ERP129 - Less Admin, More Impact: AI for MSPs

Today I'm joined by Matt Tougas, CEO of Mizo.

We'll explore how Matt's background as a software engineer and MSP operator led him to develop Mizo, an AI-powered help desk automation platform.

We discuss the realities of AI adoption in the MSP industry, why AI agents can sometimes outperform humans in quality and consistency, and why Canadian companies have such an outsized impact in the global MSP software ecosystem.

We're talking MSP innovation, Canadian tech pride, and the real-world impact of agentic AI in service operations. Let's get started!

This episode is brought to you by Opsleader Pro. A place for MSP owners and managers to get the systems and tools they need to build a stable and growing MSP. Part group coaching, part peer group, everything you need to run a successful MSP.

What is Evolved Radio?

Evolved Radio Podcast: Interviews with technology experts, industry thought leaders, business leaders and other interesting minds. Exploring the evolution of business and technology.

Matt, welcome to the Evolved Radio podcast.

Thanks for having me, Todd. Happy to be here today. All right, so

we'll get started as we usually do. Kind of give me a bit of

a background on yourself, kind of where you've come to in the MSP

industry, bit of your background and we'll jump into it from there.

Yeah, definitely. So my background, I wasn't as an

MSP operator at first. I've been a software engineer, worked in

the sphere in different roles, mostly in operations in the end. And then

I joined an MSP four years ago now

and just as a CEO operation director, helping them

scale and really going into nitty gritty and learning the trade

I guess by doing lots of it. And that's how I came into

the MSP sphere. So really by doing everything

from tickets to trying to manage and scale operations in the,

I would say lower, smaller to mid size MSP if we were growing.

Okay. And current role kind of being a bit different as

well. Yeah, definitely. So

this MSP adventure led us to realize

that we were well positioned to develop an AI agent that

could help solve some of the issue we were trying to solve at first

through human labor, through processes and in the end we ended

up developing Mizo and that's now I am CEO of Mizo. So we're really

deploying AI agents for our customers, helping them scale their

operations with digital labor. All right, so, and this

touches on one of the things I find really interesting about the MSP

software industry is I feel like most of the

really sort of darling success stories of the industry and probably

50 to 70% of the software out there

really sort of develops out of some MSPs need. Right.

Like I had front row seat to the development of IT glue and

this was sort of a product internally that we needed as an

msp. And every time we had a strategic session we talked about our, our

strengths as an organization. Our documentation platform was always one of those. It's

like there's an opportunity here, like this is good for us. There's, there's

definitely an opportunity for others to kind of leverage the same strength. So you

know, the, the software in the MSP industries is

so much driven out of MSPs solving their own needs.

I'd love to hear you kind of talk about how you guys identified that

and how I guess sort of three stage

process. A very long question, but I'll let you work through it

and ask me if you need me to re prompt you on this as well.

So how did you identify that this was something you guys needed to do,

then determining you should do it yourself. And

then when you sort of realize like, hey, there's a market opportunity, we should

actually start selling our own solution, how did you guys sort of work through

that process? Yeah, definitely. So

our goal at first wasn't to get into the AI

software or the MSP software company. Our goal was to

fix our problem internally. So we were scaling,

we had good traction in MSP. We were growing 50 to

100% annually for two, three years. So that meant revenue

growth was great. Operation growth has its challenge.

So we sort of tried everything and not everything. I

mean there's a lot of things we could have tried too, but just

trying to always hit the same nail the same

way doesn't fix the same issue. And we were seeing that as

we scaled. Adding more bodies to the

problem, adding more people internally didn't get us the

efficiency we were aiming for. So the

realization was really we were at the conjuncture

of AI agents or it wasn't even agents back then, it was

AI or chatbots or the hyperscale of this world

were getting more, I would say commoditized.

