Ep. 20 Navigating Cloud Complexity: FinOps, AI, and Remote Teams

00:00 - Introduction and Career Journey

Matt Pacheco
Hello, everyone. Welcome to the Cloud Currents podcast that navigates the ever-evolving landscape of cloud computing and its impact on modern business. I'm your host, Matt Pacheco. I'm the head of content marketing at TierPoint, where I help businesses understand cloud data center and data protection trends affecting various industries. Today, we're happy to welcome Rolf Bekkstrand, the director of cloud and engineering at Visma. With over 15 years of experience in tech leadership, Rolf brings a wealth of knowledge in cloud strategies, engineering performance, and AI integration. At Visma, a leading provider of business-critical software, Rolf oversees the delivery of cutting-edge cloud and engineering services across a vast network of autonomous companies. His experience spans from hands on project management to high level strategic leadership, giving him a unique perspective on the challenges and opportunities in today's cloud driven world.

Today, we'll explore topics ranging from cloud adoption strategies, AI integration, managing decentralized tech teams, and navigating how businesses can stay ahead in this rapidly evolving field. Always very rapidly evolving. Ralph, welcome to cloud currents. We're so glad to have you.

Rolf Henrik Bekkstrand
Thanks for having me, Mats.

Matt Pacheco
Excellent. So, can you walk us through your career journey, starting from your early days in tech to your current role at Visma?

Rolf Henrik Bekkstrand
Yeah, I graduated in 2008, educated myself as a programmer back then, but turned pretty fast, actually into project management, then later into department management. Then I also were part of founding my own company, where I had the joy of going back to actually doing programming again. We kind of saw ourselves as a bit early adopter of the Google cloud back then, but ending up as a CTO of a midsize company, then ending up in Visma, having a large international development organization, and then later into director of cloud engineering, where I am now. So it's kind of like, with the technology, it's evolving super fast, and you kind of never know where it's going to go next. Right? I kind of feel it, like, the same way for my own career.

I didn't really ever know exactly where I want to go and what to do, but I always found some new opportunities for all the roles. It has been in different industries, different domains, but it has always had something to do with software. And I find it very fascinating how much everything has changed just since I graduated in 2008. We didn't even, I mean, I recall my own secretary asked me to buy a smartphone. She thought that you are the only one in the company not yet changing to a smartphone, and look where we are now.

Matt Pacheco
Yeah, quite a lot of changes. So you've worked in various roles from hands on development to high level strategy. How has your diverse experience shaped your approach to leadership?

Rolf Henrik Bekkstrand
Oh, that's a good question. I think that these days I having maybe a more, of course, a strategic perspective to things. But I also am very commercially having a very commercial mindset to everything I do. But I think that to be able to kind of combine that commercial focus, but also really having that technical background is maybe one of the things that gives me a bit of personal strength, if I can say so. Yeah, but I usually often say that I'm educated as a developer and by DNA, I am a developer. But you should probably not hire me as a developer today. Right.

04:42 - Visma's Structure and Mission

Matt Pacheco
All right, so you work at Visma. Can you give us a little bit more about the structure of Visma, what you do and how it's set up?

Rolf Henrik Bekkstrand
Yeah, sure. I would say that we are a huge collection of companies and awesome people, highly engaged people, currently 15 and a half thousand highly engaged employees all around Europe, but also Latin America. What we really focus on is what we call mission critical software. It means that you cannot really run your business successfully without having it easy to say, accounting software, payroll software. Those are those who are very easy to explain. It's kind of mission critical for your operations, but it might be that you are heavy into also logistics. If you are running a store, selling products, maybe it's actually e commerce that is critical for your business. So it's kind of, all these type of software are delivered by the Visma companies all around Europe and Latin America.

And it spans from software that is meant for a startup up to quite high in the market, where you need super advanced ERP features to actually successfully run your business.

Matt Pacheco
The structure you guys have at Visma is really interesting. It's allowing all of these acquired companies to kind of maintain autonomy. That's really fascinating. How do you ensure consistency in cloud practices across all of these autonomous companies under the Visma umbrella?

