? Like because it it felt like, you know, OpenAI's model with ChatGPT was ChatGPT was the de facto for a while. Now it feels like Anthropics Claude is leapfrogged that. You've got Gemini kind of moving up and down the ladder a little bit here and there. You've got OpenClaw, if you don't want to have something on the open internet, like you've got all these different options. Like, what are you telling people about like these models? How do you how do you help them understand what they are and why they should use a particular one or multiple ones? So my favorite model changes all the time. Right? Right. Claude, OpenAI, Gemini. It's usually a rotation through those three. And also for different things, right? For image generation, I love nano banana, right, from from Gemini, right? And then apparently GPT Image Gen 2, which just came out, is pretty good. I haven't even used it yet. Like I don't I haven't used it either, but I use Gemini's nano banana and it's sick, like the stuff it can do. Yeah. So we what's happening though is OpenAI at the time that we're having this conversation, open AI has doubled their price for um models twice in a month. So when you go from GPT 5.2, they skip 5.3. 5.3 was only a coding model, 5.4 and 5.5, the prices ticked up quickly. You look at Claude, um, Sonnet, and then Opus and then Mythos, right, which is the unreleased as we're having this conversation model, right? The pricing is going up and up and up and up, but also these models, right, where they used to be able to like, you know, maybe call a tool tool or two and do a search or something like that. Now they're running um you know, for very long periods of time using lots and lots of lots of tools. So the and and everything's starting to get clamped down, right? Like so like the prosumer market where people were buying personal licenses and using them for for um for uh business use, um they the usage on on these from the model providers directly was very generous, right? Even like hit rate limits and that kind of stuff. That's that's sort of ending. Um enterprise agreements for the most part don't have unlimited usage. They're they usually require paying API prices, like all that kind of stuff. So what we found is it's become very, very important to route tasks to different models. And the age of uh I use the same model for everything uh has sort of ended, unless you're just fine just paying lots of money because things get expensive. Like internally, my engineers spend around $1,500 a month per person on AI per engineer. Right. And we we get the cheapest prices because we're direct to these model providers and all that. Like we've always been like we're an AI company, we're doing lots of engineering tasks. Like, of course, we use and spend a lot of money on AI, but sort of just to give you an idea of where things are going, so we decided in January of this year that it was extremely important for our company to be able to focus on control and optimization of models. So it we we actually have model routers that are newly released and very, very important where we take um open source models, best of breed models, um uh fast models, image models, all of that, and we combine them into a single sort of box where you're you you just hit this type of router. And we have um right now we have uh like light sort of uh credit saver mode, right? Uh performance, which is just like every day, and turbo, which is like, you know, I need the highest intelligence, I don't care what it costs. Right. Um and so by doing model routing, we have been able to um reduce costs uh in many cases by about 75%. Um

How intelligent model routing cut costs by 75 percent

Jimmy Hatzell 23:30

because we found that in SMB, a lot of users of our platform, the complexity required for many tasks, you don't need the best debris models for. Like, yes, it's required for certain decisions or certain parts of a conversation or certain things. But for the most part, if you can route to models that are more efficient cost-wise, um, but can perform the same at a given task, like, hey, check my emails and tell me what's going on. Like, you can give me the top 10 models, you're gonna get nearly identical results, right? Um then we've also looked at the the cadence of a request, right? A request, I'm calling it a request, but the technical term would be an agentic loop. So you you type something into Hatz, and our harness says, okay, we're gonna go reach out to the AI model, and the AI model is gonna tell us to go use some tools. Hatz is gonna go use all those tools, they're gonna give that back to the AI model, we're gonna guide the AI model to do this other thing, we're gonna compress this, we're gonna manage that, blah, blah, blah, blah. So there's like a hundred things happening under the hood on like your one chat conversation many times. So we found out if we can take parts of that conversation or that agentic loop that are high costs, right? Because they use lots of tools, but but um do not require high capabilities, we can dramatically reduce costs. So uh for example, uh, when you use uh Hatz, if you're doing web research, uh a web research agent, a subagent will kick off. So let's say you're using Claude Opus or something like that. It'll actually kick off a um uh research subagent that'll do all of the searching, all of that kind of stuff, gathering the information and summarize that back up to OPUS with a haiku, which is like a hundred times cheaper, right? So, like in that case, we can reduce the cost of that query by 80%. So um I get like how we got onto this conversation is like, where do you see things going? What models should you suggest? Now, where we're moving to over the next few months is around intelligence routing. So it's really looking at a problem and saying, okay, how much intelligence do you need for this? Can we match it to um, you know, something that we already have? Um, and how can we do this for the most efficient sort of price or cost associated with that? And that's exactly where we're going and the direction where we're going and how we sort of like moved