ChatGPT had been around for a year or two and people

were starting to see the value in that. We saw that technology as

something that could help us. And that was really the, the moment where we

thought internally let's try to do something out of it and let's try to

fix some of the problems we have. And the first one we

started with was, I guess the obvious one was triage and

dispatch where we saw we spent a lot of time, we

were one of those MSPs which probably you wouldn't recommend, but

having one of the level 2, level 2.5 tech doing the

triage and dispatch because we thought we needed that high level person

to understand issues and be able to know who can

work on that, how should they solve that issue. So, and this

was obviously very costly. That person, we aimed her,

that person to be sort of a

service delivery manager, have service delivery manager, have

dispatcher. It was just too much they couldn't do

was more than a full time role. And in the end all that person ended

up doing was tickets like the so,

so the, the idea came from that if we had that problem we,

we tried with processes to fix it. And we thought maybe the

way to fix this is just to externalize that function and

instead of hiring an admin to do it, we can do it through an AI

agent. And that was really the start. It was,

I guess more than two years ago. And we worked on the product

for about six to seven months internally as an MSP and

just iterating on it, starting with really basic

categorization, renaming, just very basic stuff and

seeing can we have a value of that. And

it was like very, I wouldn't say hacky, but

very low level integration with the PSA we were using back

then, which was Halo and getting those

features on and getting those,

those replacement on or delegation on. And what we saw is that

our back then dispatcher, level two

dispatcher was able to delegate those tasks and we

had even better results than humans sometimes.

So I say humans are

flawed in many ways. And what we saw is that

the operations we did were often better from an

ITIL perspective, from a standardization perspective and what our

dispatcher would have done in that case.

So after that six months we sort of

stopped and said what did that project cost

us and what did it give us? And our conclusion was that even

though there was a lot of R and D investment, it's obviously expected and

we knew about that, there was definitely a lot of value

we got out of it. And we, I think we saw that it was just

getting started. We as a company, as a software were just getting started. But also

the hyperscaler, the AI capabilities on the model

side and the product were also very early.

I think we still are in many ways. So that was

that point, that was December 2024

when we decided, yeah, let's try to make a product out of this.

We had great results for ourselves. Let's try to see if others in the

ecosystem could benefit from it.

So we did that and we went into the market,

went to other MSPs in our peer groups, other MSPs we knew and

asked them do you guys have that problem? And the

conclusion was a lot of people ended up having the same issues

and we tried it out, developed some more integration

with new PSAs, new softwares, just adapting what we had

and we saw some good results there. So this

brought us all the way here I guess more than

a year later, a year and a half later. But I

think I'm missing one part to your question. So we had the, I

think you got it is like why, how did you do it? And then the

pivot basically. I am curious though,

how much of this was people being

attracted to the solution based on what you had versus like

I think we have something that we could market, right? Because like I've seen both

of those where like a lot of people are going to peer groups or they're

just talking to sort of like peer MSPs in the industry and

you know, you're like, oh, you're doing this. Like, how are you doing that? Oh,

like, can you show me this? Oh, this is really neat. Like, can I have

this? Like before it's a product, was there an element of that or was

it sort of a conscious decision? Before it was really sort of public to peers

and other people in that you guys had in peer groups and stuff?

I think it was a bit of both. So

coming from having other businesses

before, I've always liked branding. So we sort of thought of a

name for the product and logos and I mean we started

ironing on that of creating that company. But the first

few MSPs we talked to, they were just seeing us as like, yeah, their

friend developing something nice and sort of their in house development

team just helping them do more. So I think,

and it's a great relationship to have with your customers too, that proximity. But

it was a bit of both. We thought about it as a product, about having

that company, creating company out of it. But I think the first customers were really

like, yeah, I want to use this too and show me how you're doing it

and get it in my environment. And this is the problem I'm having too. And

then you get a list of problems that you're having to which you can

help solving. So that was really great. Cool. Okay, great.

That also lends to. The other thing I wanted to explore here is like, I

am continuously shocked. I've talked about this on other podcasts and

all over the place because as I was traveling last year and

late this year as well, about how

we, the people that are leading edge on sort of AI and automation

are pretty far ahead of, I think the industry average. So you guys

were, I think, way ahead of the curve in the fact that you were developing

this in house and with enough time to be able to then turn around

and start or as a separate product. But I'm still

shocked by the number of MSPs that are not on

AI in a meaningful way. Right. And I see a couple of

points of pushback about this. And you guys,

I think like having a lot of touch points with potential client

prospects for the solution, probably see a lot of

this. But you know, the couple of points of pushback that I see

are like, people just don't have the time to know that the solutions are available.