06:29 - Cloud Adoption Challenges and Strategies

Rolf Henrik Bekkstrand
Yeah, that's a difficult question. The good thing about this kind of autonomy that we have is that our entrepreneurs, they stay with us. If a company is acquired by Visma, they stay with us. So that's the really strong side of it. Right? We don't really force that much. We, of course, we have, as any organization, we have some requirements to the financial part, and we also put quite strict requirements when it comes to security. But when it comes to cloud, it's more about, we are encouraging. We also really like data in Visma, so we can collect a lot of data from our development teams, from our companies, and we use that data and try to find correlations looking into, okay, companies that are on public cloud, they perform financially better than those who are operating in a traditional data center.

And then we share that information with our companies as an encouragement and usually a managing director for a company, or most, actually people in that organization, they want to do even better commercially. Right. So, of course, usually they buy the story and they start their cloud journey. Right?

Matt Pacheco
Yeah. Constantly evolving. So a lot of this development happens in the public cloud. Correct. So from what I understand, Visma has a goal of 100% public cloud adoption. What are the main challenges you faced in pursuing this goal of 100% cloud adoption?

Rolf Henrik Bekkstrand
Yeah. So first of all, we will probably never reach 100%, especially not public cloud. But we are encouraging all of the companies in the Visma family to use the hyperscalers if they can. And that leads me into some of the challenges we see today, because especially some governments, municipalities and such around in Europe, they might be a bit scared of the cloud. The political climate is actually affecting that a bit. You had the European Union in the EU, and then you have us and the hyperscalers, they are owned by us vendors or they are us vendors. Right. And that is actually challenging when you want to deliver mission critical software to some governmental institution or a municipality or things like that. So in many ways, I would say those are the biggest challenges with the cloud adoption today.

While in the past, that's a more kind of complex question. Before you had it, organizations, they were managing the data center. It's kind of, we take care of operations, you do development, but at some point some small smart people figure out that DevOps is the thing. Right. Let's merge development and operations and let actually one team take care of everything. So from that point of view, maybe in the past it was more like a culture thing and kind of how we have been organized for the last ten years. You cannot change that. Right. It's not easy to change things. While these days, I think that most teams have kind of overcome that and they really like to work with a DevOps kind of way.

Well, then you kind of end up in like, do the customers actually accept you to use one of the hyperscalers, for example, because of data protection rules and regulations that you find in the European Union?

Matt Pacheco
Very interesting. So you kind of mentioned the way of working in DevOps. What are some other ways that you kind of overcome those challenges of getting customers to accept and getting companies to accept public cloud in those regions.

Rolf Henrik Bekkstrand
There is always a difference between what you are a bit afraid of and if you have a kind of firm requirement and where you have a firm requirement from a government, then you cannot really change it. But there are many kind of gray areas here, and that's where you kind of need to educate your customer a bit. Explain to them that actually security, that is usually something that a hyperscaler, they are most likely better on security, taking care of the physical security, the network security, everything, than a traditional data center with maybe a very few employees. And how can a small company invest in a lot of physical security, for example? So you can play on those kind of things, but also when it comes to where we can actually, with a hyperscaler, you can actually store your data right on multiple locations.

That is also something you don't get out of the box with a traditional data center, they have one location, so if that location goes down, then you're basically out of business until they are up and running. Of course, you can have your backup somewhere else, but you don't have a smooth transition. You still have to wait until your data center is backed and you can start the restoring part. So it's all about educating the customer about why is this good for you and why is this actually better. And then many customers, they will probably say that, yeah, well, I actually see your point. It's actually better. What you're offering is actually more secure than what I have today.

13:25 - Cloud Cost Optimization and FinOps

Matt Pacheco
Excellent. And what we hear from a lot of companies is one of the other challenges is cloud cost optimization. What are your thoughts on approaching cloud cost optimization while ensuring performance isn't compromised? What are your thoughts on that?

Rolf Henrik Bekkstrand
That's a good question. In Visma, we are focusing a lot on the FinOps activities. We try to educate all of our teams to kind of look into your cloud costs, use some specialized finlops tooling to get recommendations, and use that as an advice to actually optimize. And of course, by reserved instances when you can do that, when you actually know that we are going to have this traffic, it's quite stable. If anything is going to increase because we are so good on selling our product. Right. And then, I mean, I hear that sometimes from the teams that, yeah, well, we cannot really change this, but, well, have you tried? And sometimes this is maybe also a bit hard.

