AI thinks, automation acts: Rewst founder Aharon Chernin on building the automated MSP

Robert Dutt: Hello and welcome to In The Channel from ChannelBuzz.ca, bringing news and information to the Canadian IT channel community for the last sixteen years. I’m Robert Dutt, editor of ChannelBuzz.ca, and your host for the show.

If you’ve been following the conversation around AI and MSPs over the last year or two, you’ve probably noticed that a lot of it is pretty fuzzy. AI is going to transform your business. AI is the future of service delivery. AI this, AI that – but not a lot of specificity about what it actually means for the way an MSP runs its operations day to day.

My guest today has been building technology specifically for MSPs for more than twenty-five years. He founded Perch Security, which was acquired by ConnectWise back in 2020, and then turned around and founded Rewst, a workflow automation platform built from the ground up for managed service providers. Rewst now has over 1,500 MSPs on the platform, which means he has a pretty clear view of where the channel actually stands on automation – not where vendors wish it stood, but where it actually is.

We talk about the difference between an MSP that does automation and an automated MSP, and why that distinction matters more than any specific tool. We get into why AI and automation are not the same thing, and why confusing them leads MSPs to make bad decisions. And we look at what the operational stack of an MSP actually starts to look like as AI moves from advising on workflows to generating and executing them.

Aharon Chernin is the founder and CEO of Rewst, and he’s been thinking about this a lot longer than most.

Let’s get right into it. My chat with Aharon Chernin.

Aharon, thanks for taking the time. I appreciate it.

Aharon Chernin: Nice being here.

Robert Dutt: Rewst isn’t your first time building specifically for the MSP market. You come out of Perch Security, acquired by ConnectWise. What did your time inside ConnectWise’s world teach you, and was Rewst a direct response to something specific you saw from there, or from the community there?

Aharon Chernin: I was actually only at ConnectWise a couple of months. Really, the Rewst idea came from working with MSPs for years while I was doing Perch. I was learning more and more about how MSPs operated. I saw the tools they were using, the problems they were having. And, mind you, this is the era circa 2017-18, and they’re using tools called PSAs – professional services automation – that didn’t really automate anything. It had me scratching my head: what’s going on here? But then I just continued moving forward because I had security stuff going on.

And then there were RMM tools called Automate, and I was trying to figure out what these things actually automated. It was just endpoint stuff, right? But there’s so much more than just an endpoint. And then I saw a bunch of single-point solutions – software products out there doing a single automation, not even calling it an automation, just calling it a software product. So once I had the opportunity after Perch, I went and started investigating what were these quirks in the market I was seeing, because automation is much, much bigger than what the market actually thought it was.

Robert Dutt: Well, and it’s for so long been one of the key premises of managed services – the idea of automate everything you can towards success. So it’s interesting to hear those observations. You’ve got the platform now with 1,500 or more MSPs on it – you probably have a better view of the real state of automation adoption in the channel than almost anyone out there. How would you honestly characterize it today? Is this a story of meaningful progress from the scenario you saw back in your Perch days, or is it still pretty early for many MSPs?

Aharon Chernin: It’s still the beginning. We only have 1,500 MSPs. And how many MSPs are there – depending on who you talk to, 60, 80,000. We only have 1,500 of them. So in my mind, these are all early adopters in automation.

But when it comes to what adoption actually looks like in an MSP that successfully automates, it’s cultural – 100%. If you look at cybersecurity – Perch was a cybersecurity company – you’d be looking for the correct size MSP that focused on security to resell through. But when it comes to automation, every size MSP needs to automate – small and big. And we’re actually seeing that. We see really, really small MSPs automate and really, really big MSPs automate. And we’ve also seen both those sizes fail.

The number one reason they succeed is culture. The buyer – the buyer could be the CEO of the MSP, or a director of managed services, whoever can write a cheque – that person has to want to automate. And the engineer who’s actually going to do the automation, they have to want to do it too. If the buyer wants to automate but the engineer doesn’t, it’s not going to work. If the engineer wants to automate but the buyer doesn’t, it’s not going to work.