The MSP support problem: aggregating models for clients

Jimmy Hatzell 26:00

uh the company. So um even like we're working on uh releasing our own models that are very good at picking tools and and uh routing models and things like that. So um that's where I see things going. Well, and so like because that's been a thing where I've been talking to folks out in the channel and one of the things that they've been I've been talking to folks out in the channel, and one of the things that they've been asking is how do we pick something that we feel comfortable that we can support for our clients, right? So when we go out and our we start to have conversations with our clients, they're like, great, my clients are like, I I really would love to get rid of shadow IT in my environment. Um, I would love to, you know, secure my data, make sure that it's being you know used in an appropriate way, that you know, somebody in in um, you know, one part of my company can't see my accounting data, right? All that sort of stuff. So then they've been asking, like, so which what do I choose? Do I start saying that we're gonna be a, you know, you know, an anthropic provider? Am I gonna and I'm like, as like I sit back and I'm like, that's actually a really good question that I don't have an answer to. So when I talk to Jimmy, I will ask. Right. So, like, how do like because when we we've had these conversations with MSPs, and the common conversation that we have, and I'm an operations person, is okay, that's great. But what happens when my service desk gets a ticket for that? And if every single one of my clients is using some sort of a different version of an AI tool, how do I, how do I support that? And then the immediate thought is from an old school mentality, which is, well, we need to pick up a a specific vendor that we're gonna work with, such as you know, we did with firewalls like Meraki or Fortinet or SonicWall or whatever. So what are you telling people? What's the what's the way that that MSPs can create some sort of standardization in what they support and how they deploy these models into their client environment so they don't get stuck not being able to support this or they're having to support everything, which makes it untenable and difficult? Yeah, I mean, look, that's that's what we set our business up on to be a curator of really everything, an aggregator. So, you know, we we don't standardize on a single platform. Um, we have access to all the models. We get access on the day they're released in most cases. Um, obviously I would tell them to partner with me because that's what I do. Um that's that's what I want to know because we've had that conversation. I'm like, I was like, my understanding is that Jimmy's team and his products aggregate these models together and give you a single interface in order to help you create, you know, agentic workflows and things like that. So like, why don't we talk a little bit about how that works within your product and like why you think that's the winning solution versus like I'm gonna be a cloud, you know, you know, a cloud partner or I'm gonna be a you know an open AI partner. Yeah. I mean, go ahead. You know, like I'm I'm all for people using AI in general. And and um like I have seen people who like their client uh signed a year-long contract with uh open AI um teams, and now they're like, oh, but we want to use cloud because everybody's using that now, right? So it like the the sort of like constant jumping of who's the best is like makes this very difficult and also make it very confusing for clients because they'll ask one thing on Monday and then on Tuesday they'll be like, oh, I want this other thing. Um so like I think like certainly access is a big piece of it, right? Like, how do I in my organization get approved access to cloud models, to open AI models, to Gemini models, to open source models, right? How do I do that? How do I ensure that they're hosted in the US or wherever they need to be, that they're logged for compliance, that um there's a zero data retention with the inference providers, right? Like, so that's like one piece of it. The next is like control about through an organization. So um how do you control users and usage? So, for example, you could use Claude's um lowest price models for a month all day for every conversation, and you won't hit their sort of limits or spend craps in our platform or in theirs. You switch to their most expensive model, you'll hit it very quickly, right? Like sort of like running through it. And these models are being released at mid-month, right? In the middle of like it's not like there's a procurement cycle. It's not like, hey, we're gonna evaluate this and then turn it on. It's like this is available today to everyone. And it's the the pricing fundamentals of it have completely shifted, where now it's actually right, 50 times more expensive. So, how do you manage that spend and governance and what like do you want to lock users out when they hit limits? Do you need to set different limits for different roles? Do you need different dials for flexibilities? Right. So, like all of those sort of like unsexy problems are where we really shine in the direction we're going. Yes, we also help with the management of everything too. So, like when you set up your hatch, you're really setting up like your AI storefront, like for to really, you know, for lack of a more mature description, right? What models do we want to have available? Right? How do we uh what plans do we make available? What are our different user roles? What are they allowed to do? What agents have we already built that we're gonna uh push into our client tenants? What workflows, et cetera? Uh so like there's a lot of that. And if you're an MSP who specializes in like a couple different verticals, being able to create agents or workflows that work with those specific tools or that industry really well and push them into different clients, right? Like we really shine with that. But over the next quarter or half, you know, six months, where we're really focusing is from the management control, right? What if you need audit logs and you need to ship them to uh a Microsoft peer review per peer review or something like that? Or um you need to show that you can check in on uh uh hedge fund traders to make sure that you know they can back up trades or whatever. Like, what about that? What it what about um access to tools? And you know, they this this department should have access to these tools, this department should have access to these tools. Um what about sharing credentials, right? Um uh some integrations require sort of service roles, like ConnectWise is a great example of this, right? They their uh API is is designed in more of a service role, not like a user role persona. So you need to create that user and share it among uh different people. How do you do that? So like we haven't solved all of these problems yet, and like like we we're in various stages of them,