Sometimes, at least for me, when looking into all these type of different setups that the hyperscalers are offering, this machine versus this machine is actually okay. It can actually appear that the hardware on this machine is less good than another one, but actually it's not. There is some letter there saying something so it can be quite complex as well, and I can understand it, but our software needs to be operational 100% all the time. Right. It's mission critical software. If you cannot do your payroll, then you will have very unhappy employees not getting their salary paid. So kind of, we don't really sacrifice on performance reliability. I wouldn't say that. But we still have a very high focus on applying FinOps techniques in all the different teams, and we'll do a lot of education around it.

And we're offering some centralized tools where they can get advice upon what to do and even integrated their gyro or other ticketing system. So that kind of becomes part of their bi weekly or whatever cycle they are having for their scrum and their planning rights.

16:17 - Cloud Migration and Best Practices

Matt Pacheco
Excellent. So you're trying to get a lot of these companies under the Visma umbrella to adopt the cloud. Can you share some of the insights on your cloud migration approach with them and how it helps those companies transition to the cloud?

Rolf Henrik Bekkstrand
Yeah, sure. So our companies, they are really running fully autonomous, but yet we have a quite small group function also in Visma, and those are about educating, sharing best practices. Of course, when it comes to cloud, it's also about the supplier management to kind of keep a good relationship with the cloud suppliers and really make sure that we understand their offering and can make sure that we spread the word around to our companies and development teams. So that is one part of it. We also have actually a hands on cloud migration factory, as we call it.

Because if you are migrating either if it's just to migrate from a traditional data center and into hyperscaler, or whether it's actually to take software from on prem and to the cloud for the first time, it might be that you are especially for the smaller team, smaller organizations, maybe they don't really have the capacity, the competence and those kind of things internally, and maybe it doesn't make sense either to hire that many people either, but then they can use our cloud migration factory to get help on the way, they get some hands on migration help, and also they can, as a part of those projects, actually educate their own existing people.

Matt Pacheco
And does that include, I guess, engineering and DevOps practices as well? Is that something you share with best practices with those teams too?

Rolf Henrik Bekkstrand
Yeah, definitely. Where I'm sitting in the cloud engineering, we do the cloud part with a lot of focus on Kano cloud finops, also the cloud migration factory, as I told you about, we also have an architecture and technology team that is really our expert in how to, well, architecture a cloud solution, how to make good software architecture, but also really experts on the DevOps principles and those kind of things. We also have what we call the Visma cloud delivery model, and that is a model that is perfect for teams who want to transition into being a totally autonomous team that takes care of development, deployment and operations completely on their own. But it depends a bit. What's the starting point? Kind of, what kind of tools and frameworks that we have in the group that you want to pick.

But I would say that since we are kind of a big group, around 200 business units in Visma 7800 development teams, they are all different. Right? So in kind of the group toolbox, we need to have a little bit for everything. We also try to kind offer something that is a bit unique, something that you cannot just go out in the market and find there, right. It kind of has to be something that is a value added or kind of what we like to call so called the secret saucer.

20:08 - Data Regulations and Compliance in Europe

Matt Pacheco
I envision there's a level of complexity with managing different companies or different security regulatory compliance across various European countries, right? It's very complex. Can you tell us a little bit about how you overcome some of that complexity with stricter data regulations?

Rolf Henrik Bekkstrand
Yeah. So the good thing about Europe is that you have European Union, right? And you had a couple of years ago, your first time, it was I think 2016 or something when you heard about GDPR for the first time. Right? But at least to my knowledge, it applies to European Union countries, right? We have a few countries that are not in the European Union, in Europe as well. For example, Norway. Norway where I'm sitting myself. But in Norway, for example, you're following the same. Actually, I would say that most rules and regulations when it comes to data protection and privacy in Europe, they are more or less similar. You might have, as we see, some countries that put some specific requirements when you want to deliver software to governmental organization or municipalities and those kind of things.

