Atera’s AI Shift: Gil Pekelman on Accountability and Risk in Autonomous IT for MSPs

[00:00:01] AI is moving from suggesting actions to taking them. The managed service providers who handle thousands of endpoints for their clients are about to find out whether that's a feature or a liability. In this episode, we sit down with Gil Pekelman, the CEO and co-founder of Aterra, to talk about what autonomous IT actually looks like in production, not just a demo. He's made a bold commercial commitment.

[00:00:25] Robin, Aterra's autonomous AI agent, will close half of every Tier 1 and complex Tier 2 ticket within 90 days of onboarding before the fees get waived. That's a real performance guarantee in a market full of AI promise. We talk about what it takes to build that kind of confidence, what happens when the MSP community pushes back on the price tag, and where the guardrails are actually when a machine has the key to your client's infrastructure.

[00:00:53] This is the Business of Tech. Gil Pekelman, you are the co-founder and CEO of Atera. Welcome to the Business of Tech. Thank you, and it is great to be here, Dave. Well, I'm excited to talk to you, and I'm going to dive straight in because you recently rebranded your IT Autopilot product as Robin. And you shipped custom AI workflows with playbooks.

[00:01:17] You've got a flexible resolution policy that allows admins to control specifically when Robin escalates versus resolving on its own, and you added multilingual support. So that's a lot coming on all at once. And let's put on our MSP hat, somebody who's running with less than best-in-class margins or running a little leaner than they'd like. What's the concrete operational change that you're focused on with this set of rollouts? Okay.

[00:01:46] So Robin is the only, the world's only, autonomous AI agentic platform, a lot of big words. It is able to solve technical problems, technical incidents, autonomously, both tier one and tier two, complex one.

[00:02:07] So this is an essence, an agent that knows how to solve and do the work of an IT person, and both simple and complex things. The end user actually interacts with it just like he interacts with a human. So if he has a problem, he doesn't need to do anything different than he would do with a human. And it is able to understand his problem.

[00:02:34] It understands who he is, where it networked, what's his computer. It can, it goes using the Atera infrastructure and the old type of agents on his computer and network to do diagnostics. And then understands the problem, looks for the optimal solution and executes it. So it does this whole thing, the mean time to repair.

[00:02:59] So the, the mean time to, to solve this whole cycle is 120 seconds, two minutes, under two minutes, up to two minutes. So this in essence, from the MSP side is a revolution because there's almost no limit to the number of customers you can add on, et cetera. The minute you activate it, you're giving a 24 seven support. It answers in a second and solves in around 120 seconds.

[00:03:29] It's able to solve thousands of different cases, technical all. Well, there's a whole world of non-technical tickets, which it does easily, but the core competence and capability of Robin and a terror is the ability to autonomously solve technical problems. Everything starts with a frustration because you have a problem and you're going to your IT, you're going to your MSP and you're going, you're starting with a problem. And then you wait, you always wait.

[00:03:58] It depends how big your MSP is and how big you are and where you are in the hierarchy of the company, but you wait, just wait for him to get back to you and then wait to fix it. And to fix it, usually to do the diagnostics, you'll need to connect to your computer and check remotely connect and check, understand and try some things.

[00:04:19] So, AI that within a Atera, it's, by the way, patented technology is, is able within seconds to do the diagnostics, within seconds to, to realize, reason what is the best solution and then run it and execute it. So, the user has this very different experience of engaging with IT. It's something that never existed before.

[00:04:44] And we are seeing, actually, this is interesting, a big uptick in tickets because of Robin. So, because people suddenly feel very comfortable to go to it, to fix things, to install things, to get things to work better. You don't feel like you're, you're wasting someone's time or, you know, we always feel like, we're stupid, but if it's AI, you don't have all that.