Like I put out this, this is the one I've sort of

banted on a Lot about is I put out this call out to

a group of MSPs and said, if you had an AI

army to build a solution for you, what would it be? And

invariably, nine times out of 10 people described exactly what you

guys do and for that matter, several other solutions in the market.

I'm like, come on you guys, what you're asking for

exists. Go get it. So I'm partly

flabbergasted by the fact that these solutions are not as prevalent as they

probably can and certainly should be in the,

in the wider ecosystem. And the other part

that I see is, you know, people feel that it will somehow

change something tangible about the human connection. And I see

this maybe as an extension of a lot of owners are really

sort of precious about live answer to a point where it

becomes actually problematic for the MSP where like we have to have a

live answer, you know, we don't want, you know, it has to go straight to

the tech, you know, like we can't have anything go to voicemail, blah blah, blah,

blah, blah. And that becomes problematic over, over time. And I don't think it's as

important as people think it is. And maybe it's sort of related. Like you said,

you actually had better results utilizing the AI than you did with

human labor in a lot of those cases. Right. So maybe if,

again, kind of a long, long question, but just exploring that idea of like,

why isn't this more prevalent yet, considering how we should

be more technically a bleeding edge in this industry and

AI has been around long enough that it sort of blows me away that it's

not more prevalent in a lot more msps right now.

Yeah, it's, it's actually a great question. And I mean it's,

it's something we, we often ask ourselves. I think one

of the, one of the

issue or one of probably

something that's slowing down this AI adoption is the hyper

customization of most MSPs.

I feel what we see is that no

processes are the same. We often hear like that's my secret sauce and

that's how I'm doing it. And it's also

my, I want to keep it for myself. I don't want to share that secret

sauce necessarily, but I think a lot of

MSPs have been built on hyper

customization or hyper specification of

processes where everything is very specific to their

operations and their company and their context and build

around that. Most people probably feel that AI

wouldn't be able to understand that

or to adapt to those situations. While it's completely

the Inverse. If there's something that's good at getting

context and getting an answer out of it, it's no,

it's good at being adaptable to adapt to any context.

So I think that's one of the

things that's slowing down adoption.

The other one is probably

maybe, and this is really speculation my part, but MSP

is being really at the leading edge of technology. I think they've been

the first to see the problems with AI. It was

probably two years ago, like there was some hallucination you would ask ChatGPT

whatever question it would probably give you something wrong and for,

for many different reasons. This happened back then. And I think

a lot of the hyperscalers have worked on that. The AI labs have

worked on that for, from their perspective. Also the prompting side on

the agent side, there's a lot of improvement that were done

but I feel like sometimes people are still stuck there. Yeah,

I'm giving it something and I get AI slot out of it.

This is maybe something we're scared about and it ties into

the service quality. I feel like most

MSPs want to keep their service quality at the highest level possible

and feel like having a conversation with a badly

trained AI agent with their customers would lower

their service quality. But while it's

inverse, if you have something really good that is fine

tuned to how you work as an msp, you get those results much

better. And people would, I think would rather

chat with an AI agent, have a conversation that's

tickets all very fast and they have a human in the

loop where it needs to be and still be able to talk to a human

if they need to while instead of waiting for a human

to be able to solve it and then having that really inconsistent service quality.

So these are my takes. I mean there's probably

more to it and I'm curious to get maybe your impression how

you feel about that. I think you're right, the quality does matter.

I'm not saying that people are sort of misguided

in wanting that obviously better quality service, especially at the front line

is important. It's a service based industry and if your first interaction with the

help desk is terrible, then that sets a really bad tone for

that relationship as a whole. But I think to your point, like I had

Megan Giholy on I guess a couple of years ago now,

like it was a long time ago and we were talking about AI agents, right,

like customer service agents and she said, you know the problem, like AI

agents are actually really good. The problem is most of them are terrible. So like

our expectations are so low because like the implementation of them are bad, but

when they're done well, they're amazing. And that's sort of like it's the quality

of the service that you provide, both in a human interaction but also in

an agentic interaction I think matters. Right.