Where to start: security as the on-ramp to the AI conversation

Jimmy Hatzell 33:00

but that's like the the type of thing that we're working on. Uh and even going as far to setting up account teams. So, how do we notify the VCIO uh or the account manager for Acme Inc. Um, to go reach out to the customer because this customer just sent up a bunch of agentic workflows on schedules that are going to cost a lot of money and they might not even know that they just did that. Right. And and putting MSPs in that um situation where they can be the thought leader, where they can be proactive, not just drive people to the most expensive models and try to get them to spend more every month, right? Like we're we're trying to guide people through this journey, right? Yeah. Well, and it's and it's changing every day. And I think the the biggest thing that I'm hearing from MSPs is where do I even start to have these conversations? Like to your point, there's a lot of people that are burying their head. Um, and I don't think it's because they're Luddites or anything. I think they're really focused on, hey, like I don't even I don't want to look stupid. Like, so it's better to not have the conversation than walk in front of the client and have them ask me a bunch of questions I can't answer. And so I this is an approach I've been taking with folks that I've been coaching and working with. And I want to hear if you think this is the right approach. I said, listen, before you start landing on, like, you definitely should be figuring out what models are out there, what works for you, what you think you want to support. I've told them I was like, go look at, you know, Hatz, like there's there's places in the channel that you should be looking. Talk to Pax A, talk to whoever, right? Like figure it out. I was like, however, I was like, they're already there, this stuff is in their environment, right? They're people are already using it, data is probably already leaking out. I said, why don't we just start with what you're already good at, at least you should be, which is to and from your background, which is let's start with security, right? Let's just start, let's have a conversation and it and you can sit down in front of your client and say, listen, I can scan your environment. I because I have RMM tools, so I can see what's installed in the computers. A lot of the times I have access to your firewall logs or I've got a DNS filter on your on your laptops. I can tell you if people are going to these different, you know, chat GPT and you know clawed sites. I can tell you what's going on there, right? And then we can talk about maybe an electronic use policy. Like, like, do you have that written down that says, you know, you guys as employees are not allowed to go do this? And then just start setting some guidelines, right? Like just start having that conversation so you sound like you know what you're talking about while you're in the background trying to figure all the other stuff out. Because if you can start having those conversations, you'll start getting input from your clients about what they're actually doing with AI and what they want to do with it. And I think that'll help you start to get information in order to start making better decisions about how you want to support it with your user base. Um, do you think that's a good approach or do you go about it a different way when it comes to how you're teaching MSPs to engage with their clients? Yeah. I mean, look, MSPs are very comfortable with the security conversation. You can go in your Cisco umbrella, you can go in your DNS filter, you can go in your your uh threat locker, whatever it is, and you can pull these reports of what people are actually using and doing today and have a conversation about shadow AI. Um, we see partners doing this regularly. We see partners incorporating this into their QBR process or the equivalent of a QBR process, whatever they're doing. Um, and then you know, moving that into we need to get something governed and uh an approved tool in place and in AI policy, and we need to have a conversation on what data we're allowed to and not allowed to put in AI. I don't think I think that that is a great place to be right now. I think in the medium to longer term, uh being able to have a conversation from a business value perspective as AI, um, like what's happening is people are hiring slower because of AI's capabilities. So the budgets for AI are sort of changing because it's like, well, if we can do a lot of this with AI, maybe we only need to hire five people instead of seven people, right? Which sort of like changes the the dynamics of things. So if you're an MSP who can help um manage that and roll that out and have that business conversation really work through the process, there's a lot of money to be made. This isn't just tool spend that's happening. Um so like I think that um over the longer, like a message that works now is crawl walk run. Like, let's just get you started using AI in your business as an approved tool. Um, let's get it, let's get some easy win use cases that people can do, and let's push this pie in the sky, sort of like an agent manages everything out a couple months until everyone's familiar, trained up, and using AI. Um, and that is a message that resonates, it works well, and it's practical because a lot of businesses will either want to do nothing or they'll have one person in their organization who wants to jump to like, we need to do blah, blah, blah, blah. It's the equivalent of someone saying, Hey, I need security, I need a pen test. And you're like, you're like, dude, let's get some endpoint protection in place and like, you know, get some email security and like update your devices and enable MFA and like then maybe we could like do an audit and then maybe we could do a pen test. Like, I can, you know. Um, so like, yes, I think security is the perfect place to start. Um, but quickly moving into a business conversation will unlock a lot of