[00:05:09] Any small thing you want fixed or to run better or smoother or you're not happy with, you go to Robin. And it's always nice, always friendly, always has some sense of humor. And it executes the whole, from answering you through diagnostics to execution and solving the problem under 120 seconds. We are, yeah, good. What I want to understand there, so, you know, understand the way you're executing. Let's talk a little bit about the failure mode.

[00:05:36] And what I mean by that is like the engineering version of that, not the, what I mean is what happens when it isn't able to execute? Like, how does that look like? And I'd also like to get a sense of like, how often does Robin escalate when it probably should have been able to resolve? That gray area, you know, in terms of execution. Okay.

[00:05:56] So, first of all, there's this whole system called playbooks where you, just like an engineer, think of it as an engineer, tier one or tier two, that joins the company. And there's different engineers here that you guide it how you want it to behave. So, for example, there's companies that when it's a security issue, they don't want tier one. Forget Robin for a minute. They don't want tier one to deal with that. They want it always escalated.

[00:06:24] So, you can implement the same type of logic. When it's a security issue, don't even try to, don't even try to deal with it. It can fix like, for example, VPN issues. But let's say decide that in your company, you don't want it to do it, so you can teach it. That every time it's a VPN or other security, escalate to a UI. That's one kind of scenario where you teach it when you want it, just like you would teach a human.

[00:06:53] And you can be more conservative in the beginning, less conservative, etc. Robin also knows when it doesn't know how to solve. It has a reasoning engine. It's a very, very sophisticated AI. And it understands. Sometimes it diagnoses, and then it goes and checks what's the best solution. And it has thousands of available solutions, but they're limited. So, there is a limited number of solutions. You can also control that.

[00:07:23] And if it doesn't have that solution, it knows I can't do it. I understand what's the problem. I look at my available solutions. I don't have that one to solve it. It will escalate. An interesting thing about escalation with Robin, when it escalates, it shows the technician all the tests it's done, the results of the tests, what it diagnoses as the problem, what it diagnoses as the solution.

[00:07:51] And it creates the script for executing the solution with a button. So, also, the time for a technician to solve the problem is condensed by 90% because all the tests that he would have done are already done for him. And the results. The thinking around what that could mean is already laid before. Now, if it's not correct, he can do his own tests. But usually, it'll be correct. So, those are the escalation mechanisms of Robin.

[00:08:20] There's another element that you asked the engineering side, what happens if he should have solved something? So, there's also a mechanism in the system where you go over time. You don't install it. Maybe I'll step back for a minute. You don't install it like software. You don't onboard it like software. You onboard it like a human engineer. And it's the same if you would bring a new engineer and he would do a few things and then there's something that you expected him to do and he didn't do and he escalated it.

[00:08:49] For example, you would go back and sit with him and understand why he escalated that and maybe he needed some. So, it's exactly the same thing. You go over the tickets periodically and you can see exactly what Robin solved and what he didn't solve. And you look at that and you say, but I want him to solve those things. Why doesn't he do it? And then you can teach him to do it either by creating the solution or a playbook, etc.

[00:09:17] Implied in your answer there is a management process of Robin. And I want to make that explicit. Like how is Atera recommending that someone inside the practice owns Robin? Like either from a configuration perspective but also from an onboarding human. I'm putting that kind of in air quotes like treating him like an engineer. Like how do you own that? What's the recommendation?

[00:09:43] And how much time does it take to invest in the training? And I mean that in the almost human sense of investing in that engineer. Okay. So, first of all, we have two commitments, two guarantees that we give. One is that within 72 hours we can show you a POC of exactly cases that you'll say, I want to see Robin solve one, two, three, and five.

[00:10:11] We guarantee that we are able to demonstrate that within 72 hours from start to finish. And it really is pretty flexible what you want to see. And we guarantee that within 90 days from starting onboarding to 90 days, 50% of all your tickets will be handled by Robin. You won't deal with them anymore. They will be handed autonomously. Now, Robin as a system has a backend.