And to some extent, I think a lot of the MSP industries kind of run

off their feet and they look for cool solutions around

automation, but they don't really have the time or the

bandwidth to capitalize on those things. And I think it's maybe just sort of

the same thing. Right? So yeah, those are some of the big

ones that I tend to see as sort of the sticking points for

this. But the other point that you make I think is actually

really important for how you guys are a bit different than a lot of the

other solutions in the industry is that so from a

philosophy standpoint, like you guys are much more facing the technician

than it is facing the client. So you guys don't really have those

problems. I mean right now, depending on what your roadmap looks like.

But right now a lot of the agentic interaction is

with the tech, like trying to make them more expedient, providing

them context and information rather than trying to front end the

conversation with an AI and then sort of creating that risk

of a bad interaction between an AI and an end

user, I guess, right? Yeah, definitely. And

we really see the value, and I

mean the value of technicians are into solving hard problems to solve

and also creating that customer relationship. You have an

MSP as an SMB. You do business with an MSP

because they are your IT department, you know them probably by name

or you have a relationship with them. And that's the value for

technicians in the msp. And we see ourselves as

just helping those technicians do more. And for

us the future is about having technicians manage fleets

of agents and mostly delegating most of their tasks,

if not all their tasks to AI agents and just being able

to have them in the loop where they need to be, if there needs to

be human contact for some reason, if there needs to be an approval for some

risky operation. But our goal and our

philosophy is to have those technicians perform better and reduce the amount of low

value tasks and non rewarding tasks they do. And if you

ask a technician what they do on a daily basis and what they like doing

on a daily basis, I mean it's that Venn diagram isn't

that big. And they like to solve problems, issues and they don't

like to create reports and document

resolutions and send customer requests

for a meeting or whatever. So all those lower value tasks that

are not, first of all they have low value for

customers, they have low value for technicians. It's something we can easily help them

with and also elevate those technicians to

higher tests that devalue more. So this is really

philosophy we have and all the interactions we're

having are always transparent. So even though we have interactions

through the PSA with some of our, with end

users, everything is done on behalf of the technicians. So

we're just helping those technicians do more and

reduce the amount of work they do by themselves

for those interactions. Okay, great. I guess the other thing

I wanted to touch on here is as people may have, may have picked up

with your, your fantastic accent is.

You're a Canadian. French Canadian. And this is another

aspect of the MSP industry that I find really fascinating. And like there's a long

history to this. Like I obviously, you know, it Glue and

Scalepad, fairly recognizable Canadian brands. Also

Passportal. But even all the way back to like Enable,

you know, being out of Ottawa, like a Canadian company. There's a really

long history of MSP software in Canada which is not a

hu, especially relative to the us.

I mean you guys must have thought about this. But like why do you think

that is that there's, you know, as Canadians we tend to punch above our weight

in the MSP industry when it comes to software.

It's a great question. And

there's, there's probably two parts to it. I think

the, the first part is the, the

SMB rich ecosystem in Canada. So a large part

of the Canadian economy is based on SMBs. Like it's

probably higher than most Western countries. So this

means more SMBs, more MSPs to serve them.

So this is definitely something that we see

that there's a lot of MSP and this is probably a lot, there's even

much more in the United States, but there's many MSPs in Canada.

So as a, as a Canadian

company or a Canadian msp, you have much,

you have a lot of ground to test your product on different

customers. And so that's, I think that's one of the

first one. And what we see is that In Canada most

MSPs will have smaller clients than what we see in the United States.

Less co managed. So they have larger client bases,

meaning that they have well, larger

amount of clients or count of clients which means that they have

more diversity to test their ideas.

I think that might be one of the first one just being

like ASMB fertile ground.

The other one might be just related to

innovation savviness in Canada. I think

as Canadian

we're much less risk prone than

United States. And you see this in VC world, you see this in any

startup world and that's why we have such a deficit probably

in productivity and in innovation. But

launching a product in the Canadian market is probably much

higher bar to launch because people are very, very demanding.

I think since they are less prone to innovation, Kinean

MSPS will want to have a product that's much higher quality.

So this might be something just in having the

philosophy in how canium companies build product they want to build.

