February 2025

The future of AI: Key trends to watch in 2025

Bjarke Arreskov-Hansen
VP Business Development & Partnerships
AI Trends
AI Governance

Welcome to AI Watch Episode 23!

In this episode, we delve into the future of AI in 2025 with our experts, Bjarke Arreskov-Hansen and Peter Søndergaard. They discuss the significant transitions in AI architecture, the evolving role of AI agents, and the impact on strategic software providers. Additionally, they discuss the potential for cloud exits and “silicon shoring” due to geopolitical risks and the changing cost dynamics of software development. Tune in to learn how organizations can navigate these changes, manage AI governance, and prepare for the exciting developments ahead.

Transcript

Bjarke: Hi everybody and welcome to this session of AI Watch. With me today I have the pleasure of talking to you Peter who has joined us today to talk about a bit of an outlook into 2025 and what we can expect of AI. So welcome to you Peter. Peter: Thank you, thank you. Always a pleasure. Bjarke: We are going to kick it off with your perspective of where this is leading. So I wanted to sort of get your take on what do you see of the biggest differences between what we’ve seen with AI in 2024 and what we can expect of it in 2025? Peter: Yeah, I think I mean… it’s very clear that now we’re kind of in the scaling phase of things for most organizations. And the second thing that’s also very clear is it’s almost impossible to have a unique view of how AI is going to evolve because it seems like everybody has a perspective and everything is kind of known. But I would say then that there’s probably a couple of things that will be new for 2025. Not surprising, but new. I mean, I think the agent AI and the role of agents across organizations. It’s very clear through the product announcements that we’ve seen from some of the technology players will actually be a substantial topic to which organizations should need to determine what therefore is their architecture stack. How does it change their architecture stack? And I think we will, therefore, through the year, start to consider what’s our AI architecture stack and how that impacts the overall technology architecture that an organization has implemented. So, this push and pull from, do we drive more over towards the architecture stack view of the large emerging AI providers, or do we simply embed this into our existing architectures? And in fact, how does that scale? The other major trend I think that we will start to see become more prevalent as this aspect of how does it actually change the existing large strategic software business process software providers that we have implemented. So the SAP’s, the Oracles of the world, because 2024 was about them announcing what they were doing, which is really their software product plus AI. 2025 is really the software product with AI. And it also lays the groundwork for how do we actually go in and re-architect so that our application is architected with AI as the core. And that will drive a replacement cycle in software, not in 2025, but a few years out when we will start to see companies opening up for the possibility of changing some of their strategic applications because new software providers have emerged. And so, 2025 is a transition year because once you scale AI, you start to consider transition, not just in the year, but also prepare for things that move forward. Bjarke: Wow. So quite a fundamental change, this plus AI to actually AI at its core. Where do you see that sort of, if I’m a CEO or a CIO, what kind of questions does that raise for me? Peter: Well, it starts to raise this as a classical thing in essentially IT procurement and IT strategy and architecture is you have to start to ask yourself, how well are my strategic software providers positioned to exploit this? And do they do that in time before you start to see new emerging providers that may enter the market? We’ve seen this. This is a classic thing that you see. You’ve seen multiple cycles in the ERP marketplace that has played out this way over the last 20 or 30 years. And so this introduces another cycle of that. And it introduces a question mark around your architecture. as a CEO, you’re going to get requests at the board level, at your level, for large-scale infrastructure investments that you do for technology and software. And so now you have to ask the question, the provider that we are selecting, do we believe that they actually have an ability to transition through what’s happening now? It puts more pressure on the procurement department because they’re the ones that actually have to evaluate this. And of course, as a CIO, it means I have to go back and revisit my technology strategy. And at least be certain that the providers that I have based the primary part of my technology architecture on that they’re actually able to move me forward to the next couple of years. Bjarke: So that’s if I look at it in the context of enterprise software, you talk about that. If we look at the providers of say, language models. Do you see a similar thing here where it kind of comes as a standalone? Peter: Well, think, I mean, I think the large language players obviously play in the market that’s more consumer oriented, as we’ve seen. I