How AI breaks per-seat MRR and what replaces it

Jimmy Hatzell 38:30

uh opportunity to create value for your customers, is you know, dollars. Right. And so there's two things I do want to touch on real quick. One is I did not even think about until you just said it. It was kind of like an epiphany. Um so we talked about how people are slower to hire, right? And that's true. And you look at an example that you just provided is if I could, you know, two, let's just say 2x the productivity of one of my developers by spending $1,500 a month, like that seems like a no-brainer to me, right? Instead of hiring two developers, I just spend $1,500 more, whatever that is, 10%, 15%, 20%, like whatever your your costs are on a developer, right? As an example. From the MSP model, uh-huh, over the last 10 years, everybody has shifted to per user pricing. So if we're not hiring as many people as a country, right, and we're supporting these small businesses, obviously we're gonna have to figure out a way that if we're used to getting $150, $200, $300 per person per month as an MSP as we're selling, what uh where does that take us, right? Because now we're gonna start seeing like, like, are we good, are we creating new packages? Are we having to do add-ons with this product? Like, are we gonna completely fundamentally shift the way that we bill for our services? Like, what are you hearing in the channel? Like, what's going on, or if people people haven't come to that realization yet? Where are we at there? Have you thought about that? Oh, no, it it breaks a lot of things. It breaks a lot of things. And a lot of people's reaction to it breaking things is like, okay, well, then it's not gonna work and like let's move on. Which is like like I that's like that's like oh, they're like, what do you mean I can't bill by the hour? Like, that's stupid. Let's just not go there. Like you know, like that it changed once, it could change again, right? Yeah, no, I look, it it it if you believe that people are going to hire less people in their company and SMBs are gonna stay leaner or scale with less people than they would have in an alternate reality where there's no AI, and also spend more money on AI in the near future, then you're right. It is a contradicting thought to I'm gonna set my entire business up of bill per user per seat, right? So how can like the where where is that revenue made up? Where's the opportunity? And the opportunity is in helping, managing uh and selling that AI. So like like there's like approaching the situation of, oh, I don't want to do that because if my if my uh customers are more efficient with AI, they're gonna hire less people and then you know I'm gonna make less money is a short-term, you know, is is a short-sighted way to think about it. Um, so like, you know, it's about getting started right now. Like I I the the the the the answer no matter what is like we need to build the internal muscles to support this, uh, have a skew for this, offer services for this, etc. Um But you're right, it it it like if those things are true, it does break some long-fought MSP traditions, which is almost going backwards, right? Because MSP went from like break fix, you know, sort of like, oh, let's move into like advisory services. We sell business process, we sell solutions, to oh, standardize everything, get your lowest operating cost possible with a standardized stack per seat and be able to scale to almost moving back to like business process, like let's go through this, right? And that's where people with big uh VCIO practices um uh you know are really shining with this stuff or advisory services. Uh a lot of MSPs, uh, like who are our partners, they're getting a lot of their AI revenue and deal sourcing through their advisory services because it is an opportunity. So there's there's a, there's, there's, there's an initial

Why strong account managers are the new MSP leverage point

Jimmy Hatzell 42:30