And then usually that company who has that software, they of course need to decide, is this kind of a sweet spot customer that we want to deliver to? Do we want to adapt to those special requirements or do we not want to do that? And that's when it comes in where the hyperscaler might not really be an option.

22:35 - AI Integration in Visma

Matt Pacheco
So jumping from security one complexity to another potential complexity, I want to talk a little bit about AI, because that's what everyone wants to talk about right now. And I have some interesting, there's some interesting use cases out there for AI in the cloud, but I'm really curious. AI is a focus for you. How are you integrating AI into your processes?

Rolf Henrik Bekkstrand
Yeah, so in Visma, we have been speaking about boosting people with AI and boosting products with AI in a boosting people part, from a development point of view, you can think about, okay, how do we utilize AI in actually the product development itself? Right. So my organization is for focusing a lot on that part where obviously we are rolling out copilot, GitHub, copilot or similar to the Visma companies, and we look into the adoption of those kind of tools every month, how many percentage of our developers are using it. We also correlate data from the users of that with the productivity data of the teams trying to figure out, are those teams using AI in development more productive than teams not using it? We just concluded a study and have the webinar together with Plandec a couple of weeks ago about that.

And we have also during this year looking into how can we use generative AI to modernize a code base. So let's say you are on.net framework and you want to go up to.net eight or perhaps.net nine. Now, I'm not programming anymore, but then how can generative AI be used assistant to actually speed up that process and not to make any promises there, of course, but we have seen a 40% productivity gain in at least one very big project there. We have also seen even better, but we don't really have that much data yet that we can really draw a firm conclusion on it.

So that's one of the parts under people then you have obviously in the customer support, we see a high, really high adoption of AI in the customer support to offer chatbots, but also even more advanced maybe than just Armstrong questions, but also kind of guiding you in the software, or at least giving you the right documentation of those kind of things. Also, when it comes to the products, I mean, of course, I'm not totally objective here, but I would say that Visma is kind of a front runner when it comes to actually applying AI in the products. We developed our own models for kind of reading, for example, doing OCR many years ago, or we have something called auto suggest that can kind of fill in the empty holes for you, right?

But also, of course, last year, now when all these kind of models have been available to everyone, then of course, then we are looking into, I think that more or less every product team, and we are looking into how can we utilize generative AI to make our product better. For example, in e commerce, right? You are selling products for your e commerce platform. You spend a lot of time maybe on writing a text. Thats perfect task for generative AI to do for you, right. Then you have time registration for example, you can do anomaly detection of that. That's an old model that we have developed in Visma. You can have for example where you in workforce planning, right. That's also a place where we have developed our own model. It can make you quite more efficient.

But we are super eager, we are monitoring every day what is happening out there. We are really trying to encourage all of our product teams to what can AI do for you, what can it do for your product? I kind of feel a huge enthusiasm around it in Visma, but it's going super fast. That's.

Matt Pacheco
I think we all could agree it's going very fast. And we know that AI is kind of an umbrella for a lot of technologies like machine learning. That's not new, that's been around for a while. Generative AI is just one flavor of AI. When looking at all AI, what potential do you see for it in optimizing and analyzing the cloud?

Rolf Henrik Bekkstrand
Yeah, so I just came from a meeting before I met with you Matt where I looked into a bit disruptive product on observability, trying to, typically you are getting a lot of alarms in the observability part today, but those are really maybe more symptoms, they might not really be a problem yet. Right. So at least that company were showing me that kind of, okay, here we use AI to really look into what's the actual problem and what you traditionally see in observability problems as a platforms, as kind of alarms, it was kind of more symptoms. So you can see that tree as one really cool way of using AI. But also looking into the finops part, for example.

I mean, one thing is about suggesting that here you can do something, maybe also the AI can be excellent, that we dare to actually let AI go in and make changes for us on the fly and save costs. But also, you asked me the question about sacrificing on performance, right? Maybe I can kind of test it out for you. See, no, it's not working. Okay, go back and do kind of some kind of those things. But I do believe that as we also see with our customers, that AI is great on giving advice. You should not blindly trust it. Right? Same when you write code. So I think that at least for now, you would like to have a lot of actionable advice and kind of go through them, does it make sense? And then apply them.

