Reducing IT Firefighting With ControlUp

[00:00:00] The leading issue of agentic AI in businesses right now is ensuring agents act with compliance guidelines. And Denodo applies guardrails across your entire data estate. By aligning your company's data infrastructure under one system, these guardrails perform consistently across your platform. So start scaling your business and start with Denodo. Simply visit denodo.com to learn more.

[00:00:27] What would change for your business if your IT department could resolve endpoint problems before employees even had a chance to report them? Well, for many teams, the working day still involves moving between dashboards, investigating alerts, passing support tickets to colleagues, all with different levels of expertise.

[00:00:53] But the question is whether AI can help complete all that work reliably with controls and evidence that the problem has been resolved. Well, today I welcome Jed Ayres and the CEO of ControlUp and Gaddy Feldman, the Chief Product Officer, to join me here on Tech Talks Daily. Jed will discuss the business cases for autonomous IT and what it could mean for employees and support teams.

[00:01:22] While Gaddy, he will help explain the technology, including specialist agents, connections to existing systems and the company procedures that these agents need to understand. I will also discuss a very real example to bring all that to life. Talk about human approval and ways to assess value across support time, employee disruption and device performance. But enough scene setting for me.

[00:01:50] Please join me in welcoming not one but two guests here on Tech Talks Daily. So a big thank you for joining me on the show today. We've got not one but two guests. Jed, a repeat visit for you. But can you just remind everyone listening with a little about who you are and what you do? Sure, Neil. Great to be on again with you. Thanks so much. I'm Jed Ayres. I'm the CEO of ControlUp. I've been here for about three and a half years, but I spent the better part of the last 30 years

[00:02:18] in enterprise technology and most of it focused on the end user compute sector and kind of focused on how people experience technology through work. Great to be on again. It's a pleasure to have you join me. Of course, you're talking to me from San Francisco today, but I'm now going to beam everyone listening over to New Jersey where Gaddy's joining us as well. Can you tell everyone listening a little about you too, my friend? I'm Gaddy.

[00:02:42] I'm the CPO of ControlUp and the company for quite a while and a lot of experience in the IT space. From the beginning, like been through different companies like HP, Citrus, Google, and the last couple of years I'm responsible for managing the engineering and product of ControlUp. Awesome. Well, thank you both for taking the time to sit down with me today. There's a lot I want to talk about.

[00:03:06] There's a lot of hype and buzzwords out there at the moment around AI and autonomous IT. So, Jed, just to set the scene today, when we talk about the move towards autonomous IT, what does that actually mean for the typical enterprise listening, wherever anybody is listening in the world today? And how far can IT operations realistically run without someone constantly watching and intervening? Yeah. Well, I think in the enterprise today, this is the holy grail, right?

[00:03:36] Of kind of taking an operational problem all the way through its life cycle. And a lot of what has happened in the past has been more around observing it or detecting it and lots of great dashboards. And I think now where we're headed is the ability to take that data set intersected with AI and, you know, see what's causing it, take an authorized action, and then most importantly, also verify that the problem has gone away.

[00:04:05] And, of course, that last step matters enormously, you know, because running a script doesn't necessarily mean that it fixed anything, right? So, yeah, I think this is an exciting part of why Gotti and I get out of bed with a spring in our step these days is that, yeah, we're seeing,

[00:04:24] I would say in the next four or five quarters, you're going to see, you know, very big movement and what, you know, for the last three, four decades has been kind of a cycle where, you know, the same problems are solved over and over again with a lot of, a lot of tooling, a lot of tickets and a lot of human intervention. And I think a bunch of that is set to be wiped away.

[00:04:50] And that, I think, is what's getting a lot of people excited that, you know, we can completely rewrite how we manage devices and really improve the life of both the employee and IT. That will be music to the ears of a lot of people listening. And, Gaddy, to bring you in here, endpoint teams have already had enormous amounts of monitoring of data. So many different dashboards. Alert fatigue is very real.

[00:05:19] So, is the real problem now less about visibility and more about turning all that information into decisions and actions before employees even experience a problem? Yeah, exactly. This is what we aim to do eventually. Because, like, currently, like, thinking about that, actually, what Jadis said, that this IT ecosystem has never been changed for the last 25 years, right? It was the same thing, same tools usually, same, like, structure in the teams. What you're trying to do is actually finally disrupt it.