revenue opportunity of rolling out AI tooling and access to models and whatever that is. And then there's a long-term, much larger revenue opportunity on being the go-to company that manages the intelligence and AI spend for an organization, make sure they're spending in the right places, doing the right things, uh, and optimizing the right processes or re-architecting processes entirely. So the latter is is the big opportunity, and not everyone sees that right now. No, and I think that that's the shift that I'm gonna that I'm gonna wind up having to really focus with my my coaching clients on, which is like my opinion has always been the success that I've seen in MSPs where I've worked or managed or or built, or the ones that I've coached with, they have strong account management teams. Like, and whatever you want to call them, like it's not typically a sales-based account manager, right? Where it's more of that consultative technical account manager, VCIO, whatever you want to call it, right? Where you move from five, seven, $10 million, where you're in that range, that no man's land, if you can shift responsibilities off the owner or the primary operator to smart folks that can have intelligent conversations with your clients about their business, right? And and use the technology to help them, those MSPs scale faster, they scale more profitably, right? They're in a better spot. And so I'm finding that if those folks aren't in a position to have conversations with their clients about AI, start talking about processes, like how that works, what those business outcomes are. Like then there this whole like, hey, let me show up and like try to sell you something and hey, look, I'm looking at the stack and you don't have web filtering in place, so I'm gonna go ahead and you know, you know, approach this white space conversation differently. Clients right now are like, looks, I I appreciate the fact that you think that I need some extra security, but you know, I've got three people over here that I notice they're using different websites to upload my personal or my business data in to you know, start generating PowerPoints and stuff. First of all, I'm not happy that they're doing that, but also I'm not mad at the fact that these PowerPoint presentations and these um you know proposals that we're putting out are happening ten times faster and they look twice as professional as anything we've ever distributed. So I think that the account, like having trained account managers and outcomes, like having that middle layer of consultative um employee within the MSP is gonna be absolutely critical to the ones that are gonna survive this, because the ones that are just selling commodity technology services, they're gonna really struggle. Like I don't see how they get past this over the next couple of years. But I don't know if maybe I'm just being, I don't know, hyperbolic about the whole thing and just like, oh, this is like a terrible situation for MSPs who aren't ready for it. But I I don't know, man. Like I I don't know what you're seeing, but I think it's I think it's scary for MSPs that don't have a good account manager at this point. Yeah, I mean, look, like you know more about the MSP bus. I see a lot, right? And I see adoption trends and I have an unfair view because we have a thousand Martiners that I get to, you know, work with and and see. But you're in it every day. Like you know much more about how this will play out with, you know, having conversations with the customers and all of that. But, you know, I I tend to agree with you. Like, um, we're going through a huge shift in business as a whole, not just the MSP industry, but every single industry because of AI. And things are gonna change fast. And if you're not ready or nimble enough to move, like you could seriously get hurt. And the core of your business could sort of like change seemingly overnight if you're not ready for it. If you're been in it and training and already a year into this journey, you're like, oh, that wasn't overnight. Like, I've been watching that happen, right? But but what it'll look like is you know, renewals start getting harder, right? Uh people start leaving, uh, employees starts going, start going down, employee accounts start going down, growth slows because you can't hire a new customer or can't find new customers. That's sort of uh um