mean, that’s why they’ve grown so popular. But they also will need to now distinguish what they’re offering. And I think you’re seeing that from OpenAI as an example, in terms of some of the agents that they’re introducing is that some of them are directed more towards us as individuals. Others are starting to be more directed towards assisting in business processes. And so I think they’re equally playing. And I think they are the unknown in a market where you then also have classical technology players that have an installed base out there. And so I think it’s still too early to speculate what happens because it’s not the first-order impact we should focus on. We need to speculate on what second and third-order impacts of this would be. Bjarke: If we look at this transition, I imagine for some it can be almost a bit of a, you get into a bit of a paralyze, a paralysis if so much is changing, what am I going to choose? Do you see people take a backseat in this or lean in? Peter: I think for larger organizations, I think they’ve already built sort of what they believe would be their strategy, obviously something that always can change. not necessarily seeing paralysis among organizations. I think organizations do suffer and will continue to suffer from the inability from time to time to execute things at scale inside their organization. And so where you could talk about paralysis is that there’s a POC paralysis in some instances. Is organizations don’t get out of the number of proof of concept or pilots that they’re running. They find it hard to select and then really manage through the solutions that they want to scale. That happens in some organizations and that’s potentially paralysis that we could see. I don’t think it’s sustained because companies usually eventually figure this out. But obviously in a world that’s moving faster, you don’t have the luxury of time. So I would strongly recommend organizations to really focus on how do we effectively manage our POCs. And POCs is like managing an innovation process. You’re basically managing an innovation from ideation all the way through to scaling and deployment of this. And because artificial intelligence changes so many things in an organization and has so many opportunities to do innovative things for the organization, the number of options are exponential. And that means that you need a structured approach to manage this because you’re not managing two, three or four options, you’re managing hundreds of options of things you could do in an organization. So if you don’t have a process in place now, you will not be fast enough. And it’s proven that people that put together processes for managing innovation do far better than people that have a haphazard or unstructured approach to innovation management. Bjarke: So being structured is important here and I guess in a dynamic world if you don’t bring the structure then everything will just be very dynamic and you kind of get nowhere. Peter: Yeah, mean a little structure sets you free. Bjarke: I like that phrase. So one thing about this and organizing it a bit, I saw Jensen from Nvidia recently come with a statement that IT will be the HR of AI agents in the future. How do you see that talking about structure here? Peter: Yeah, I mean, I think it’s an interesting proposition. And I think he may have meant something specific with that. AI agents is essentially software. IT does have the primary responsibility for coordinating the usage of software in an organization. So from that perspective, yes. But increasingly business units are the ones that actually manage what software does and which software is relevant for solving their areas of responsibility. And so if his view was narrow, that software is coordinated and I choose that word carefully, I don’t say owned, coordinated by IT, then he’s absolutely right. Will IT become HR in the future? No. Because HR, and I also choose this word carefully, coordinates people and response to people in an organization and managers are accountable for those people and ensuring that they work. They are also accountable for the goals of the organization. And if those goals are fulfilled by agents that act and perform tasks of people, then ultimately the manager needs to be accountable for it. You can’t make IT accountable for it because otherwise IT would perform the function of the CEO. Bjarke: True. Peter: Now, is there going to be a change in the balance between what’s done? Possibly. Will it vary by vertical industry and organization? Absolutely. Because nobody has the same approach to how they do processes and how they manage and the metrics that they measure and what have you. but to be definitive in a statement that says, which is not really what he said, but if you interpret it that way and say, it will do the work of HR. That’s not correct. Bjarke: I agree with that, but I, but I do think it raises a point that it seems AI is not just a technology that is being implemented in the classical sense, but it’s moving to something that is being employed in a sense. Peter: Yes. Bjarke: To take on tasks, to act on behalf of, and that raises all kinds of questions of how do you lead this technology? And then it’s not a CIO question, then it’s a broader leadership question. Peter: It