[00:05:48] And we want to, like, to, finally, we're actually in the beginning of control, we try to do it always with kind of, like, different, like, manual work or, like, some of our automations as well. But, finally, we have, like, the engine, the AI agents that actually were able to do it, to actually run the IT autonomously or almost all autonomously. To get all the data, to create this whole connectors to different, like, three-party solutions. Get all the data across our agents, across other products.

[00:06:16] And actually take decisions and take them actions to adjust the environment before users start feeling the issues. So, yeah, this is exactly the case. And, Neil, I would just add one thing to that that I think is important for us. Our origin story came from the VDI world, as you know, where we were trying to solve one of the largest and hardest problems in end-user compute around virtual desktop and the observability there. And a lot of those users are front-line users, right?

[00:06:42] They're in a healthcare setting or a retail setting or a manufacturing setting, and they actually don't have the opportunity to even open a trouble ticket, right? And so that's where we also see kind of profound impact here, right? Because those environments, many times the experience just degrades and they kind of put up with it. And it affects a lot of things downstream.

[00:07:04] And so that's where we also get quite excited that this technology can do an even better job of kind of proactively seeing those environments where the end-user just can't open a ticket and play ping-pong all day trying to solve it. Yeah, and I do suspect we have many business leaders and indeed techies listening.

[00:07:26] They've seen IT teams spend huge amounts of time firefighting reoccurring issues across endpoints, AVD and EUC environments, of course. And maybe they're thinking, you know what, I've seen this movie before. We've been promised this utopia before. So to bring this to life, what we're talking about here, are there any examples that you're seeing where AI and automation is already identifying, diagnosing,

[00:07:51] and remediating some of these problems we're talking about without waiting for yet another support ticket? So I can give you some examples. For example, we have this concept of skilled agents. So skilled agents are actually kind of the replica of SMEs in the IT team, like skilled agents for Intune, skilled agents for networks, skilled agents for security and so on, storage and so on, so on. These skilled agents actually can work together to understand or manage or fix an issue.

[00:08:17] And then, for example, we can put the skilled agent in front of the eldest tickets or just monitor your environment. But let's take the eldest tickets, for example. You can just listen for each eldest ticket, go to the ticket, understand the ticket. If needy, you're going to ask the storage guy, the storage AI agent, analyze something or ask the vSphere skilled agent, analyze something else.

[00:08:40] Then combine this oldest analysis together and provide with a solution or just kind of analysis of the ticket. And then we can add it as a comment to the ticket or just solve the issue by itself. So it's something that before then, it's going to be maybe work of a week or a couple of days between moving these tickets between different teams. Now it's just a matter of minutes. So this is just a really quick example of how it's going to be changed.

[00:09:10] I guess I would add one to Neil. Like, yeah, we look at like the data set that we can collect for our customers. It's it's tremendous. Huge. Right. And now with AI kind of working at machine speed, looking across these environments and doing it. Yeah. At a very fast speed, we can see patterns, distinct patterns that are causing problems.

[00:09:33] I would say there's a good one on our website for Christus Health, where our pulse insights, your pulse AI insights identified an audio driver problem that was affecting 900 different devices. Right. That, you know, in a hospital setting where they were having problems with the dictation. And so it sees that. And so instead of, you know, people having to sort of call in individually and, you know, solve these device problems, you know, one at a time, we see it holistically.

[00:10:02] We know the remedy. We update all those drivers and, you know, hopefully also do it before it becomes a real issue across the organization. And I guess, Jed, that there's also an interesting workforce question here. If autonomous endpoint management, if that continues to succeed in removing much of that repetitive operational work, how does that role of the IT professional change?

[00:10:28] Where should organizations redirect that time and expertise that they get back? I know for many years, IT departments have been looking to be more of a value add department, maybe get involved in project resources, et cetera, and adding value around the business. But how do you see this playing out? Yeah, well, I think I just spent the weekend up in Napa with about 30 CIOs and it's the same old reframe, right? More, more, do more with less, right?

[00:10:56] And of course, now that's even more extreme with AI and the budgets that are needed for AI and then the security and governance that is needed for AI and the board level mandate that, hey, AI could, you know, help us get ahead of our competitors, right? Or if we don't do it, we'll slip behind them, right? So I think there's more pressure than ever on IT right now.