What freaks Jimmy out: AI and the human element

Jimmy Hatzell 46:30

those signs sort of like happening. So, real quick before we get into the the questions here at the end, I do want to ask you a couple of real sm uh really quick things. So what what really I think we've talked about what excites you about AI, but I want you to know, I want to know what freaks you out a little bit. Like what is what is on the horizon where I mean, is it lack of investment in data centers where you've got communities that are kind of pushing back against this sort of thing now? Is it you know limitation of the ability of of us to get and retain chips so that we can actually build these systems? Is it is it this shit's gonna get r really smart and we we're like gonna wind up having overlords? Like what what are what what concerns you about where AI is going a little bit, just that you're keeping it your pulse on it, you're watching it like presciently. Yeah, I mean the constraints are the constraints and it just affects price. Like it's it's fundamental economics. There's not enough chips, there's enough data centers, things are gonna get more expensive. Like that's that. Um what really concerns me, like I can deal with that, right? Like it it sucks, but like I can deal with it and manage it. Uh what really concerns me is the human element. So I've seen it for a while because we've been so on the bleeding edge and been working with software engineers, um, you know, where AI has primarily been focused on on innovating. Where when someone who has worked very hard for a very long time develop a certain set of skills that used to take years to acquire, and then someone with much less experience aided with AI can get to the same outcome in the same amount of time or a shorter period of time. People don't think about the human element of what that does to people's identity. So AI, 2026 is the year of AI um automating the white-collar worker, right? Where AI is getting crazy good at making presentations, looking through documents, Excel, like all this stuff, sort of like this business process where you can go and upload a you know spreadsheet and ask for, you know, your marketing numbers, your whatever, your synthesis, and it's pretty good, right? You definitely need humans to look over it and review it. But now someone who's a lot more junior can get to a lot of these senior places a lot quicker. And and if you're not there to train and coach and help your valued employees, they can feel some type of way. Because someone who has all the experience but isn't willing to learn AI versus someone who has way less experience but is willing to really push the boundaries of AI, the the latter could uh end up a lot more productive than the former, which really messes with people's heads. So I'm very concerned about what's going to happen with the workforce. Yeah, I could I can imagine that, you know, especially within the United States, like so much of people's identity is tied to their career, right? And the idea that you could spend a a lifetime learning something and suddenly somebody else could be really good at it really quickly. I think the the the side to that right now that I'm I'm finding solace in with what I do on a day-to-day basis is I still feel like at this point, and maybe it'll change, human in the loop is critical. And there's times where I'm like, give me this information and it spits it out. And I'm like, if I were not fluent in what I do, right, I would look at that and go, that looks correct. The minute I look at it, I'm like, what is that? Like that doesn't look like what is that? Like, fix that, like that is a weird and they're like, oh yeah, we inferred this, or you know, I I took the data and just thought maybe that would be something you would want to see. And I'm like, that's not, that's not what I ask for. Like, like that's that's not great. So I think that there's still a I think there's still a place for people who are experienced and understand their business. Again, to come back to this consulting thing. Like if you're consultative and thoughtful and you understand a certain industry in a certain way, I think AI can be a huge tool to augment you. But I think if you use it as a replacement to that knowledge, that I think you're gonna be in a dangerous situation where you're getting data and information back that looks fairly accurate, but if you're making business decisions off of it, you could be in a world of hurt. You could you could make some really bad decisions. One 100% agree. And I think that that is becoming less and less as these systems get better. Um, but the people who just don't use them at all, like that's the those are the ones that are that are really gonna get like a person with experience who knows AI enough to know the limitations and have a feel and a sense for it, and like a oh wait, it couldn't do that a month ago. Now I can do that, right? Like, like that's where you need to be as a professional worker to stay ahead. And I think a lot of people won't