And that cultural thing extends further than just the want. The CEO of the MSP should be running around saying, “I want to be an automated MSP,” and excited about it. If they’re not excited about it, they’re going to be a part-time automation MSP. The way I like to say it is: you’re either an automated MSP, or you’re an MSP that does automation.

Robert Dutt: What do you find helps flip the switch from one to the other? What is it that gets those teams that are either misaligned or not aligned at all to get things lined up and moving in the right direction?

Aharon Chernin: It’s really an open line of communication between the buyer and the person implementing. Because if the buyer has an automation idea – just one, a single simple basic thing that would save the company time or help improve service delivery reliability – and that engineer performs that small automation, and they talk to each other, and the engineer says, “Yes, it’s running. Yes, it runs fifty times a day. We’ve saved eight hours today running this automation” – and that actually gets back to the person who writes the cheque – there is alignment. The tide has turned. Suddenly the MSP says, “How do I dedicate more people to helping automate this business?” It’s a matter of getting that first win and getting it in place.

Robert Dutt: There’s a lot of talk, obviously, about AI. And Canadian MSPs are being sold a lot of things that blur the line between AI and automation. You’ve talked about that distinction – the idea that AI thinks and automation acts. Can you expand on that? Because I think getting that framing right can help change how MSPs make decisions and think about how they’re structuring things internally.

Aharon Chernin: AI can’t touch anything by itself. This goes back to: AI thinks, automation does. Take ChatGPT, for example. ChatGPT is not an AI. ChatGPT is a tool on top of AI. The AI is GPT. The tool is Chat. So just having AI gives you a lot of answers to a lot of questions, but nothing gets done. You need the tool on top of the AI. I can’t think of an easier way to define it than that. There are an infinite number of possibilities of what you can do with a tool that leverages AI.

Robert Dutt: So you guys have RoboRewsty now. You’re moving from AI that guides building workflows to AI that generates those workflows. That sounds incremental, but to your point on thinking versus doing, I suspect it’s more significant than that. What actually changes for an MSP team when anyone in the org can describe a workflow in language that’s natural to them and have it built for them, rather than having to go back to that one person who knows how to build out the automation?

Aharon Chernin: AI is easier to understand than even that. We need to think of it as just another employee. Now, depending on how much the business trusts that employee is how much governance we’re going to put around that employee. If there is one hundred percent trust, it gets free will and can run freely. If there is zero to ten percent trust, every step of the way needs to be gated by a person.

But there’s a balance. One of the funny things about cybersecurity is everyone looks at the cyber risk side of the equation, but no one looks at the business risk. The business risk of having zero trust in your AI means all the other MSPs can surpass you. You need to balance business risk and cyber risk. It’s never one hundred percent cyber risk, or your business won’t move forward.

How do you put people in the middle? When you tell the AI to build the automation, you require approval gates along the way. And the AI will build in those approval gates for people to get in the way of automation and approve or deny things. That solves the problem of the AI acting fully independently.

Robert Dutt: You’ve announced an MCP server that lets external AI agents trigger Rewst automations directly. That feels like a pretty big architectural shift toward that AI-doing side of things – toward operations that are genuinely agent-driven rather than just automated. What does that look like in practice today, and where do you see that going over the next couple of years?

Aharon Chernin: I see three main agent use cases. But if we sit back and think, “Where does this MCP server fit in?” – in my world, in my vision, when I create products, I have to try to predict the future. Because it takes time. If I create a product for what’s needed right now, by the time it gets done and goes to market, it’s already a bad idea.

In the future, people will be interacting with agents, not with the user interface of any product – a PSA, an RMM, maybe even an operating system. When you think that way, without a capable MCP server in front of your product, you’re forcing people to always log into it. If everyone is in Claude Code all day long and you’re the one product they have to pop out of Claude Code for, that’s not good. I don’t want to be there.

And for folks who want more approval controls in an automation process – what’s cool about using Claude Code to interact with the Rewst platform is you can put requirements in your Claude Code, in your AGENTS.md file, for example, to be enforced however you want them enforced on your side. Which is pretty neat, in my opinion.

As far as the three major use cases for agents: I immediately get brain-fried when someone asks what an agent-driven MSP of the future looks like, because I think – what agent are they talking about? I really see three specific, distinct agent implementations that an MSP could ever have.