But maybe ten years from now, we will let actually AI do more of the job for you. And of course, one thing is writing the code. But imagine as a small startup, you are using AI to really get your code fast, and then you can just tell AI, by the way, I want to host this somewhere. Can you put it out there for me? Right. And it happens.

Matt Pacheco
So a lot of interesting use cases, really exciting use cases that you just talked about. How are you, Visma, and on your team balancing that excitement around AI with the practical implementation of your cloud strategy? Like how you, how are you making sure that the hype doesn't overtake what you're doing?

Rolf Henrik Bekkstrand
Yeah, that's good question. A bit different mechanism. I think that most of our companies, of course, or all our companies, they have the board meetings where they report on kind of the product strategy, and they make a plan for the future. Right? And of course, the board will by nature go in and challenge, is there a real business case here? Are you doing just things for fun? But I assume in many of those cases, the board will never see it because their own organization, they have stopped it themselves. But it's also about, we like all our product teams, of course, to be really good on product discovery and product analytics. Right?

So it's not like I don't think we have many products teams, if any, that goes out there and decide what to make or changes or new features without actually having some evidence in their data about how the users are using it. What do the users really need? And of course, then you have the prototyping part. You don't really make a huge new feature without involving the customers in the journey, testing out the prototypes. Right. Getting feedback going back. And I guess that some ideas were stopped by nature. Already there.

32:42 - Future Trends and Advice

Matt Pacheco
Thank you. So now for a few fun questions, or more fun questions, advice from you and forward-thinking trends. As we close out this episode, I'm curious, if you could change one thing about the current state of cloud computing, what would it be and why?

Rolf Henrik Bekkstrand
Oh my goodness, how should I even answer that? Is it allowed to say the price?

Matt Pacheco
Yes, you can say that.

Rolf Henrik Bekkstrand
No, but really, you come a long way with the finos. But of course, I mean, the cloud cost is getting considerable, right? And especially if you don't really apply finops techniques and you kind of go blindly into the cloud, it can get really expensive. So my wishing in, at least in Visma, is that all teams are getting really good on doing their finops. Right.

Matt Pacheco
Cool. What advice would you give to companies just starting their cloud journey today.

Rolf Henrik Bekkstrand
Well, seek for someone with an experience. It's really expensive to go out there, move to the cloud without making sure that your cloud architecture is correct or somehow correct from day one back to is getting expensive, right? You might end up with not the best performance and you end up kind of just scaling out. But really what you need to do is to make some changes, maybe in your software as well. If you don't, if it's your first time moving to the cloud, make sure that you have some competence in the team, people that have been doing this before. And if not, go out, hire that competence or go to a consulting company and hire someone from them.

Matt Pacheco
Great advice. How do you envision the future of work and cloud, especially considering trends towards remote and distributed teams? How do you envision that changing in the future or remaining the same in the future? What are your thoughts on that?

Rolf Henrik Bekkstrand
I mean we are pretty remote already, right? But obviously I think that, excuse me, that the pandemic that we had showed us how important it is to actually have access to your software every day. We saw that moving from on prem to cloud products. Customers that were stuck in on prem, they were really seeking for new products quite fast. Right? So there are a lot of on prem software out there today. I think that it's not so many years left for that trend. And eventually I think that all software will be something that we require to be there wherever we are on the planet and whatever is the time or the day, we expect it as a consumer that is there for us both as a private consumer but also kind of as a business consumer, we kind of expect it to be there.

And it also changes the way we work. Right. When it comes to before it was like office hours, it's eight until four, you go come to office, you leave at four and maybe you sit a bit longer some days. Well, today, I mean unless you have some important meeting going on at 08:00 it's totally fine. That one is showing up at seven and another one is coming nine, because that one who comes nine can just even leave three and go home to a dinner and can work from home a couple of hours in the evening. So it has been, I mean this accessibility part, how accessible things are to us, has really changed. And I think that in many ways the pandemic was a boost for that, I would say.

Matt Pacheco
And I agree. So looking forward the next 510 years, if we could even do that with the cloud world, what emerging cloud tech or trends are you most excited about?

