Where AI ambition meets infrastructure reality

Louise Walden 

Hello, thank you for joining us. I am Louise Walden. I’m a senior marketer at Expereo and our theme today is the very real challenge of AI ambition meeting the reality of infrastructure. We know that every board today recognises the need for AI at scale, but who is asking whether the infrastructure underneath can really carry that weight? This is the gap we’re here to talk about. 

So I’m joined today by three eminent individuals from the IDC, McKinsey, and also Expereo, and I’m going to ask them to introduce themselves. So Martina, firstly, please. 

Martina Longo 

Thank you, Louise, and hello, everyone. I’m Martina Longo from IDC, research manager for the CIO and CTO Buyer Insights program. So when my researcher looks to devolve in CIO and CTO agenda and how technology leaders are turning AI ambitions into value, how they are aligning with these C-suite leaders, and how they’re navigating this changing technology environment. Thank you. Next, Gerardo, please. 

Gerardo de Geest 

Yeah, he’s the partner at McKinsey in Amsterdam. I lead our networks practice. And part of that, I’m looking at the impact of AI on networks and how enterprises work with that. 

Louise Walden 

Thank you. And last but not least, Julian. 

Julian Skeels 

Yeah, thanks, Louise. Good morning, good afternoon, good evening, wherever you are joining us from. My name is Julian Skeels. I am Chief Digital Officer here at Expereo. and are very proud to lead product management, software network engineering and user experience teams. 

And I also lead our solution sales teams who work directly with many of our largest and most complex customers to solve their problems, which increasingly is about how their networks can keep up with their AI ambition. 

Louise Walden 

Great, thank you. So over the next 35 minutes, we’re going to work out where that gap tends to open up. What to do about it before it costs you a deployment issue or a deadline, or even if it turns into a board level problem instead of an IT one. So we’ll take your questions for the last 10 minutes. So please pose your questions in the chat as we go. 

Thank you. So I’m going to start with Martina. You surveyed 800 companies for the 2026 Enterprise Horizons report. You asked them where they were with AI deployment. Can you tell us if this is mostly ambition or is it now real? 

Martina Longo 

Yes, absolutely. And I have to say that it’s real. So there’s almost a consensus around the AI adoption, 91% of organizations that we survey in the technology leaders survey are already implementing AI, you know. So I have to say that we are seeing also from our broader IDC research that the real question is not anymore on should we use AI, but more on how do we get value from AI? And I think that’s the real question today, especially because we are seeing these widening gap between AI ambition and AI real execution. 

So for instance, one of the elements that we are seeing is maturity. So while AI adoption is almost universal, Most organizations are still relatively at early stages. So like 62% of the organization with a survey are using AI only in a limited way, such as a man-business function or on some processes that are piloting yet with AI. Only actually 5% of them describe their use of AI as transformative, meaning that AI is very impacting their business in terms of, for instance, revenue growth or operating models. And another element that I would like to highlight again is what I mentioned before, delivering value from AI. 

So the first gains we are seeing right now from AI also highlighted in the survey around efficiency. So already 90% of the respondents highlighted how they are seeing improvements in productivity, speed of work. Also better quality outcomes from AI decisions, but you know, get some incremental gains in terms of customer experience. But when you look at really hard metrics such as evaluating cost reduction or revenue growth, in this case, the situation is more nuanced. So, you know, they’re seeing time savings from automation, but at the same time, the growth story is only starting to emerge. 

So the 5% that I mentioned before saw the organization that are already seeing transformative outcomes from AI. They’re already talking about growth, thanks to technology. So idea here is that they are investing the efficient incremental gains that I have seen in the past one months into growth expectations, especially these companies in the terms of growth outlook with the uncertainty of economic environments that we are seeing, they’re respecting more revenue growth compared to the peers. So somehow they are really reinvesting efficiency into innovation. But let me say that in broader terms, real money from AI still needs to come. 

And the first consideration is that you need to look to have more money from AI and to close this execution gap to real foundations. So from the CIOs that we have surveyed, when you look at technology priorities, AI, of course, is pivotal, but it’s not number one. So number one is cyber security. Then we have AI, and then we, at third place, we have network data platforms and cloud computing. So foundations are those elements that are really needed to scale AI. 

Let me say this is the scenario that we are seeing. So AI is real, yes. They’re even in already incremental efficiency gain and value. But those organizations that are pulling ahead are not only considering AI by itself, they’re looking at the real foundations to allow AI to scale and to create this long-term business impact. 

Louise Walden 

Thank you very much. It seems like our CIOs here in the report are really concerned about value and growth. I’m just going to turn to Gerardo now. Across your clients, Gerardo, what is the deciding factor when it comes to whether an AI deployment works or whether it stalls? 

Gerardo de Geest 

It’s a great question. Actually, maybe let me first say that if you look at our latest state of AI report, 78% of the organizations say they are using AI already in one business function. However, if you look a bit deeper, like the average of business functions they’re using AI in is only three. So in my mind, it’s not so much about let’s say whether it’s adopted or not. I think most organizations have adopted. 