One is an agent that they build for their customers that just does customer-centric business things. That’s not something the MSP would run internally – that’s something they want their customer running. Then they may want an agent that the customer uses to interact with the MSP – “Hey, I need to add a new user,” for example. That may launch Rewst automations, because it’s just an agent – it’s just AI, it can’t go do those things on its own. And then the third type is the agent that an engineer uses at the MSP itself, because that agent is going to have more privileged access and communicate in a much more technical way than you’d want to present to a client.

Robert Dutt: How do you see those rolling out in terms of timeline? My first assumption would be that you build the internal side first – get the side the engineers are working with up and running, learn from that, then start building out the customer-facing side, and finally toward the end-user type stuff.

Aharon Chernin: You always eat your own dog food first. It’s got to be useful for you. If you’re just doing it to say you’re doing it, you won’t be able to sustain it. It won’t scale.

I would have your help desk and your engineers doing simple things first. I love the owner-operators that have grand visions of their AI-automated futures, because those are the ones who are going to succeed. The one downfall they may have is they won’t settle for anything other than that grand vision – they work on the most complex thing first. Or they get a ton of small wins under their belt and never go for the big win because the big vision feels too far away.

So: keep it small initially. Eat your own dog food. Get really good at it. Then move bigger successfully internally. Then maybe the next step is the agent the customer uses to interact with your business – do the same thing, start small, then go big. The agent the MSP builds for their customers’ businesses is the toughest. Not because it’s technically challenging – it’s tough because the MSP has to go learn how that customer’s business actually operates. And if you want your agent to provide value, you’re going to have to do that work.

Robert Dutt: That’s a chance to really test the “trusted advisor” theory out and get real about it.

Aharon Chernin: Exactly.

Robert Dutt: Let’s talk about the economics of this. Especially as a Canadian MSP – we’ve got a smaller talent pool, cross-border salary competition, the Canadian dollar. The headcount argument for automation is particularly acute here. Where does break-even actually land? What does an MSP need to have in place before this really starts to visibly move the margin needle?

Aharon Chernin: You start with understanding the economics of automation. The most basic is: how much time will you save? And even the definition of saving time can be less obvious than you’d think, because saving time might mean “I don’t do this work today” – so how do you calculate that? Well, one of those savings is: will it reduce churn? How much time does it take to go get a new client? Then you realize, oh, there is a time savings there. Or it may not save time, but it improves service delivery reliability. How much time do you spend each month on ticket re-dos because they were done wrong the first time? And some of these savings are security-related. If you can take away privileged access from your lowest-level help desk because automation handles those tasks instead, how much does an incident cost you each month?

The next big thing is understanding the cost to build an automation. We’ve dramatically reduced that over the past three months through RoboRewsty – you just describe an automation and it goes and builds it for you, then runs, tests, and fixes it. It could save you roughly ninety percent of the build time. Just like writing code with Claude Code saves you a lot of time building software.

So: you’ve got an automation idea. It costs you four hours to build and saves you 400 hours a month – you break even really, really quickly. You have an automation that takes four hours to build but only saves you thirty minutes a month – do you still build it? Yes. Because automation is not a one-month profit thing. It builds off itself each and every month. And that’s when you start to realize: maybe I do start small, because the big grandiose ideas might save you ten hours a month but take you a hundred hours to build.

We’ve got a couple of Canadian partners who’ve had great success and are okay with me naming them. Resolved IT: last month they did 420,000 tasks and saved fifty-nine hours – 1,800 endpoints. Ideological Systems: 28,000 tasks, forty-five hours saved – 750 endpoints. And Yardstick: 744,000 tasks, 327 hours saved – 6,500 endpoints. Just some numbers for you.

Robert Dutt: That’s certainly some solid data. You’ve talked about where you’re at now with RoboRewsty. Can you share a little about where you see things going in the short term – in terms of the direction of the technology and how you’re automating processes?

