When AI Outpaces Humans: Jason Kemsley on What Happens to MSP Roles as Automation Advances

[00:00:00] Dave Sobel here, reporting from Channel Con, talking with Jason Kemsley. He is the Chief Revenue Officer at Uptime Global. Now Uptime is a channel-only outsourcer. That's white label help desk NOC and dedicated engineers and project teams built for MSPs that's running since 2009. Jason's here to deliver a session called AI Regret, Chasing the Hype is Setting You Back Years. Jason, welcome to the show. Thanks for having me. I've been watching for a long time, so I'm glad our paths finally crossed.

[00:00:30] Well, I'm excited to have you on. So tell me about this session, AI Regret, Chasing the Hype is Setting You Back Years. Give me the argument. What does AI Regret actually look like inside an MSP that move too fast? So I speak from what I see, which puts me in a unique position this time around in that I'm speaking against the grain of the jump on AI and so forth. And one of the things I've really been noticing is a lot of AI adoption, but a lack of AI completion.

[00:01:00] So everyone will start a project, vibe coding, putting in automation, whatever it might be. But then when you ask in three, six, nine months time, oh, where did that get to? I'm seeing a lot almost effectively die and sit in a pile of 80% or 70%. And so what I was thinking about when I was looking at this is our business, the MSPs I've seen, where they get a project to 100% and delivering real benefits.

[00:01:26] It could be money, time, whatever it may be. What have they done that's different to those that have failed? And effectively, I boiled it down to one simple thing. The people that chose to take that project or AI solution on as the starting point usually failed. The people that identified an issue in their business then tied it to a solution that was available typically succeeded because they already had buy-in.

[00:01:54] They had enough of an issue that it would gain traction in the organization. So it was just issue into solution rather than a solution into an issue. Gotcha. So the ability to identify the real problems first, be specific about it and said, oh, now we're willing to solve it with this. That's the key differentiator? Absolutely. Because I think I could walk around any vendor hall and look at something and say, oh, I've got that issue somewhere because everyone's job is to tell me the pain that I might be feeling or generically.

[00:02:21] I want to know what's my biggest top three issues as a business. Is there an AI solution or something tailored that we can go use? If so, great because it's a top three problem. I don't have to get employee buy-in. I don't have to change the culture. I don't have to drive usage because we all want to drive it anyway because it's a problem. Makes sense. So you told us that MSPs are not too far behind and that no one has this nailed.

[00:02:47] You see hundreds of MSP ticket queues with the visibility. How many of your partners have put AI into production in the last 18 months? And what happened to their ticket volume to the ones that did? I would suggest, and I don't know the stats, so I am going to guesstimate, I would say 25% have used a notable amount of AI where it's clearly visible in some way, shape or form.

[00:03:14] And I'm talking beyond triage. I think triage is a fairly mundane task that most people can do without. I would then split those in half. Half of them, I don't think it's having the effect they thought it was going to have. I get to see the tickets where the customer is saying, I just want to speak to a person. I just want to speak to a person. And then the other half have had real noticeable improvement achievement.

[00:03:42] We've seen 10, 20, 25% ticket savings. I'm never happy at a number. Unfortunately, it's just my position. I'm always driving for the next. When I see the 10, 20, 25%, I want to know what types of tickets are those. What impact does that have on the customer? If I've just silenced some things that actually surfaced useful things. So we're in that, what I would call a balancing seesaw moment, where we are getting more efficient.

[00:04:08] And we're seeing it. We've got 10, 15 very case study-esque MSPs that have gone to do things that have had a real good impact. But what effect does that have on the next six months, 12 months? Do their customers enjoy it and retain, keep their contracts? Do they look at other providers? This is the bit I'm waiting to see. So seven weeks ago, Uptime published something more specific than almost anything I've seen in the marketplace before.

[00:04:36] An actual confidence threshold. And above 98%, the AI acts. Between 70 and 95, it flags for a human. Below 70, a person takes it. Where did the 98 come from? I'll condense this because I'm very passionate about it, as I hope comes across. We went on the AI journey or the machine learning journey three years ago. That was when we started.

[00:05:06] All of the owners of Uptime, myself and Brad, we're huge tech lovers. And so it's a passion of ours. We looked at machine learning, AI and so forth. And we thought, right, if we're going to do this, how do I stand in front of my partners, friends as well, and say we've done it in the right way, the ethical way? So we decided to build the model ourselves. So we sat there and we looked at what a human does in all our dispatches, etc. What sort of accuracy percentage are they?

