ITSM in 2025 with Magnus Sundset

[Magnus] (0:04 - 0:12) Democratizing data and having it spread all over is not always a good thing, because in the wrong hands, it could be not so good. [Darren] (0:14 - 0:22) Welcome to the DevOps Sauna, the podcast where we deep dive into the world of DevOps, platform engineering, security, and more as we explore the future of development. [Pinja] (0:22 - 0:32) Join us as we dive into the heart of DevOps, one story at a time. Whether you're a seasoned practitioner or only starting your DevOps journey, we're happy to welcome you into the DevOps Sauna. [Darren] (0:38 - 0:44) Welcome back to the DevOps Sauna. Today, we're joined by Magnus Sundset, our head of ITSM Global Practice from Stockholm. [Magnus] (0:44 - 0:48) Hello, hello. Magnus from Stockholm speaking. Good to talk to you. [Darren] (0:48 - 0:53) You too. And of course, we're joined as always by Pinja. [Pinja] (0:53 - 0:56) Hello. And Magnus, welcome, warm welcome to the Sauna. [Darren] (0:57 - 1:12) Thank you. Okay, so the topic we have today, ITSM, which I don't think is a new acronym for anyone, but I think it's always good to get a bit of a grounding for those of who perhaps don't know its origins. So let's start with, where did ITSM come from? [Magnus] (1:13 - 3:19) Yeah, well, it's sort of a long story, right? When I started 20 years ago, ITSM was already there. So like service, how do you cater IT to the population internally and then externally, and how it has evolved into good practices and best practices? And how can we sort of, from the IT perspective, give good services to our internal customers or consumers, as we now say in ITIL version 4, to cater for service management to the outside of the company, and how the evolution has been going is rather cumbersome. And some people may even say it is bureaucratic and boring. I would say the contrary, because we need some sort of bureaucracy and, you know, to have a structured way of working with our IT operational side, because you don't want to put too much effort into troubleshooting stuff that has gone wrong, right? So that's still sort of the foundation of service management, having a stable environment, and just to cater for that could be really troublesome. So the evolution, I don't know, we speak, at least in my domain, a lot of ITIL practices and how that has evolved. But now the trend has become more like we talk about service management rather than just IT service management. We talk about service management, Enterprise Service Management, more these days. Nothing of it is really new on the planet, but it's becoming more clear, at least to me, that we have three big pillars. We have the business, we have the operational and support side of it, and we have the development side. And how can we get the whole shit of all of this in a good structured manner and deliver something that the buyer doesn't really, I mean, they don't know what's going on behind the scenes, right? It should just work. That's what people think. But there's a lot of men and women who work behind, man hours and women hours behind it. So that was the short description. [Pinja] (3:20 - 3:57) So we've gone from the IT background, the IT orientation, now it's being used in other departments. You mentioned Enterprise Service Management. And I guess the question from my side is, isn't every company now an IT company or software company these days? At least when we look at the job market and we see, let's say, a bank that we might not traditionally consider an IT company or a software company, but then they say that we actually are a software company. So is it a branding issue or is it somehow now tied to the service management of that business around it? [Magnus] (3:58 - 5:38) Yeah, it's like IT has become a commodity. So IT is business, business is IT. So more and more companies stop talking about IT. It's just out there. Everybody's using IT. It's in your phone. You have AI in your phone today ,and you can chat with Gemini or chat with your phone. So it's more and more a commodity out there. But I would say the evolution into service management and enterprise service management for businesses out there is that we can reduce the good old practices for incident management and service requests, et cetera, to cater for human resources. How are you onboarding new hires, for instance, and how can we reuse the platforms that we already have? Because people and organizations out there, maybe they have to reduce cost and start reusing the platforms they already have, and sort of getting more value for the money and their investment. So choosing the right platform is becoming more and more crucial so that the business also can use and reuse the practices that you already have. Software development departments, they may not use the ITSM tools, but they will sort of link into it, so everybody knows what's coming. You want to know from the service desk side what's coming in the next release, so you can answer the questions when people call in, et cetera. So those questions are still valid, but more and more people out there in the business, they will sort of start using the same tool sets because they want to get closer to their native IT, if you say. [Darren] (5:38 - 6:20) And kind of entertaining how it mirrors development so well, as well, because we're looking quite frequently at systems where people are distributed across GitHub