[00:10:40] And the backend has these playbooks and analysis of what it should solve. It also learns. Once you install it, it learns as it works on all kinds of things. It sees that your technicians are solving all kinds of things. And it says, I can solve that. I can do that instead of them. Do you want me to do it? Here's the script that I'll use. Let me just approve it. So, there is a system behind it that enables you to do that.

[00:11:09] And we train on that. We train you how to do that. And we work with you. The onboarding, the 90 days onboarding is we work with you there. We don't let you do it on your own. It is a new world. It is a new world. It's different. There is an expectation. We just talked about it as a person. It is. The closest thing it is is to a person. It's much less close to a software and SaaS platform than it is to a person.

[00:11:37] And you can do something with it and say, oh, why didn't it solve it? But then you can train it and teach it. We'll be right back after this message. This episode is brought to you by Control Map. Growing MSPs are using Control Map to build recurring revenue by expanding their GRC services. Starting now, Control Map is offering a free plan for MSPs looking to get started with providing compliance as a service. Create a free account and run an assessment.

[00:12:06] Track key items like policies, risks, and evidence in one place. It's a practical way to prove value to a client before deciding to expand your compliance offering. Try Control Map for free today. Visit scalepad.com slash Dave to get started. That's scalepad.com slash Dave. Within MSP, what role, position, title, what human role do you anticipate is the one managing Robin?

[00:12:36] There needs to be one person to do that in the company. And he needs to get the goals and KPIs, the 50%. Or we have customers that 80% is what they... And he's the one person that he manages the AI system. And he will work on that. We have customers that went down 90% in the headcount.

[00:13:06] And they don't... These people usually go to do other things. But they have one person for every 1,000 employees or endpoints. And then there's one person on Robin. So the architecture of the IT department is very different suddenly. For every 1,000 people, you have a human. But then you have one human that manages Robin across all these devices and customers.

[00:13:32] And the total number of people is about a tenth of what you would use today. And it answers in a second and solves in under two minutes, etc. So... Is a pattern emerging on who that human is? Like if they're... Are they the head of engineering? Are they the owner? Owner? Is there like a pattern on how that's working for the most successful MSBs? It's usually somebody that's in a sense dedicated to it.

[00:14:02] So they live it. And so it's not the owner or head of engineering. It'll be somebody more a system engineer or somebody that knows the system and understands it more hands-on. I just had something a week ago. So I was at a conference and in the morning, I had to read a contract. And I opened... I installed something. I opened the Word file.

[00:14:30] And I didn't have a lot of time before I had to go to the conference. I opened the Word file and I got this error message. I can't even repeat it. I don't remember that. And I looked at it. And the first thing, the first reaction is, okay, now I can't read the contract. So I'm going to have to do it later in the day. And my whole plan is now disrupted. The second reaction is I have to engage with IT now, which is going to be a hassle. But there is Robin in Atera.

[00:14:56] And all I did was copy-paste the error message to Robin and said, fix this for me, please. And two minutes later, it was fixed. And it actually was an interesting case because it had to go into my computer, go into Word, find out that there's a plugin, an Adobe plugin into the Word. I didn't even know I had that. It was not behaving properly. It uninstalled it, reinstalled it, and solved the problem. And for me, I didn't talk to anybody. I didn't engage.

[00:15:26] 120 seconds later, I was back in the flow. I never actually lost the flow. It was so fast. So there is also this impact on you as an MSP and what happens to your customers. It's not the old days of chatbots where it didn't understand you and went back and forth and clicked this. And that in itself was frustrating. This is an amazing, amazing, cutting-edge AI experience.

[00:15:54] Now, you've actually commercialized your guarantee around this, calling it the autonomy guarantee, right? And so what I'm really curious is how you're measuring Robin's capabilities on the back end that makes you that confident. But I also would like to also know, like, what does the kind of a failed guarantee look like? You know, because you would be able to measure those outer cases. Like, what do both sides of that look like? So I hope I understood.