And it's always a challenge when you're building a startup but you

can never launch too early probably and if you feel like you've launched at

the right time, it's because it's too late. But

I feel like this is something that Kenyan companies

might be more inclined to having that higher standards, higher bar for product

quality and that might lead them to creating better products.

But this is honestly just

our take on it and speculation. I mean you guys are a Canadian

software development company, so your opinion matters in this matter.

Just curious, did you guys leverage like shred for those not

familiar and especially those not in Canada, this is scientific research and

experimental development. It's like a, like a tax incentive program that

the Canadian government has for, you know, anything scientific, but it also

lends to software development and process innovation. Was that something you guys were able

to leverage? Yeah, definitely. So as

you mentioned, there's shred, but there's many other governmental

programs to accelerate. Yeah,

Iraps one of them too that you can.

So yeah, we leverage all of those actually. So

as Canadian also we have that advantage of being able to

subsidize most of our development R and D,

I wouldn't say subsidized, but

get those tax returns for it helps, right?

Yeah, definitely. It does make a big difference. That paired with

having great talent pool is probably something that that

helps a lot. For those again,

not familiar with Canadian sort of markets and

you guys based in Quebec. There's a lot of software development like

Ubisoft and a lot of the Canadian software development houses.

EA is Canadian west coast. But Ubisoft being huge

in your region, that's the one that sort of comes to mind for

me. But I'm sure there's a ton of other sort of companies

that tend to create a bit of a gravitational force for good talent in the

region, right? Yeah, definitely. And I think

us being based in Montreal Montreal has the longest survey with

AI being one of the first cities to have AI labs from

Meta that have been here for a while now. Google also has an

AI practice in Montreal. This AI I think

there was a good movement around AI five years ago generating

a talent pool in Montreal and generating

those various deep research on AI

development in Montreal. And this obviously shifted a lot

to applied AI now as the fundamental

research is mostly taken care of by the big labs. But

I think we have that talent pool that stayed there that has been doing

AI before AI if you want. I mean even Geoffrey Hinton,

the godfather of modern AI from Toronto, right?

Yeah, exactly. So we have a lot of. So we have a

lot of those talent in Canada that have been at the forefront of the

research before it got before it scaled as it has. So

I think that talent is really leverageable

to apply those principles and start from first principle

and not just learn about how ChatGPT works from

reading the API doc, but really understanding the fundamentals,

fundamentals of models and being able to really

fine tune the usage to how it should be used. Yeah,

cool. The other piece of this kind of lends to

small upstart David and Goliath

type approach. I'm curious sort of how you think about your

opportunity as a small development company relative to

a lot of the PE led companies,

you know, the big giants in the industry.

I personally feel like there's a strong and movers advantage for

smaller companies to be able to iterate and innovate and improve

upon those platforms certainly at a faster cycle and than

some of the larger companies. I assume you think that that's true as

well. But I'm be curious kind of how you feel like you guys have a

competitive advantage as a smaller entity being able to

capitalize on on sort of market opportunities and be a bit more

nimble. Yeah, I think

we try to do what

we preach really internally and I feel one of our main advantages

being an AI native company and what we define as

an AI native company, and there's probably a lot of buzz around it, is that

even though we're a small team, we have agents working for us doing all sorts

of tasks and on the marketing side, on the development side,

obviously on the account payable size

on the accounting side. So we deploy our own AI agents

for many tasks internally so we can leverage that technology

and we can also create

processes around using those AI agents and not having

to rethink already scaled

processes to how we should use AI internally so we have

the opportunity of leveraging AI everywhere. We need, and this is

really, I think one of the main advantage and it leads to, as you

mentioned, faster iteration cycles. And I think we have iteration cycles

that you can match

in any industry. And I mean we try to stay on top of

that, but we ship multiple times a day. And this is something that

with such a small team that we can do as an AI native

company that a lot of the incumbents wouldn't be able to do.

And I think even starting five years ago, you still

have a lot of that gap to bridge

because you've built your processes and you hired people and you build

your company processes into working how they should work around

people. And we are building our processes into

how we should work around agents and how humans can intervene in the

right place in the loop through agents.