and GitLab, and then they're building in four different pipelines. And it seems like ITSM just mirrors the principle of trying to get everyone working together in one way and pushing everyone forward in the same way, instead of having this kind of disparate spread that just keeps getting wider and wider and less focused. And then all of a sudden, you have 50 different methodologies you have to unite. Only with development, it's about languages and pipelines. In ITSM, it's about people and practices and how they do things. Yeah, exactly. [Magnus] (6:20 - 9:05) And I think the trick here is to get it as simple and easy as possible and think about content rather than the process itself. It should be an easy process. So if you really think about ticketing, for instance, you have something that's new, in progress, and done. Those three main things that go in development in any case where you are doing, even if you're renovating your bathroom, it's like new. Well, I have an idea of renovating my bathroom, and then you're doing it, and then it's done. And between them, we start making things more and more complicated. But visualizing what you're doing for the other departments like development, hey, we're seeing a lot of bad stuff going on in this domain. There may be vulnerabilities or incidents happening, the system is going down, etc. And then we may call the operation side, hey, what's going on here? Because I can see there's something fishy going on here, we can see it from the monitoring. So all this visualization can cater for the conversation that we will have with our colleagues. So I will talk to them as a service desk agent or talk to the operation guys, hey, what's going on? Well, we just promoted our deployment, it was a new release from this area and was a little bit erroneous. Now we're backing out from it. And we had to take down the server. Okay, great, then I know and I can talk to the customer. So there's like, who is this person that knows everything, you know, you don't know everything, but you need to have the backup systems, and that the colleagues are backing you up. If you're fronting the customer, you need to be able to answer those questions just by punching down a couple of questions to your knowledge base, you will probably get some answers. But when it comes to the new stuff, you need to know what's going on in the operational world as well as the guys in the development area. Are we going to promote more features in a very erroneous area, creating more havoc and more complexity. And nobody wants that because the developers probably don't want to do troubleshooting from some old releases that they never fixed the bugs for. So everything's tied together, I think. So I think the holistic view, keeping it simple and visualizing for everyone involved, is a key point to it. I probably derailed from the question already. But I think this is such an important thing that we need to open up the systems, visualizing what we're doing. So everybody's on board. So we are on the same team as a company. It's not over there. It's not them. No, we're on the same team, all of us. And who can help me as a service desk agent in the service management field? [Darren] (9:06 - 9:18) As a core concept, do you think the idea of ITSM is really just to break down silos, to cut through these communication blocks we seem to see every time we move forward with a lot of these things? [Magnus] (9:19 - 11:08) Exactly. But when I started my career at the telecom company, the manager said “What really caught my eye on that job was, Magnus, you're going to be the spider in the net.” “Oh, what do you mean?” “Well, you're going to know everything about, a little bit about everything. And you need to know who to call when something goes down.” It was pretty much in a networking area type of thing. That was 20 years ago. So like, okay, I need to know people. So I had to break down the silo myself, to be this spider in the net, to network. And I loved it. It was fun. But then, after all the evolution from, back then it was IT version two, it was, we're talking about applications. Now it's like a system of work with thinking holistically. We even talk about philosophy, like DevOps and the System of Work from Atlassian and the ways we're working together. How can we break down these silos? And I think it opens up the gates between teams. How do we do that? Well, visualize it somewhere and open up that dynamic page of whatever you have there and talk about it in your standup meetings, in your Scrum meetings, or in your town hall meetings, or wherever, talk about it and lift up the problems and see it holistically. Because you never know, maybe somebody from finance is wow, you can do that in the system. I can actually reuse this ticketing system for our purposes. We can put in tickets and start planning our financial year here. We don't have to have another system for it. We can just reuse what we got, you know. So open up the communication channels and start talking to each other. I think that is the key to good service management for real. [Pinja] (11:09 - 11:43) We've come down to this in many conversations here in this podcast, that it always comes down to the people, it always comes down to the communication. One thing I'm going to focus on here with the next question, perhaps, is that you mentioned that it is crucial to keep it simple. And it wouldn't be