[00:11:21] And so when you think about the budget that, you know, is applied to sort of the mundane, you know, that we've been talking about, right? This is an area where if you can solve for this, right?

[00:11:34] If you can have L1 tech help desk, you know, solving what used to go to two or three, this allows the, you know, the sort of intelligence level and the sort of effort on strategic projects to kind of redirect is what I would say. And that's very important, right? As things are moving faster. And I kind of call it Groundhog Day, Neil, right? Where it's like, we've been in this sort of like loop.

[00:12:02] We walk into an organization right now and we can take 30 to 50% of the help desk tickets out immediately. Just like they're gone, right? And then you're taking the ones that are left and you're allowing the guy that might have just been triaging in the past where he was just getting kind of baseline information. And most of those were going to the more expensive two and three level to get solved.

[00:12:26] Our AI is allowing them to kind of, you know, have the full intelligence of everything that has ever been solved in the universe of that environment. Plus the rest of our millions of endpoints, right? So that's where I think the world changes dramatically, right? It's like people get worried. Oh, this is going to, we're going to lay off lots of people. I see it the other way around.

[00:12:48] Actually, we're going to allow people to sort of stop digging trenches and do much more high value work. Yeah. And there is never a shortage of projects there. And I think the days of IT saying, well, we can get that project in 12 to 18 months. We'll get it into our pipeline. And those days are gone too, right? Because so much changes now every 12 months. You've got to move so much faster, right? Exactly. Yeah.

[00:13:16] The speed is the number one thing here. And yeah, that's where I think our origin story around the data and God, he's been here from the beginning of time. But the fact that we are trying to solve this really hard problem of observability and VDI, it meant we had to have a very wide data set, about 10,000 metrics that we're collecting every three seconds. So very wide, very fast.

[00:13:42] And now as we enter into this AI era, it turns out that data set has become a real competitive advantage for us. Yeah. And Gadi, to bring you back in here from a tech perspective, especially there'll be many techies listening to you saying, yes, but autonomy can introduce additional risks.

[00:14:02] So how do you decide which actions an AI system should be allowed to take independently, which require human approval, and ultimately how organizations can maintain accountability when machines start making operational decisions? And maybe something that might cause a surprise further on down the line. How do you handle that? So you're right. You're totally right. So because of that, we're adding a lot of different layers of the guide rails in our AI agents.

[00:14:27] So for example, each one of them running into that kind of a containerized environment that we actually control what it can be reading outside and getting inside the container and so on. We have a human in the loop. So each agent will be able to work with specific setting of human in the loop. So when it's going to be approved by an admin, when it's not going to be approved by admin. And we're going to add different connections and different layers of connections. So what is risky MCP? What is not risky MCP?

[00:14:56] So we're adding different layers of responsibility and there's some guide rails to control these agents. But overall, I think it's also kind of a journey that we need to go together with the customers, because some of them are still very afraid of it. They're kind of like coming to this position. Yeah, I'm not sure. Maybe whatever. Or I not want to touch it at all. Something like that. It's a journey. They need to trust it more. In the beginning, it's going to be mainly like analysis and maybe insights or like some recommendation.

[00:15:25] But going forward, I think like they're going to trust it more and like allow it to do more actual work. But yeah, we will provide whatever guide rails and whatever security layers we will need to make sure that actually not doing any harm to this environment. Yeah. So many great points there. And I guess another danger with automation is simply automating a bad process faster.

[00:15:48] So for anyone listening, before an organization moves towards autonomous IT, what needs to be in place across their data, their workflows, policies, and all that organizational readiness for automation to genuinely improve operations rather than just create new problems? And that old garbage in, garbage out problem that we've also seen many times. I think it's mainly about connectors, by the way.

[00:16:10] So this is why we're spending a lot of effort to create a different connector to different systems because eventually we're not going to replace any kind of IT system because there are hundreds or thousands of them. And like a customer, a usual environment using kind of like 30, 20 to 30 different tools in these environments. So we're not going to replace Intune. We're not going to replace like Ruster or something like that. It's all about connectors.