is. Bjarke: Where do you see that? Peter: I think, so all this being said, I think the HR function needs to… the leadership of HR needs to upskill massively on technology because they need to be as versatile on what agents will be doing in terms of which tasks and how they perform and how you monitor and govern what they do. so HR has a significant task on its own plate to upskill what they do and certainly from a leadership perspective. We also have a requirement to upscale leaders generically in an organization, leaders and managers. And that’s something I’ve been pointing out for years is every manager will manage both people and agents. In fact, it is likely that we will end up in a place in which every human decision is AI supportive. And if that is the case, then every one of us and every manager that manages us would need to be knowledgeable about how we coordinate that in the most effective way to the organization. So this does truly move how we look at labor and work in a way that I think is hard to predict because there are so many external factors that also pull on this. There’s a theoretical linear thought process that just basically says yeah, it’ll take all work But there are so many other unknowns in that equation that I think we don’t really know today And I and I think you can produce theoretical models about this you can produce reports But once you see this technology hit an organization and hit how we human beings react to it I think the outcome will be different. Bjarke: Yeah, and hard to predict. Peter: Very true. And then, I mean, you didn’t ask this question, but then I’ll jump in with it. I mean, it obviously changes the culture of the organization because every which way we look at definition of culture as a combination of strategic ambition for the organization, the processes that define the organization, the mindset of the people, every one of those elements change. mean, our processes change the way we work as people and therefore how we think changes. So culture will change. And evolution of culture is hard for organizations to deal with, because we all are most comfortable with the present and the constant. And so this will be an interesting transition where, again, HR needs to be very involved in this. And we need to make sure that we monitor what these software environments do. So this aspect of AI governance, which has been portrayed as being something that is very much about regulatory environments. It’s really about managing people. It’s about managing the business, the business model. It’s about managing your brand and what you stand for. That’s why you do AI governance. It’s not because the European Union has put together some legislation or you put together some legislation in California or somewhere else. It’s because you have, a as a leader and a responsibility to manage the organization that you are accountable for to the investors, whether or not you’re private or public. Bjarke: So the discipline of AI governance needs to be elevated from a compliance to a broader leadership exercise. Peter: I think we will end up there. Yes. I think right now it’s becoming well embedded inside the compliance department and inside really managing risk in particular for public companies. Because it isn’t just about safety and ethics. It is also about brand financials, customer relationships. So it’s much, broader than I think people have portrayed it to be. Bjarke: And I do agree. think in client conversations, we see the representation of my company as an agenda that often takes lead in the conversation and of course, legislation and so on is important, but the concern tends to start there really because we are hiring these machines to represent our company and it’s a very relevant concern. So I think you are right. Governance is becoming a leadership task. Peter: Correct. And once it does that, it sits in between the CEO and the board of an organization. And so they’re accountable. In fact, they’re legally accountable for this. so ignoring it is not an option. Delegating it is an option if you are sure what it is you’re delegating. And that, course, requires you as an executive leader or a board member to be aware and knowledgeable about this to a degree that you know what you’re delegating. Bjarke: It’s going to be interesting to sit here towards the end of the year and see where the world has moved on these kind of things. One other topic that I was curious to touch upon. I read an article recently in Computer World, raising the question of whether cloud exits was beginning to get a big sort of attention. And I mean, why bring this up? Well, A lot of the AI has really been driven by some of the hyperscalers and they’ve been providing a lot of the exciting models, especially in Gen AI. Where do you see such a question come in and what does that mean in a boardroom? Peter: Well, it’s a risk. Again, it goes back to sort of risk and managing the cybersecurity aspects of an organization. So this has to do with sovereignty. Where are you headquartered and what environment are you headquartered in? What is it that you believe is important to manage from a perspective of risk for the business? And what does the political environment dictate that you need to do or potentially might be need to do, right? And