the year 2025 if we wouldn't talk about AI and agents, obviously. And it has won its way also into ITSM, hasn't it? And let's say, how do out-of-the-box agents play a role in this? We can, for example, take the example of Atlassian's Rovo agents. How do they work with ITSM? [Magnus] (11:43 - 13:14) Yeah, since it's now open and sort of included in all the packages. So it opens up Pandora's box. And let me explain a little bit. This is really fun when we get new tools, where everybody's excited over it. We start testing it, experimenting. So that's where we are now. Before they opened up after Team ‘25, on what was it, 9th of April this year, 2025, it was rather expensive. Now it's open. So once you have, you're on the platform, you can use it. People start using it. And then comes the next tricky thing. What do you get out of it? Do you know what you're getting out? Because the permissions you get on the platform will permit you to see everything, of course. So if I ask a question in the chat to the chatbot, hey, can I get to see all the release notes from last month? Okay, I'll get whatever I get. But if you two, Pinja and Darren, when you are doing the same question, exact the same prompt, you will probably say something else, because the permissions are not the same. So your world is different. Your AI world will be different from mine. So ,if I put that in a report, can you guys trust it? And where does it come from? Yeah, it comes from me. Oh, your permissions look like that because you're in the ITSM app. So if you're in development, you will probably see much more deeper down into the stacks of AI. So I think, what was the question again? [Darren] (13:14 - 13:23) I think we're talking about Rovo and the whole out-of-the-box agent, the new release from Atlassian, and what people can expect from them. [Pinja] (13:23 - 13:41) Yeah. And how perhaps use them? And are there any prerequisites for how to use them? And why should they use them? And is there anything that needs to be done in addition to that? And just like, can I just get an out-of-the-box agent and start using it? And wow, now my ITSM practices are perfect after that. [Magnus] (13:41 - 15:59) Yeah, because, I mean, first, the permissions gives you one idea of what you will get out from the agents and the chats that you will have with AI. The second thing is the knowledge and the data behind. They could also be different. So if you have a document, a Word document or something, you send it to somebody and it's on average saved like nine times, I think, in different email boxes here and there. And then people start taking that document, they are updating it. So all of a sudden you have nine versions, live versions of this. And then you think, okay, which one of all these versions are okay? So you need to start cleaning up the knowledge databases, the areas, the pockets of data that you have out there, and clean up those. You can use AI to rewrite your articles, your knowledge articles, and stuff. Absolutely. When you do that, you should also set a timer on that document to, in a year, we're going to come back and refresh the data, or just delete it. So if you have really bad data, you will also get a bad system to ask your questions to, right? So you have the permissions, maybe you don't see everything. So that could be some erroneous thing. You can have bad data that you are taking out your reports from. So when you put AI on top of that, it could create a mess. So I would suggest maybe some sort of an AI coach that goes in, where do you fit those agents in? Like the AI agents, where should they work? What kind of workflow, what kind of purpose does it have? What kind of use case do we have here? Does it fit? Because if you look in a year or two, we will have hundreds and maybe thousands of those agents working for you in your organization. And what's going to happen? You know, people will stop using them because they're afraid of what's going to happen. If I use that one, I don't know what's happening. The last time I used that one, something happened. I didn't understand what happened. Somebody called me, Hey, what's going on? And so you need to have control of those agents, also know when to use it and how they should be used in your workflows. So it's setting the strategy, how to implement AI, especially with the agents when they're becoming more and more autonomous in the future, you need to have a structured way of dealing with them. [Pinja] (15:59 - 16:20) And as we discussed before, the crucial thing here is again, the communication between the humans, the people. So what are you trying to do? How do you communicate that in the organization? And as we have discussed many times in this podcast, that AI is nice and it will make things more convenient, but it's still just a tool. [Magnus] (16:21 - 17:24) Yeah. And when you have this tool and when it's open to all the Atlassian users, then people start experimenting, and all of a sudden, you will see that there's a lot of nice agents, and you start using them, but you don't know if they are live or not. I mean, if they are sort of drafted or public, or they should be used, because when you start creating them, you will see them in the list of agents. And so if Darren, you create one and I start using it, but you just tested it, and you're trying to do something magic. So once they're there, everyone can use it. So I think in