What's next: multi-agent swarms for MSP go-to-market

Jimmy Hatzell 51:30

even do anything and they're gonna be in for a rude awakening. Um, before we get to the questions, last question for you on this is what's super exciting for you right now? Like, what do you what are you like without giving away something that maybe you guys are working on, or maybe you want to hint at something like, what are you just like, holy shit, by the end of this year, like this is gonna be hot? Like, where where are you getting jazz when you get up in the morning and come into the office? Like what's going on? Yeah, we're building basically a multi-agent swarm that an MSP can come in and connect their data, like their PSA or their their customer list or upload it or whatever they want. And all these agents go out and start researching the companies, um, start pulling in whatever relevant data they can have, suggest surveys to send out, that kind of stuff, um, and do the hard work preparing for AI meetings where you can, you know, I mean, obviously you have control over it over an MSP, but we I think by the end of this quarter, maybe middle of next, um you'll be able to click a button and a couple minutes later an AI will come back and say, hey, Acme Corp, uh, here are their top uh five ways we think they can use AI. I've actually drafted up proof of concept uh demos uh for that. Um here are the right people you should talk to. Here are the things that they probably care about, and here's some information um that you have as an MSP that nobody else has, either from you know, ticketing history or DNS or whatever, uh, that can help you uh navigate that conversation. Click here to draft an email, click here to generate a presentation. Um, so like that's what I'm working on, and it's like very fun because this is a hard thing for MSPs, right? Being able to articulate this stuff, being able to communicate it. So we want to make it super easy um from uh a go-to-market perspective, but also an AI knowledge perspective. It really is. That's hot stuff. So um all right, so