[00:16:33] So whatever we will be able to get the data from this like Serapathy solution to our engine, to our brain and understanding and understand how to read it. This is the story of like how we will be able to bring more autonomous solution to this environment. Without connectors, it's kind of like, yeah, it's kind of the same thing, right? You see what you see eventually. And to allow this kind of very smart engines to make analysis and make actual work, you need these connectors.

[00:17:04] So this is the first step. The second step, again, that creates kind of like policies to what the agents need to do, like kind of instructions, provide all the knowledge base you need to be familiar with. For example, it's not just a connection to CrowdStack, but you need to actually understand who are your VIPs, for example, or what is your policy? What are like your procedures you need to follow in the organization? And it's also kind of like a step that you need to take.

[00:17:33] I kind of usually kind of like trying to create a paralysis in between like agents and IT workflows. So whenever you're building your IT team, you're usually looking for someone responsible to be responsible for your security environment, right? So you're looking for something in the wild, people with some experience with Cisco or WEEZ or whatever, kind of understanding with the security.

[00:17:59] But they come with the knowledge, but you still need to teach them your procedures, your actual day-to-day process of your environment. And it's kind of the same with the agents. So the agents, the brain or their experience coming from the LLMA models, but all your currently like processes or your procedure inside the company come in kind of like of knowledge base or knowledge you're adding on top of the agents and the instructions. So kind of the same.

[00:18:27] It's like you're getting the basic of knowledge coming from the experience or whatever from the wild. And then you're putting on top like your specific procedure for the company. So without it, you wouldn't get a lot of like value from these agents without adding this kind of like layer of like specific company procedures. So many great points.

[00:18:48] And Jen, I'm curious, when you're speaking with CIOs, what are the kind of business outcomes that are convincing them that autonomous IT is worth pursuing? Are they instantly looking at lower support costs, employee experience, resilience, productivity? Are we still starting to see completely different measures of success? What are they looking for? What are they measuring here?

[00:19:11] Well, it's interesting because, you know, we just spent the last three years kind of making nice progress as a category leader and digital employee experience. And this is a very expansive, almost discipline, if you will, inside of IT that we've gotten very good at walking into the C-suite and sharing with them the value. But the challenge with DEX was that it was so expansive. There were sort of like a lot of different tracks you could take.

[00:19:39] But I think the nice thing with autonomous, you know, IT operations and this platform we're building is that, you know, it really is very clear. And there's some clear starting points right around the ability to lower support costs, doing it through all the things we've just been talking about, right? Fewer tickets, fewer people working tickets, way more of them just being automated out.

[00:20:00] So, and then, you know, this has been a challenge across the board is the opportunity to show how much more productive capacity the business gets back, not only in IT, but also at the front line, right? The actual end user. These can be a little softer, but I think the more sophisticated organizations are figuring out how to sort of, you know, run a clock on both of these times, right? The time that IT spends fixing it and the time that the employee spends waiting or working around it.

[00:20:29] So I think a strong business case for the C-suite measures both. I will say as much as Gotti said, you know, we're not in the business of, you know, outright replacing tools. And he talked about all the, you know, integrations we have. I think there's 87 of them now and more coming. There, you know, where we're almost just creating this massive real-time data lake, right? With the ability to do, you know, automations off of all that data and workflows.

[00:20:58] But I will tell you, there is a lot of bespoke tooling that we find in organizations. And our ability to take those out actually is quite meaningful. I will tell you one other big one that's quite topical right now is device lifecycle management is another big ROI on these platforms. Because the same precious resources that people are trying to use to build these data centers for AI is the same ones that go into the endpoint, right?

[00:21:26] And Gartner just published something last week that says between now and 2030, you can expect the prices are going to rise by 50%. And the ability to get these devices is challenged. And so we're seeing for the first time in my career, organizations are looking at our performance data instead of the warranty, right?

[00:21:47] Where you used to have sort of like a third of the or fourth of the enterprise fleet of endpoints would move off the, you know, out of the organization every year. We're now seeing, hey, we can save millions of dollars by just keeping these, you know, well-functioning devices in service. So that's a very interesting use case that drives ROI. It's also right alongside that is software reclamation. But the hardware one is especially topical right now.

[00:22:16] If you've talked to anybody that's tried to buy fleets of endpoints, they're very frustrated, both in the availability and the volatility and increased price. So this is kind of an exciting use case that we're seeing. Awesome. And one thing I always try and do at the end of every podcast here is give people listening a valuable takeaway. I'll throw this question to both of you. I'll send it to you first, Jed.