the future, and I had a talk with Atlassian and also Sven Peters, we had an ACE event here in Stockholm and he said like, there's probably going to be more permissions to those who can create those agents. Because if everyone is creating those agents, there will be so many that you can't keep track of all of them. It's like everyone in the organization is creating new features for your system, right? You won't let people do that. But now it's at your fingertips already. [Darren] (17:25 - 18:12) And I do think that kind of echoes the problem with the data, because what we have in almost all situations is data that's not in an ideal position to have AI applied to it to begin with. So then you basically build up this faulty stack where you have bad data with a like massively, let's say, saturated AI space above it, which no one's really sure what the features are. And then if you expect results out of this that are the same or useful, you're going to be disappointed. And that's what we mean when we say AI is a tool, you know, a tool can be great in the right hands, but you need to know how to use it. And without the data and without the understanding of the agents, I think people are going to run into a lot of problems before they find solutions. [Magnus] (18:13 - 19:18) Yeah, for real. I mean, if you have, I talked to this, it's not a customer of ours, but it was a lady that came to our ACE event, and they spent like two and a half months just cleaning up there over, I mean, it was hundreds of thousands of bullish articles. And again, you can use AI for that. And that's really, really good when you can use AI to clean up, you know, just to comb through all your data and get it structured in a good way. So I think that is the foundation, and then start building a strategy where you want to, what's your targets? What, what, what are your goals with this new tool set that you have? So, I mean, democratizing data and having it spread all over is not always a good thing, right? Because it's in the wrong hands, it could be not so good. So, restricting the data and understanding the output of whatever you have, I mean, all the agents will give you some sort of output, right? And understanding what it really means and in what context you're giving the data assets. I think that is super important as well. [Pinja] (19:19 - 19:42) And I guess the more we're using this, if we still stay a little bit in the AI versus the humans, because of course, many articles have been written, like, are we going to lose our jobs because of AI? And sometimes it might feel, even personally myself, like the more I use it, I will get this fallacy that am I going to lose my job because of AI? But Magnus, is it actually the contrary? What do you see in ITSM here? [Magnus] (19:43 - 21:27) Yeah, I don't think you will get a job if you don't know about this. So, in order to keep your job, you need to know about this. I think that if you are missing this train, because it is here to stay, it's a huge revolution. This is, I mean, Industrial Relief, I don't know, like version 5. I mean, this is huge. It is, really. It's groundbreaking news, like every three months, there's a new thing. And there's a graph I saw from a Norwegian company that they are sending out that in 2026, we will reach IQ Level 140, I think, for those AI LLMs that are out there. And that's amazing. That's huge. And that has just become, like, you could see the graph, you see all the tools are sort of shifting right to a higher IQ level, which is also very scary. It's like, what's going to happen with my job? Am I going to lose it? Or yeah, I think it's like when IT first came into play many, many years ago, and the internet and all that, we shifted. We shifted from printing presses to printing in a printer. We went from faxes to email. So, of course, there are companies going out of business if they continue to just create and build faxes. Well, when other people are upgrading and jumping onto the new technology. So I don't think there's a big difference between the old times of changing technologies. Now, we also need to be on board the new boat, so to say. If you're not, you're standing there on the dock, you could see all the boats sailing away, and you're still standing there. Yeah, it's going to be a beautiful ocean. [Pinja] (21:27 - 21:29) But you are not on any of the ships. [Magnus] (21:29 - 21:32) You're not on the nice cruise ship, exactly. Yes. [Pinja] (21:33 - 21:38) And I guess it is, we still need people to bring context, right? [Magnus] (21:38 - 23:14) Yeah, for sure. As service management has been evolving into, I would say, if you're lifting up the phone, somebody's calling you, and you could sort of transcribe the call into text, send that text over into AI, get the answer back immediately when you're talking to the person. Then we have won something, and we're already there. So what if somebody sends in a video? Look, there's an error here. Yeah, can you just send me a video of it? Yeah, sure. And then you just send in the video, and then you have the video transcribed, and you can raise bug fixes already there and send it into the ticketing system. Because I've been working in QA, and you're trying to recreate issues that the caller is posting. And that could be really difficult, right? Like, what did you mean? Where did you go? In my screen, there's no fault, but in yours, it is. And there's