Closing five: books, bands, and the worst meeting ever

Jimmy Hatzell 53:30

let's hop into our five questions here. So uh my first question is what's the best business book you've ever read? Like tell everybody out there the book that changed your life when you read it. Like, you got one? Um there's lots of them. I think uh one of my favorites is Beyond the Cloud by Mark Benioff. Um I'm actually a Benioff Salesforce stan. And I don't know if that makes you think less of me if you're listening to this, but I I do really like that. But nobody gets no one gets in trouble for for Salesforce. Like everybody's like, yeah, let's just do that. Like it's expensive, but it works. Why did you why did you like that book so much? What it what what did it do for you? Um I don't know. I mean, Benyolf is like uh one of the best marketers I've ever seen. Um and at the time that I was reading that, I was in a marketing role. Um, and you know, he he um like invented SaaS basically, which was pretty cool. Um did some some some cool things along the way. And um yeah, you know, like I I just I thought the building of Salesforce, maybe not the Salesforce we see today, and and um I might not agree with Mark Benioff on everything, but he's just like uh he's just an impressive person. Um yeah, it's it's uh it's a great book. It's very good. Cool. Um all right, what's your favorite curse word? You're welcome to say it or not. It's up to you. I don't mind. I put that these are that there's explicit language used on the show, and I post on YouTube. So it's it feels weird weird saying my favorite curse word is fuck, but it absolutely is. I drop F bombs left and right. I I use it like pepper. Um it's like I spice up my conversations with it. So yes, fuck is a good curse word. I don't mind that one at all. It's it's you can use it in so many different ways. It's so fantastic. Um there was a we had a a guest on recently and he had a he had a word that I actually couldn't, I was like, we can't use that. Like I was like, I was I was scandalized, and I am uh I'm a sailor. I was like, whoa, uh uh yeah. So uh I was like, we we're we're gonna go ahead and edit that. Um but I was like, but thank you for for coming prepared. Um so if you were on a desert island and you could have the music of one artist or band, like it's your band, they're in town, you're always going to see them. You'll go see them someplace wild. Like I'm a music guy, so I just like to ask this question like, who is it? Who's your band? Who who who who are you going to see? Um, I mean, I'd have to say the dead or fish or something, because they're just gonna keep iterating and jamming, and you know, they could play the same 500 songs, but they're gonna be a different way every time. Yeah. Did you go seeing with the sphere with John Mayer when they were doing Dead and Company? Did you go there? No, I didn't. I didn't. Yeah, I did get invited to see Fish at the Sphere last week when I was in Vegas, and I didn't do that either. Dude, I've basically stopped going to concerts almost entirely since I founded this company. I think I've been to two where I used to go to like 20 a year, so it's been it's been rough. But you have to take care of you, Jimmy. I know. I think you gotta go. Like a fish concert at the sphere with Anastasia and everybody, that would have been so much fun. So that's a bummer. Yeah. I haven't been to the sphere. I've seen it. I walked around the outside of it a couple times, but I've never been inside. But I heard it's sick. Really sick. There's apparently uh there's saying there's a rumor that Tool's gonna do a residency there. And I'm with Oh, there you go. Yeah. Yeah. And my my friends are going to see Metallica there, I think, in February, which I think would be. Tool in the sphere could get dark. Like that. I might like freak out if I'm a little bit like some of their, like especially they're doing the visualizations. And if people remember I'm dating myself, but like back in the 90s and early 2000s, they had some music videos that were real like weird. And I mean, if they if they incorporate some of those visuals into what they're doing, like that could get wild. But uh that would be super cool to do. Um so you can change names. Uh you don't have to be you don't have to name names you don't want to, but if you had a business meeting, a sales call, an HR meeting, something that was like the worst, like the worst ever, where you're just like, yeah, no, that was just weird and it was awful. Like, is there one that comes to mind where you're just like, yeah, that that probably never should have happened that was that was terrible? Well, I'm notorious for hating meetings. My whole team knows that I hate meetings. Yeah. And like, like hate meetings. And anytime there's a meeting about meetings, I like absolutely like lose my shit. I do. It's like pretty bad. Yeah. We're gonna get together and have this. I hate that. I've we used to do that. I'm like, is this a meeting about a meeting? And they're like, and everybody would get real quiet. I'm like, come on, y'all. So you want to schedule a meeting to talk about how we hate meetings about meetings? Yeah, that's what we should do. We'll just put it on the calendar. That's a good one. Okay, so last question is if there's somebody that you would recommend you can make an intro to that I should have on this podcast, like who is it? Like who who do people in the channel need to be talking to right now? Who's out there who's saying interesting things, like who needs to, who needs to be heard? Oh, um. I don't know. Have you had Chris Day on? He's been like reaching out to me, like trying to get MSP AI, where's it going? You know, back in the in the like uh he he he him and I have recently connected and he's been all about AI and trying to uh really really uh lead the ship there on how things are going there. Yeah, yeah. Yeah. Yeah. Yeah. I'll reach out to Chris. That'd be cool. We'll figure that out. Yeah. That's that's a good opportunity. Cool. Uh well, dude, uh, so we're gonna wrap it up, but I just I can't thank you enough for you know being on and taking time because I know as you said, you don't even have time to go to a concert. So you're a very busy man, and I appreciate you know you you hanging out with me for the last hour and chatting it up. So thank you for everything, and I'm wishing you guys the best, for sure. Yeah, my pleasure. All right, all right. Thanks for having me on.