[00:16:20] The way we do it, the process is we ask you, after signing an NPA, we ask you to give us six months back of your tickets. And then we take these tickets and we actually give it to Robin. And Robin tells us each ticket if he's able to autonomously solve it or not. In some cases, the numbers are very high, higher than 50%. But we then take that and we come back to you and we show it to you and we say,

[00:16:50] look, these are your tickets. You gave it to us. Here are the tickets we are going to solve for you autonomously. You'll never see them again. And then when we implement, we work against that. And many times you'll see 80% in that list. So we'll commit to the 50%. But with consistent work, you'll get higher than 50%. We, by the way, started working on this in 2017. So the confidence is because we've been working on this for 10 years.

[00:17:17] In 2022, in early 2022, we started working with OpenAI and Microsoft. We are partners on what was at the time GPT 2.5 private beta. So since early 2022, we've been working with that AI. And it's been in the field almost two years in design partners. So we know exactly what we're able to accomplish.

[00:17:43] So when we get the tickets, we actually are able to come back to you and even give you a bigger guarantee than just the general 50% because we really sit together on the same tickets. And that's what you have. That will be very, very predictive of what you're going to get also six months ahead. So gotcha. Now, Atera raised $102 million. And if I look at your 2026 moves, things like autonomy day, your investment with service now,

[00:18:12] your let people work enterprise campaign targeting larger enterprise organizations, I could read that as an enterprise pivot. What is the story for that mid-market MSP or a smaller MSP shop who's looking for reliability and stability in a partner? How should they read that story of your moves in 2026? We are going to be the next Microsoft. I mean, so we have different segments.

[00:18:42] The real answer is, first of all, Ateris started as a pure MSP company. We have 13,000 customers in total. Half of them are MSPs. And we have different business lines here. We have a business line for MSPs. And that team has their sales rep and support and customer success and training. And we have a business line for mid-market companies. And we have a business line for enterprises.

[00:19:12] And we have all three segments in our company. And each one are very different. And MSP is very, very different than an enterprise company. And they are handled different. It's a different product. Some of the pricing is different. The training is different. It's an owner. It's his business. Yes. But the interesting thing there with MSPs is whatever we do immediately goes to his bottom line.

[00:19:38] It goes to the bottom line on the expenses, on the operational expenses side. And it goes to the bottom line because the level of service is enhanced dramatically. So he becomes more competitive. His customers suddenly are getting this amazing experience. And it answers in a second and et cetera, et cetera. So they start telling their friends, look, this guy has some AI technology that is really amazing.

[00:20:06] So it enhances him both on the expenses side but also on the growth. And MSPs are different. It's not the same needs and processes and things that they care about are different. Now, I would agree with you that the needs are different. And it's curious to see because the other thing I wanted to ask you about is in May, many of the MSPs were pushed back pretty hard on your pricing change. You made a price change that tied to mandatory AI bundling.

[00:20:35] And the feedback that really intrigued me here was some of the MSPs said that they were already paying for what they considered better AI separately and they didn't want it forced in. Like, how do you read that interpretation, particularly for an MSP who wants to build his own toolkit? You know, your choice to bundle it does make it a bit of more of a forced requirement. What's your response to that pushback? Okay.

[00:21:00] The actual thinking behind it is that there is no more RMM and PSA of old. It doesn't exist anymore or shouldn't exist. And what we did with the product is we've changed that into an AI product. It's a very different product. It's not actually bundling AI into an RMM and PSA.

[00:21:27] It is a whole different experience of RMM and PSA. I'll give you an example. You can vibe code scripts and execute them. You can create agents that will do things for you over the day or inform you on things. You can work with the co-pilot. You can ask it questions and then ask it to do things.

[00:21:49] So, the whole RMM, PSA construct of old, of SaaS, that I had to do things and I had to go into menus and everything was humans pushing buttons. We flipped it now. It's not that we took something and really bundled it. We just changed the whole product to one that AI is driving everything now. And the economics there are also in it.