[00:05:32] We found they were between 95% and 96% across all tickets. And we're talking 20, 30, 40,000 as an allocation. And so we said, right, if we are going to put something in, it has to at least drive us a little bit forward. Otherwise, there is no point doing it. Over the past few years, AI has naturally evolved and has got better, which has meant 98% is now the minimum bar that we will accept with any deployment. And we use that ourselves.

[00:06:00] So I'm not sure if it came across in the post, but that's actually what we have set because we want to be at least what a current human is capable of, if not a little bit better. In fact, the post does say your own human engineers run at 98% accuracy. So by your number above that line, the machine is at least as good as a person, right? Yep. So what share of ticket volume already clears that? For every 100, maybe four we flag for review now.

[00:06:30] Right. But we're much deeper into the journey. Sure. Because we're in our own model. I mean, I sat there with 12 engineers one weekend with pizza all around the room. And we were manually looking at Excel spreadsheets to see how they were allocated. Correct, wrong, what would it be? So we've manually trained this to the absolute nth degree because we're quite a niche use case. We work with different MSPs. They all have different bars. We didn't need an off-the-shelf product.

[00:06:57] So I think we're probably a little unique and have had more time. So one of the things we also have built in, which isn't in that post but is a really useful part, is when you do it yourself, anything that gets flagged, you can have a subsequent action, which is go back and re-update what I've learned from what the human told me about that flag. So, right, I've had 100 over the past six months. Now go re-update the model so every quarter a new release can come, and it only gets better and better and better.

[00:07:28] Every tool you bolt onto Microsoft 365 is another console, another login, another thing that needs your team's time. Proofpoint 365 Total Protection gives that time back. Security, backup and compliance for Microsoft 365 in one multi-tenant platform. Built MSP first. Enterprise-grade protection tailored for your small and mid-sized clients with the operational simplicity that protects your margin.

[00:07:58] Modern threats covered, resilience built in, one pane of glass. Built on Hornet security, now part of Proofpoint. Learn more at Proofpoint-Total-Protection.com. If the AI clears more of the queue every quarter, what happens to the number of engineers an MSP needs to buy from you? That has shrunk significantly. We have, in our help desk, and in...

[00:08:27] I can't think of an MSP that has a triage role anymore, a dispatch role off the top of my head. We don't have it anymore. The lines have blurred. We now have what we call first responders. And a first responder role in our organisation is probably the personality that would have been a dispatcher or a triage, but with usually a fundamental level of technical skillset. So they can now be doing 365 GUI changes,

[00:08:57] general portal administration, VoIP portal and so on and so forth. So that has absolutely lines of blood. And no longer do you have what priority is this, what queue should it be and what engineer. Now those people have blurred into... They're actually delivering tickets. For us, and I'm sharing just in case it's helpful for others, they have a cap of 20 minutes. So they look at a ticket, they think, can I solve this or make meaningful progress within 20 minutes? If yes, work on it. If no, let a first line have it.

[00:09:27] And we've found that that is the sweet spot for... It's usually a quick, simple, and I say simple because some things do develop, fix. They then obviously naturally move first line, second line and so forth. So your master services agreement has three ways to charge, right? Hourly, fixed project and dedicated monthly engineers. All three of those are priced off labor. So there's no per ticket price and there's no per outcome price anywhere in that.

[00:09:54] We have per device, per ticket. Okay. Full-time engineers and then on the knock basis per device as well. Okay. The per ticket, to be truthful with you and you're extracting something I usually don't share, but I'm not a big fan of. It creates a transactional relationship. Okay. I'm only ever going to be as good as the last block of tickets.

[00:10:19] It is the gateway for people to seeing we do what we say we're going to do and thus then looking at the other solutions. So in an ideal world, someone would not be on a ticket bundle forever because it is like you're saying, his money, his time, meat in the middle. It's slowly going to rise with the world economy and so forth. And so our goal is to show enough value that you want to look at our other solutions which are priced per device and outcome based. Okay. That makes some sense.