So this is, I think it's definitely a fundamental shift that just

creates more faster iteration cycles and in the end better

product quality for the customers. That sort of reminds me of

reminiscing back in the good old days of Connectwise when

Connectwise again they were building their own software for themselves

as an msp. So it sort of speaks to this kind of building to scratch

your own itch. But I think part of what made them successful in the

very early days was they were building it for themselves and being able to

test the things that they needed, solving for their own problems. And I think because

you guys were born out of an MSP and still have

sort of that, that high touch point of like what do we need, what do

we see internally, what's working and then sort of iterate from there. It gives you

that, that higher fidelity understanding of what's

required to make this work in an msp, right?

Yeah, definitely. And I think one of those advantage also is that

we're not afraid to change things. We

built something that didn't work, we'll just scratch it and start again. And

that's something we can do because we have that proximity

and we can create that iteration cycles, but we also have that velocity

of developing it. So where I think as a larger company you

have to plan ahead and create that big roadmap and then you launch and hopefully

it works and you do all that product led development all

around. But I think the processes are so much larger

that you end up probably

launching a product a little too late where it doesn't exactly fit

the need where it's at now. So being in

an MSP and also having that evolution speed allows us to

continuously adapt and our customers are changing on a day to day basis.

And we talked earlier about how they

are starting to adopt AI. But AI will obviously

create a lot of process change for our customers. So we need to be at

the forefront of that and being able to adapt to what they are currently

doing and will be doing in a month or in a year. Yeah, but

lends well to a question. Before we started recording I was like is

this worth asking? But I think you had a good answer to this. So we

can kind of go down this road and it sort of lends from that is,

you know, the future of AI. Right. And totally recognizing to

your point. I always joke that the MSP industry and IT in general has a

six month shelf life. Right. And now like I would say with

sort of the cycles of AI and how quickly things change, it's like

maybe three to six week shelf life. Like things are moving so

fast. Right. And I'm curious sort of how you think about, you know, planning

for the future for in an industry that is subject to

such radical change within a six month period.

Yeah, it's. I think it's a great question.

I feel like the, when you look at the fundamentals of

the industry, they won't be changing like and the MSP

industries are all about service quality. It's all

about providing service, first of all service quality. And there's a lot of

probably uncertainty about how the business will evolve.

And I think PAX8 and a lot of players are pushing for that

MIP which is obviously really great into how the

MSPS can evolve into being strategic advisors to their

customers, into uptake technology which obviously includes a lot

of AI. But I feel like

this is a direction where we're going in and this means that

all the tasks to be done will remain the same. It's just that our

customers will want to focus on tasks that have a higher value for them

and this will be all the strategic advisorship for their customers.

They will want to have their best tech players and their best technicians

into advising and upskilling them into advising their own

customers into how to take that AI

turn and leverage AI as much as they can internally. So

we have that opportunity

of having them delegate all their lower value tasks to us and all their service

desk. And this comes into all those agentic

level one, just doing the full level one

agentically and doing a level two in the same ways and really

removing all those lower value tasks so that the msps can focus and where do

they deliver the most values and change

that part from their P and L of the support team or the help

desk from being cost center to being just a cost

of doing business. And they will generate a lot of good revenue out of

that and increase their capacity to generate even more

revenue through the advisorship. So yeah, I certainly think that

is the direction of the industry is sort of a joke that

we're kind of going back to the 90s where there's a lot more consultative work

that's happening where you know the agents can take

care of the a lot of the rote work that

needs to happen still in a lot of environments.

I know you and I have talked about back in the day when working

on compact computers and we had like a, like a self

healing floppy disk that we put into it and this has always been the dream

of like telemetry tells us the things that we need to do. Computers take care

of it by themselves. Right. But, but we're almost there so

it's exciting times for sure. This has been great. Matt,

anything we haven't touched on or any other sort of last minute

topics you want to hit on before we wrap?

No, I think that sums it up. Thanks for that conversation. Thanks for

having me today and I'll. Link to you and Meso in

the show notes as well. Just a quick call out

if people want to hear more about what you do or connect with you. What's

the best place to connect with you and do that?

Yeah, of course you can reach out on LinkedIn or email me at

Mattizl Tech and I'll answer. So happy to chat with

anybody, happy to have conversations around AI, around how to

deploy it in your msp. So looking forward to that. Thanks

Matt. Thanks to you Todd.