the issue right now is we live in a world in which all of these elements are highly volatile everywhere, really. One tends to focus on one person being responsible for this, but it’s really a volatility that exists everywhere. That I think will drive certain public sector organizations that have a requirement to manage the nation and what the nation does to really consider a shift from a cloud strategy, potentially considerate. I think it will require some businesses for some process functions and some data to look at, well, maybe I go on-prem or remain on-prem with certain things until I understand with greater clarity what is happening in the world. There is another factor that is driving this as well, which is AI related, which is that we’re seeing a change in the cost of labor. Things that were more efficient to do offshore because of lower labor costs, in particular for certain Western environments, may now in fact be just as cheap and less risky to do near shore or on shore or even within your own entity. So what one could call silicon shoring. So I’m basically putting it in silicon in the environment that I operate in. And so there’s a cost dimension to this that is also different than it was when we saw, you know, just going back five years, because now I can, the productivity of my developers potentially could be much higher. And so the cost level is evening itself out. All of this is obviously, again, one is trying to speculate on first order impact of something. And it’s never that that will surprise us. It’s always the second order impact. And so I think it’s very critical for us to follow what goes on here for businesses. This has to do with cost and risk because they can really change the cost equations for things that they do. And of course, risk is a question of the factors that govern the world we live in right now, which has higher degrees of risk than it had 10 years ago. Bjarke: So that problem or potential problem. And also the opportunities. How does that translate into like a question? Again, let’s say the CIO or the CEO turns over to procurement and say, okay, how do you prepare for these scenarios? Peter: So I think the CEO would look at this as a way of first, can I do things that I have done in a particular way? Can I do that as cheap or maybe even cheaper by shifting my resources around the world? So I look at it from a cost perspective is the cost equation of software is changing. And so does that mean I can do something differently from a cost perspective? I then have another dimension is I’m accountable to manage the risk of the business and I report to the board around that. So given that you have higher risk, where does that risk translate into my technology deployments? And how do I consider actually managing that risk? Because obviously there’s always going to be some risk. The question is how you manage it. So that’s the CEO. The CIO has an inventory requirement to really understand, you know, are there things I can do differently? And I think the big task and CIOs have been focused on that all of last year as well is what’s the impact on software development for my company or organization. And so you need to now look at really how does that impact budgets and opportunities and relationships with external suppliers. And so I think this will be an interesting year in which some of the work that happened last year is now going to migrate to being execution from a CIO perspective this year. And that means procurement, for example, needs to look at things differently. Bjarke: But also a big transformation potentially of organization if you really shift the way you operate and Silicon Shore as you said. Peter: And I think this also places some responsibility on the CFO is that all of this has tremendous implications for how you put budgets together. Because I think you could potentially see significant movements of budget from department to department. But you would also need to monitor that actually the outcome of the promised projects happens, and that’s something that CFO spends a lot of time on so you need to also have the checks and balances that actually mean that you monetize the benefits that are articulated, right? Just because I save 40% on software developers because they’re more profitable and productive, sorry, doesn’t mean that people actually do anything with that. It could just mean that the software developers do other things, for example, which they could. I mean, you have a technical debt that is huge in businesses. So maybe you can just put the extra resources on actually addressing your technical debt. But you need to monitor it as a CFO. What happens? Bjarke: Yeah. So you have clear KPIs of what’s right. Use this potential uplift in productivity towards business goals. And that’s just for stuff. All righty, it’s gonna be an exciting 2025, I have to say, and I can’t wait to sit down in, let’s say after summer sometime and have a look of how much of this has actually materialized and also what these uncertainties turned out to be. I really wanna thank you, Peter. This was another exciting session and thank you for all of you for listening. We look forward to see you at the next AI Watch.

More news

Get the latest news

Stay up to date on our latest news and industry trends