[00:10:47] So when the AI makes your engineers faster, like who captures that? Does a partner's invoice go down or do you keep the margin? How does that efficiency get play out? So it depends on per ticket. They win because obviously they need less tickets from us. Usually they'll use it to then, I've got 100 tickets with uptime. I've saved 20. Now I can put these other 20 over that I maybe had or was considering. Or they can start to reduce. We have in our contracts that you can reduce.

[00:11:18] On the per device, what we're effectively saying is, Dave comes to me. I've got a 15 device customer uptime. Could you deliver the Monday to Friday for me so I can focus on projects, AI, VCO and so forth? We will drive all of the AI as much as we can with you in that whether it's scripting. Obviously, it grows arms and legs and goes in different directions. And we will deploy that in your RMM or in our RMM.

[00:11:47] And the goal being that after 12 months, 24 months, 18 months, whatever the period, we've driven such efficiency as a team in combination with each other that the customer doesn't want to leave you and you don't want to leave me. Because a sideways move to another entity would mean tickets spike again and people become less efficient and so on. This episode is brought to you by Control Map.

[00:12:12] 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. 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.

[00:12:39] Visit ScalePad.com slash Dave to get started. That's scalepad.com slash Dave. Interestingly, the white paper that you write, your tagline is stop adding employees, start adding power. Right? Your product is employees at some level too. So is that a contradiction or is it the pitch? Like, and how are you thinking about the future for uptime in three years of L1 and L2 volume halves?

[00:13:09] What we've always wanted to be is an operational partner. So I work with Dave. I've onboarded Dave's Autotask, ConnectWise, whatever it may be, Halo. We've got the APIs in. We've got accounts. We've agreed what looking at passwords looks like and so forth. You can now choose to add resourcing in whatever capacity per ticket, per employee, knock devices on the back end. Very simply with a couple of days, we can just put them on and off.

[00:13:37] And so when we talk about adding power, we're effectively trying to say whatever scenario you come up against that causes an operational pain, we should have something if we've done our job that fits that commercially and you can very quickly activate it, turn it on. So I can see how the employees ties into that. I don't want to just necessarily focus on, I've got an L1 with you. What else, when you've got a gap that comes up?

[00:14:05] Sometimes we've got a person coming off the contract and a partner said to us just in passing, Hey, I'm in a bit of a hole here. I've had someone walk out. I've got someone that's only available three weeks. Normally we do permanent. Is there a way I can help you? It's just about having that. We all need someone behind the scenes that can just give us a bit of firepower when we need it. And it could be in any area. The second point of your question being what do we see?

[00:14:33] We think we need to be even better than we are is the first point. So from a help desk standpoint, we are expecting very little if any price rises. We think that all of the, I guess, where the market gets eroded and price gets eroded, we're going to make up for with efficiencies and AI and so on and so forth. So we're expecting and hoping that each year we can drive such efficiencies that the price never needs to change again. Certainly for the foreseeable.

[00:15:03] From a Knox perspective, we think that's roughly the same. We're seeing a change in the market. We used to do lots of servers and on-prem. We're seeing more looking after big networks for hospitality and so on. Looking after Azure web apps, that type of thing. From an engineer perspective, it is absolutely not going to be in five years as prosperous as it is now.

[00:15:26] One of the things we've done to try and make sure that our business doesn't erode is we now actually do roles that focus on AI and automation. So we are one of the last to market. I'm happy to be that guy because we have tried to do the very best talent. We've just put someone into an MSP in Ireland, having phenomenal results.

[00:15:49] So we are trying to take the creme de la creme positions, which are fewer, they absolutely are, but better outcomes. And I think that's probably going to be what makes up some of the shortfall on bodies. But I think, and I know this isn't necessarily me asking you, but I think even though we are going to see reduction in workload, good people are still hard to find. And because we sit in the good people section, I don't want to place 20 engineers with a person.

[00:16:16] I want you to place for every three vacancies, consider us for one, and let's put someone excellent alongside your team. We've never been a quantity shop. Well, Jason, you've given me a ton of insight into the way the business works. Really appreciate you joining me today. Thank you very much. What would you fix in your business with the right playbook? Small Biz Thoughts members get a library of templates and operational resources, recorded member calls, and classes through IT Service Provider University.

[00:16:45] Operational education built specifically for independent MSPs. Start at smallbizthoughts.org. 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.

[00:17:15] Produced by Picture This Video. Part of the MSP Radio Network. The Business of Tech The Business of Tech