[00:22:42] But for an IT leader that could be listening today, maybe they're spending too much time reacting to endpoint problems. They want to move to a more autonomous model. Where would you recommend they actually start? What is one workload or process, that low-hanging fruit that might be able to prove value without trying to automate everything at once? Just start at the beginning and then keep moving along slightly. But where should they start if they're a little bit overwhelmed right now? Yeah. Yeah.

[00:23:07] Well, I think one of the things we have, we have something we're calling the seven day challenge where we can come in and we can in seven days, put down our agents and kind of come back with some very clear use cases across ticket reduction or across the software and hardware pieces that I just talked through. And just sort of delivering some insights on where we think, you know, from this baseline, where we can derive value.

[00:23:37] And it's pretty profound, you know, a lot of things we're done in seven days. We can do some pretty crazy things in seven days here at Control Up. So I think that's one would be my advice is, you know, take our seven day challenge and let us sort of give you a readout on what our data provides. Awesome. And just to bring you in as well, Gardy there, you must get to see and hear a lot of stories.

[00:24:03] And when it comes to that seven day challenge, anything you've learned or anything you'd pass on to people listening and wanting to get started? I mainly, like you usually said, try to understand before what you have. Because currently, like I think like a lot of people are missing the picture of what they currently have in the fleet. For example, they're not sure about like how AI agents have been utilized by their end users or what they're using, which MCPs, which models, which data insurance on.

[00:24:31] And something like just understanding is a starting point that will lead to them to actually take another step to take more knowledgeable or more like a more smart decision about how we want to continue with that. Beautiful. I think it's a great moment to end on. And for anyone listening that would like to find out more information, I will add a link to both of your LinkedIn profiles. And, Gaddy, I owe you an apology, actually.

[00:25:00] I introduced Jed as someone I've spoken to before. I don't know if you remember this, but we have spoken to, I think it was about five or even six years ago, when I was hosting a Citrix podcast back in the day. I think we talked about our synthetic monitoring solution. Yes. Yeah. We just released it. Yeah. Yeah. Nice. Yeah. Yeah. You were a part of the Citrix ready thing for a while, right? Yes, that's right. Yeah.

[00:25:23] Oh, well, so for anyone listening, want to find out more information about that seven day challenge, Control Up, anywhere you'd like me to point them? Yeah, I think control up.com is a great place, but I guess the other places we have a community as well that's run by a former pretty famous Citrix figure, Doug Brown. There's, I think, almost 3,500 people in that community every day on Slack.

[00:25:48] So it's also a really interesting place to get practical insights from people that are actually using the tool and lots of good resources there with, you know, real people. Awesome. Well, I would add links to everything. I encourage people to get involved, especially around the community aspect of everything as well. I think that's where the magic happens very often. But a big thank you to you both. We've got to make a promise to each other that we won't leave it so long before we all get to speak again.

[00:26:16] It'd be great to get you on back next year, see how things are evolving. But a big thank you to you both. Thanks, Neil. Thanks so much, Neil. I think running a script and resolving a problem are very different things. And Jed's emphasis on checking the result offers a useful starting point for anyone assessing autonomous IT. And, God, he broke the discussion back to organisational knowledge.

[00:26:39] An agent might understand a technology, but it needs the procedures, the permissions and priorities of the business where it operates. And connecting systems also provides information. Well, defining those boundaries, that is what gives teams the basis for deciding what actions to follow next. But I think, for me, the business case here also needs to account for both sides of the support problem.

[00:27:06] There is the time IT spends investigating it, of course. But there's also the time an employee loses waiting or working around it. So a big thank you to both of my guests. You'll find both their profiles, their company website, their community resources, everything we mentioned in the show notes over on the blog post associated with this episode at techtalksnetwork.com. But over to you, where would you begin reducing that repetitive IT work?

[00:27:35] And how is it that you would measure its success? I want to hear your stories, your successes, failures, everything in between. And you can do that by simply visiting techtalksnetwork.com. But that is it for today. So massive thank you, as always, for taking the time to tune in. And hopefully you will join me again tomorrow. I'll be here, same time, same place. Speak with you then. Bye for now.

[00:27:59] Bye for now.