Aharon Chernin: I feel like I’m more of a future predictor than a startup founder. AI is both scary and exciting at the same time. It’s a huge opportunity for everybody – for MSPs, for vendors that serve MSPs. But it’s also super scary because it’s a disruptor. It could disrupt MSPs, the vendors that serve them, and their customers’ businesses. You have to take risks in how you predict what’s coming.

But I really do think the user interface for people is going away, and that the primary user of all products will be AI. That’s where the world is going. Think about that – how does that change how an MSP operates? Look at the tool stack MSPs use today. Is your PSA built for people, or is it built for AI? It’s going to be interesting. Never a dull moment in this industry.

Robert Dutt: Last one for me. If I’m an MSP listening to this and being honest with myself about where I’m at in my automation and AI journey – what’s the one thing that most of them should be doing in the next six months that most of them aren’t?

Aharon Chernin: Starting at a really high level: you’re a business owner. I’m a business owner. I understand how hard it is to get out of firefighting mode. It’s hard to see thirty days out. There’s so much going on – happy people that want to talk to you, unhappy people pulling you aside. You have to spend some time thinking beyond thirty days. You can’t ignore what’s coming with AI and automation, because there is going to be an MSP that doesn’t ignore it.

The next level: come up with an automation idea. As the owner of the MSP, that’s the most valuable automation idea that will ever get identified, because it’s coming straight from the business – not from a random engineer with a random problem. That engineer can go automate something and make their job better, but you may never know about it and it may never impact the business. Come up with an automation idea. If you find that easy, come up with more. Having the ideas is half the battle – not even having a product. Just start writing them down.

Next: identify someone in your organization who will become the automation champion. As fast as tech is moving, it may not have to be a super technical person, because they’ll be able to talk to an agent to make the automation happen. But if it were me, I would still pick who I think can automate the best. Go to your team and say, “Who wants to automate?” and pick the hand-raiser. That’ll also help with alignment.

Once all that’s done, you’re finally ready to look at automation solutions. Learn everything you can – we have Cluck U, our certification program, get certified in automation. Then build that simple automation as quickly as possible and get it into production. Do not wait for perfection. If you’ve got a ten-step process that takes a hundred minutes, and you only automate nine steps, you’re still saving ninety minutes. Why wait for the last step to get that win?

And then we’ve got a conference called Flow – an automation-focused conference for MSPs. The coolest thing about it is that everyone there is an MSP that wants to become an automated MSP, not just one that wants to do some automation. This is our third year. First year we had about 120 people, second year over 300, and this year I think we’ll have 500. It’s in Nashville, June 23rd. Check it out at mspflow.events. Was I allowed to plug that?

Robert Dutt: I think you’re good.

Aharon Chernin: Sounds good.

Robert Dutt: Aharon, I appreciate you taking the time and giving us a view of where automation’s at and where it could be going. Thank you.

Aharon Chernin: Thanks – it was nice chatting.

Robert Dutt: There you have it – Aharon Chernin from Rewst. I’d like to thank Aharon for his time today, and thank you for listening.

A couple of things I’m still thinking about from this conversation. The framing that stuck with me most is the distinction between an MSP that does automation and an automated MSP. It sounds like a subtle difference, but Aharon’s point is that one is a project and the other is a culture – and only one of them actually compounds over time. If automation is something your team does when they get around to it, rather than something that’s baked into how the business operates, you’re probably not going to get the return you think you are.

The other thing worth sitting with is the governance question. As AI moves from helping you build workflows to actually generating and running them, the trust model has to evolve with it. The new employee analogy Aharon used – starting with limited access, expanding trust as it earns it – is a practical framework for MSPs who are trying to figure out how much autonomy to give AI in their stack right now.

And for any Canadian MSPs listening – the data points Aharon shared from Resolved IT, from Yardstick, from Ideological Systems – those are real Canadian shops with real numbers. The economics of automation are compelling anywhere, but they’re particularly compelling here, given our labour market realities.

If this episode was useful to you, the best thing you can do is subscribe to or follow the ChannelBuzz.ca podcast wherever you get your podcasts. We’re on Apple Podcasts, Spotify, YouTube, and most major directories. A rating or review goes a long way and is always appreciated.

Until next time, I’m Robert Dutt for ChannelBuzz.ca, and I’ll see you in the channel.