a ping pong of so many questions going back and forth. But if you could see, OK, you're on that screen, you're using that browser, you did it at this time point, then all that would be recorded, and you could create tickets from that. Using AI would greatly help you, right? Just to fetch all those small data points that are so important, so you avoid this traffic of sending an email back and forth. So you can actually sell tickets faster, for real. That is what I'm expecting from the future tool sets, that this would be a commodity that would be out there. And it's already there. [Darren] (23:14 - 23:46) That's interesting. It's like, because that's kind of been a goal a lot of times for things like telemetry. So when we see what we want, when we're developing, we want that information to come through to us. But that always requires some kind of backend that's passing telemetry. And then we have all kinds of rules regarding that because of the gathering of information. So the idea that it could be voluntarily given and adapted from easily accessed media is that, yeah, I can see that would be game-changing for ticket solving. [Magnus] (23:46 - 24:24) If we can mention one tool out there, you have Loom, so you can record your screen with your face, whatever, you can have your face in it or not. But you can transcribe that. And then you could, with AI, also create tickets and send it to Jira Buckets of Work. So it is there. It's just a couple of clicks, and then everybody understands, and you can repeat the video over and over again without even calling the end user. And you can create the bug fix and promote to production rather quickly, and then just call back. Hey, can you test again? All right. Okay. Works. Fine. Thank you. Everybody's happy. [Darren] (24:24 - 24:42) That sounds really great. I think we have this worry that we're coming maybe a bit too AI-centric, but ITSM seems like it's about giving service to people. It's the human-centric element here, that we're not serving an AI. AI is the tool to serve people better. [Magnus] (24:42 - 27:18) Yeah, absolutely. I mean, it's like either I use my mouse pad or the mouse. It's just like a way of, I can write it with my pen on the screen also if I got a touchscreen. It's just a way of working. And the way of recording something becomes more and more of a commodity. It's like, I mean, today everybody knows how to make a video and film with a camera and all those things, and take screenshots and send it in through a ticket. But capturing the data itself from that message, I think that in itself is a game changer for service management, especially. Because then you can see, okay, what kind of requirements are you actually looking for? What kind of features are you looking at now when you saw this bug happening? Oh, you have a new idea. How would that work? So, just eliciting bug fixes, ideas, product enhancements, all that is so much easier when you can render it down from the videos and the speech and the transcriptions and all that you got. So that would also speed up the understanding of what people actually want from you. So if I record a video of a mock-up of something, how I would like it to happen, and I send it to Darren and Pina, and you guys are developing a beautiful software. But how about I just do it myself and ask, “Hey, how would that look?” Yeah, like this, and with screenshots, I can get a whole set of features presenting to me in an image format. That would speed up innovation. So, how about using the service desk agents in that manner? When you see there's a lot of things going on in this area, why don't you just recreate that? Maybe an idea of a new feature that would solve a lot of problems that the customers are experiencing. So the service desk agent could spend some other time to actually be innovators, to send that further down the line to software development. Because the service desk agents, they know both the fear and they have empathy. They know the fear from the customers and they have empathy for them. And they also know how we can assist them the best way. And then if you're thinking a little bit, maybe I can start innovating, I have a couple of ideas, where do I send it? Yeah, I want to send it to some development department to innovate, to be part of the innovation. So I think that is something that we could use our staff more to bring innovation. [Pinja] (27:19 - 27:36) Yeah. So, like building better service for the customers, or was it the consumer that ITIL refers to in the framework? So if we deep dive into that a little bit, what is good service, and do we have a standard of how fast are we actually expected to give the service? [Magnus] (27:36 - 29:16) I think it's in the perspective of the consumer, there's a consumer in the service. So if I expect it to happen in a minute and I get it in a week, bad service. But the people behind delivering the service within a week, for them, may have been working their ass off in order to get that service up and running. And so it's about managing the expectations, I think. So I was in a company where we actually did exactly the same release to two different huge companies. One did not get any training or pretty much no information about the new release. The other one got all the information, we sort of worked for them, and promoted all these new features to them. But the other one also has the same features. Guess who