[00:22:14] The amount of time you save and the speed you work and the level of service just by that part of the product changes dramatically as well. That's without Robin. Robin is a separate entity in the whole package. So, Robin really does the work of an IT professional. And then there's the platform for the IT professionals, which suddenly became an AI platform. So, it's not just taking something. And that doesn't exist.

[00:22:42] You can't really build it together. It's all within a Terra. You need it to work within your ticketing system, your RMM. It's not really something that – there's a lot of tools there. For example, every remote connection that one of your technicians or yourself executes its video, it automatically creates a log from it with everything you did there.

[00:23:07] Yet then later on, both you can check if something happened or for regulatory and compliance, et cetera. So, there's a lot with – it's all become AI. It's not taking one AI and bundling it and raising prices. And the percentage of customers that were already paying for AI was very high in any case. So, they didn't pay more, actually. Whoever was already using AI didn't pay more here. We'll be right back after this message.

[00:23:36] If you're delivering Microsoft Cloud services, Nerdio is a company you should know. Nerdio builds software that helps manage service providers deploy and manage Microsoft Cloud environments more efficiently. That includes things like Azure Virtual Desktop, Microsoft 365, and Microsoft Intune. What Nerdio focuses on is automating the infrastructure work, managing multiple tenants, provisioning environments, managing policies, and optimizing Azure costs.

[00:24:03] So, MSPs can run Microsoft Cloud services without the operational overhead that usually comes with them. Instead of building and maintaining those systems manually, Nerdio provides a platform designed specifically for MSP operations. If Microsoft Cloud is part of your services strategy, Nerdio is worth a look. Learn more at GetNerdio.com. And we're back.

[00:24:29] Now, the other thing that I think is an interesting way to look at this is there's a case to be made that autonomous IT is the wrong direction for MSPs. That MSPs' intrinsic value that they're selling is human judgment and the relationships that come along with it. And if you automate that, it starts to erode both the margin and client trust. Like, how do you respond to that? And where do you think the argument is right? And where do you think the argument is wrong?

[00:24:57] It's already been proven that there's things that AI, Claude and OpenAI, et cetera, and other areas can do that. It does much better. It makes us much better. It makes us as people, as employees, as professionals, much better. It can do things that we couldn't dream of doing before. And it does it much faster, et cetera. The same thing actually applies in the world of IT support and IT management.

[00:25:26] There are things that AI can't do. AI can't go to the customer like the MSP and sit down with him and understand his business and then come up with ideas. He can go back to Claude maybe, but come up with ideas on how he should be working and how his infrastructure should work and what should the security tools be there, et cetera.

[00:25:50] So there's many things that are also the more interesting things in IT that AI won't do and doesn't replace the human interaction there. But there's a lot that it does, and it actually does better than us. It's faster. It's more accurate and can go through a lot more data than we can go through and come up with solutions. It's actually a positive all around.

[00:26:18] There is an element that the world is seeing in every area of AI that there's jobs that are being eliminated in a sense. There is a huge, huge deficit in IT professionals in the world. So I don't see any danger there.

[00:26:39] And the majority of what we see, it sounds like, is that people that stop doing the work that Robin does, which is 50%, 60% of all the tickets, start doing much more valuable things. Infrastructure, security, customer relations, understanding your customer's business. You don't have time to do that today.

[00:27:02] Suddenly, you do have time to sit down with them, understand their business, and come up with solutions that are much better tailored to them, et cetera. There's something that I say all the time. The current version is the worst version we'll ever have. We release new features every week, more or less. And every time we do that, it gets better and better. True. But we also have to understand that even if the current version is doing damage, that's to be tackled.

[00:27:29] I'm a little actually more thinking about some of the studies that were starting to come out now about the degradation of human thinking. There are some studies that are starting to show that people that lean too far into this and rely on the tools too much start actually seeing degradation in their own thinking capabilities. That piece is the bit that I'm kind of focused on because an over-reliance on the technology does lose that ability to do human judgment and build those good relationships.