Rolf Henrik Bekkstrand
It's gotta be the AI part, right. I spoke with our AI director and he is updating a slide of all the models that we are having available out in the market. And he said that this slide is kind of almost like every time I'm presenting it, I have to make sure that it's up to date because there is some changes. Every time there's some changes, this model is getting better and there's coming a new model. I think at least maybe not 510 years, maybe things have stabilized a bit more. But let's see the next three, four years, I think that we will see new models coming more or less every month. And what we have available today, we will probably have a bit fun of them already. A year from now, perhaps.

Matt Pacheco
Do you see things like quantum computing ever playing into the cloud?

Rolf Henrik Bekkstrand
We haven't started to look into it. Right. Especially from a security point of view, whether quantum computing will be kind of available to everyone when I'm still working in the industry, I don't know. Let's see. But it's super exciting and of course that's going to change that computing power we have today. Like with AI models, we will look back at it and how would funnel of it. But I guess with quantum computing, no one really saw that AI generative AI was really coming bang over the night. I think at least I didn't because then I would buy some interesting shares on the stock market. So I didn't really see that coming. I guess we might see something similar with quantum computing. It might be ten years from now, it might be 20 years, might be more. I don't know.

But it's super exciting to follow and read up on.

Matt Pacheco
Yeah, this stuff is really evolving quick. Especially you're matching with the AI pieces. There was a demo from OpenAI video that just looks amazing. It's very convincing. And looking at it, what, six months ago, a year ago, it's completely different. It was like going from Caveman to computer. It's crazy how fast it's evolving. I'm particularly interested in seeing the intersection and really out there, but the intersection between what AI can do when powered by quantum computing, that's going to be some interesting stuff.

Rolf Henrik Bekkstrand
I really have to say. There is some, there is a backside also here because we see how fast no AI is utilized in fraud. I mean, it's almost scary how you can use AI. You just take pictures of someone and you make a video. That person speaking, I mean, how will that affect us? How should we as consumer, as a citizen in the society be able to really differentiate what is fact and what is actually fraud generated by AI. And that might also lead to some political discussions at some point. I don't know what will come out of that, but there are really some things here to be a bit aware of, as well as not only positive side. So that super fast development, that's a great point.

Matt Pacheco
We actually, a few weeks ago, were talking to security experts on this podcast, and they were talking about how scammers, fraudsters are using AI, but also from a machine learning perspective, like security teams are using AIH to identify threats and points of intrusion, while the hackers are also using AI. So it's like a battle of the AI's going against each other, and it's almost like an arms race of who can build the best AI tools. So you're right, there's a lot of negatives that come with that positivity around AI. And then you also have the considerations of how much energy and how much money it costs to run a data center with AI. And the sustainability impacts, do you think about that a lot too? Because I know that's one of the big things.

Matt Pacheco
It's like one query, and an AI tool can use as much energy as a person uses in a month or a year. What are your thoughts on that?

Rolf Henrik Bekkstrand
So, sustainability is something that we are very concerned about in Visma. And of course, it goes down to how do you actually build software? And as you say with AI, suddenly you can build something that consumes a lot of energy. But we put quite strict requirements to our suppliers, and they are also pretty good at it. And our teams, they are educated about how to make software that is consuming less energy. Maybe a bit fortunate is that, of course, at least until now, it has been a quite good correlation between energy consumption and the price. So that helps, right? But we are paying attention to it. We have also developed some sustainability reports looking into the carbon emission of the cloud workloads across all Visma. So our companies, they can go in and look on their own carbon emission.

The product team can do it for their specific product, and also we can monitor it as a whole for Visma.

Matt Pacheco
Excellent. Thank you for your insights. And I want to thank you for being on the podcast today. It was great speaking with you and learning from you and sharing your knowledge and experience. I think anyone listening will get good value out of that. So we appreciate you being on.

Rolf Henrik Bekkstrand
It was very nice to take part of this. Matt, I really enjoyed it.

Matt Pacheco
Thank you. And for our listeners, we appreciate you listening. Stay tuned for more episodes of the Cloud Currents podcast, wherever you listen to your podcast. And we'll see you soon. Thank you.