got really upset and who got really happy with all the new features? So it's about managing the expectations. We didn't expect anything to happen in the QE this week. We're doing a lot of demos and we're doing a lot of sales here. And now you're promoting all this without telling us? I mean, so that was really, really bad perceived from them. Their experience got really bad. So I think the experience management from the consumers out there is the most crucial thing out there during 2025, actually, what we're seeing. If you have a bad experience, you could spread that on social media in a couple of minutes, and everybody knows that you've got a really good bad service at that restaurant, and nobody goes there. So you could be out of business within minutes today. [Darren] (29:16 - 30:01) Yeah, I guess we're at the point where communication's kind of accelerated. Like we talk about AI acceleration, but the internet led to communication acceleration far more quickly. Well, maybe not going further, but just the idea that maybe not just the instant connection, but the perceived anonymity that just allows you to say whatever you want without any kind of consequence. So yeah, I think the ITSM has probably been hit quite considerably by people's expectations for their software now, their data now, their tooling now. So, having these AI tools in the background is probably going to be really helpful to keep up with those rapidly accelerating expectations. [Magnus] (30:02 - 31:21) Yeah, and you can draw conclusions from many, many, many different data sources and sort of merge them all into one big data pool, and you ask questions to it. When you can actually ask this Oracle, if you like, any type of questions, because you have all this data, then you will receive other types of information and trends. You can identify trends in the marketplace and how people are perceiving you. If you're looking into social media, for instance, and you download all of that and you ask questions, you may have had a hunch in your stomach, oh, people don't really like our company, but we don't really know why. Is it because of bad service, bad products, or bad quality in our product lines, or what is it? We need to know. And then you ask people about it and you send out surveys, but the survey is just as good as the people who are answering, right? If they're answering, maybe you're super angry with something and you don't get that survey, then it's not really, you don't get the message, right? So getting a more structured intake of data from the data sources you got, and you can draw conclusions and insights from it, I think you're onto something really, really good for your service management. [Darren] (31:21 - 31:46) Yeah. And I think the input of data is important too, because what we know is that these datasets are built on years, if not decades of information. So making sure data is weighted to understand more recent data as priority and not grabbing a document from 15 years ago that tells you how to install the software on windows NT or something like this is not helpful. [Magnus] (31:47 - 33:05) Yeah, exactly. And, but it's also good to see if you want to have a historical track record of how we have performed to set the baselines and see the graph going up like the hockey stick, you know, Oh, we're better now than before. But I mean, mixing the datasets, I think you could draw faster conclusions than before, because just looking into things that you did not understand before, like I was looking into this Digital Operational Resilience Act, and it was a lot of law texts. I uploaded it, ChatGPT, to see and ask questions, learn, teach me about this, you know, and immediately, my skill level went up. I didn't know who to ask. Who should I ask something that's pretty new? Well, at that time, it was brand new. You know, I just have this law text. Now I'm not a lawyer. I don't need, I don't know how to read all this, you know, just asking the questions and getting some answers back. That is awesome. I think just understanding who the caller is asking questions, you do that verbally, right? But the real problem they are having, you can ask AI, because that is sort of materialized into your software and everything that you have delivered. [Darren] (33:05 - 33:27) Okay, so it does seem to be all about bringing everything back to the user experience and shifting everything closer to that immediate feedback. Yeah, shifting left for sure. I think that's actually all we have time for today. So I will say thank you for joining us, Magnus. It's been great to talk to you. My pleasure. Thank you for having me. And thank you, Pinja, for joining us. [Pinja] (33:27 - 33:29) Thank you. And thank you so much, Magnus, for joining us today. [Darren] (33:29 - 33:30) Thank you. [Pinja] (33:34 - 33:39) We'll now give our guest a chance to introduce himself and tell you a little bit about who we are. [Magnus] (33:40 - 33:52) Hi, my name is Magnus Sundset. I'm the head of the ITSM Global Practice here at Eficode, working from the office here in Stockholm, catering for good service management. Thank you so much. [Darren] (33:53 - 33:55) I'm Darren Richardson, Security Consultant at Eficode. [Pinja] (33:55 - 34:00) I'm Pinja Kujala. I specialize in Agile and portfolio management topics at Eficode. [Darren] (34:00 - 34:03) Thanks for tuning in. We'll catch you next time. [Pinja] (34:03 - 34:11)