[00:27:58] How are you thinking about that problem in the sense of like the way MSPs would be thinking about interacting with their customers? In our world, the minute I free you up from X work, you have other work to do that inherently usually is much more complex. You're going to use your mind much more. You're going to have to learn more. Learning is what keeps you sharp.

[00:28:25] So if you suddenly have time to understand your customer and you suddenly have the real time to understand what his security posture should look like, and then you have time to go out and study what are the best technologies to do that, your mind is in a very – that's our world, the MSP IT world. In other worlds, it might be degradating their thinking and they're becoming lazy and they don't – but in our world, it's not the case.

[00:28:54] We know that also in our world, there's like constant pressure. We don't have enough people and enough time. And so suddenly this whole paradigm changes with Robin and Atera. And the counterargument is, is that you build great engineers by giving them basic problems and they work their way up by learning on simple problems.

[00:29:13] If Robin takes away that learning opportunity for the next generation of engineers, how do we create the thinking process that makes them great tier three and great thinkers if they don't have the opportunity to work on the smaller problems? You have a junior programmer and he's not going to program now. He's not going to write a line of code. His job is changing. How do you train him? So you're going to have to train him.

[00:29:43] You're going to have to give him – you can give him tickets. You can – you won't have to do it for a very long time. You'll only have to do it until he's trained. And from that point on, you'll move him upwards in the food chain. But you'll have to do that. You're going to – the world has changed in that sense. You're going to have to give him, if he's a programmer, to program things and do work and feel how it feels.

[00:30:05] But at a certain point, he'll be skilled enough that he'll be able to manage 20 agents that are writing code for him on 20 different projects simultaneously. So there is going to be a part in how I bring in young people and train them that is thought about. You need to be – think about it. You need to plan it. It's not like you do it – you did it before or you still do it today. You just give him the work and he starts taking care of tickets and he makes mistakes. It's going to be thoughtful.

[00:30:34] But it's – you're able to do it. It's still there. How do you want the MSP community to think about Atera differently? Because every competitor is making their version of an autonomous AI claim. Like, what do you think is the functional difference or the key differentiator between what you're doing with Robin with all the other claims that are out there?

[00:30:56] There is no other company vendor that has an agent that is able to autonomously solve technical tier one and tier two tickets on its own, autonomously. That doesn't exist in any other vendor. If somebody claims it, ask them to show it to you. We guarantee that within 90 days, 50% of your tickets will be handled by it. It's a guarantee we give. So that's how to think about it. It's technology that's patented.

[00:31:27] It is a breakthrough. We are the only ones. Even if you look upmarket in the enterprise world, ServiceNow, et cetera, they don't have this capability either. Atera is the only company in the world today that has it. Well, Gil, I really do appreciate you taking the time to answer my hard questions directly. If people are interested in reaching out, continuing the conversation, what's the best way for them to do so?

[00:31:48] So first of all, in our website, you can sign up for a demo and a trial and you can start using the product. If anybody wants to send me a personal email, my email is gil at atera.com. Also feel free to do that. There's nothing more that I love more than talking directly to MSPs and working with them and showing them.

[00:32:12] And having MSPs become better and more profitable and more competitive is what we live for. Well, that's certainly what we're all here for. Gil, really appreciate the time today. Thank you, Dave. It was a pleasure. Want more from the Business of Tech? Join Business of Tech Plus for ad-free episodes, early interviews, extended cuts, subscriber-only shows, and exclusive member perks and analysis. Sign up at businessof.tech slash plus.

[00:32:42] And follow this show on your podcast app. And if you're on YouTube, hit subscribe and the bell so you never miss a story. Reviews and comments help spread the word too. Interested in advertising? Head to mspradio.com slash engage. The Business of Tech is written and produced by me, Dave Sobel, under ethics guidelines posted at businessof.tech. Thanks for listening. I'll see you on the next episode.

[00:33:18] Part of the MSP Radio Network.