It’s about whether you’re able to scale it or whether it’s tools over there. And then if I look, let’s say, what’s driving that? If I just look at, let’s say, many organizations that I’ve seen over the past year or two. I think there’s three things that stand out for me. One is I call it insight for rejection. 

A lot of the AR use cases that I’ve seen create amazing insights. Turn models, models around when things break, efficiency models, all kinds of amazing dashboards, but it doesn’t trigger an action. It stores a lot actually at what I call PowerPoint level. Whereas actually what it needs to do is actually trigger the action to deliver an end result and an improvement or impact for the business. The second thing I say is I call it data plumbing. 

You see a lot of use cases coming with a great roadmap, like let’s build a data lake for the next two years and then we can deliver it. That’s, I think now I’m saying it, of course, a little bit funny. That doesn’t work. Actually, the use cases that I see that work are focused on value first and then building the platform later. And the third one, I would say trust and ownership. 

And the comparison I like to make is if you are a network engineer on the emergency network, the last thing you’re going to do is have some random algorithm go ballistic on that live network and make it crash or do something that you don’t want, right? So it’s about trust and ownership in that the model or the POC works. And that’s not done by some data science team building a POC, but not owning the business or the network or whatever it is. It’s about the business owner driving it and owning it. And once that is how that works, that’s actually when AI skittles. 

Louise Walden 

Great, thank you. There’s a lot of content there about scale and impact and then finally trust and ownership. Thank you. So Julian to you, are you hearing the same things from customers coming to Expereo? 

Julian Skeels 

So I’ve just realized, apologies, my headset just played up. So can you hear me now? 

Louise Walden 

Yes, most definitely. Thank you very much. 

Julian Skeels 

OK. So I feel Gerardo’s point about action versus insight is critical here. And I think it’s worth thinking about what needs to be in place for action to take place. And for me, one of the key things is that is the connectivity needs to become, it becomes part of overall the control loop of the things that need to be in place to make this successful and to actually drive those actions and drive beyond. As Gerardo says, it looks great in a PowerPoint, for example, but doesn’t have any influence beyond that. 

And Yeah, so it only creates value when it makes safe action. So, rather than treat it as like a passive utility, it needs, you know, you need real-time visibility and the ability to respond across all of your sites for this to move beyond the pilot, to move beyond the prototype. Because without it, it’s just going to remain a bit of a dashboard. And so, I would say this year, AI is an necessarily creating these weaknesses. In many ways, it’s exposing operational debt that has been hidden for years and AI is exposing it because of the volumes because of the way it’s working because of the sensitivity of how it needs to operate. 

It’s now becoming very, very apparent. And our role often is to help customers understand that. They may say to us, I mean, see that their network impacts on the success of their pilots, but they will talk about your pilots working, but being unreliable across sites or clouds or countries, or it behaving inconsistently or not behaving at scale. Very often, the model is what gets blamed. And what we’re increasingly helping our customers see is that actually this is a much wider issue. 

This is a failure of a complete service chain that AI is now exposing. 

Louise Walden 

Thank you. It sounds like we’re helping our customers really understand what AI application means in the real world for their own business. 

Julian Skeels 

Absolutely. 

Louise Walden 

Yeah. I’m just going to move back to my team if that’s OK. What are the main challenges when it comes to AI readiness, do you think? 

Martina Longo 

I think that’s an interesting question partially covered already by Gerardo and Julian. So let me tell you what the research shows. And I can confirm that most of the challenges are not only technological, they’re really exposing those operational and organizational challenges that companies really have. And so we asked CIOs and CTOs in our survey What were the key elements that supported AI initiative exceeding their expectations? And what instead made AI initiative stall? 

And they are the same. This is quite interesting, you know. We’re very happy that we both were coherent with one another. And the first one, we mentioned that before data, data quality. So as you know, good data means good AI. 

So those companies underperform mentioned that the quality was a key challenge. Whereas high quality training data was one of the strongest contributors to AI success. And I mean, this is an interesting fact. It’s not you, but it’s still there. And the second is a hot topic right now and is all the conversation that we are having around, cost and ROI. 

So we have like 50% of technology leaders saying that the initiatives that fell short, that was related to cost exceeding the return. And almost for intent leaders say that the biggest concern right now is losing visibility around the AI costs and around the ROI they’re creating when they’re scaling these AI initiatives from one business function to another. And for those that the minority that still needs to adopt the AI cost theory remains a key obstacle for starting the AI journey really. You know, what is clear is that right now technology leaders know that both spending around AI models is on the solutions for getting value from AI is just amplified in efficiency they are seeing. And sometimes even we are seeing that for instance, many European companies are starting to quiet a little bit on AI investments due to these ammet expectations. 

So really, this is a key issue right now. And then the third one is, again, own but good talent skills. So most organizations struggling with AI again, they are pointing to a lack of in-house skills. Whereas like 50% of the organizations seeing positive returns from AI, they credit strong internal skills as a key enabler. I have to say that 10 gaps are again related around the main tech priorities. 

that organization are investing in 2026 and beyond. So again, AI first was cyber security is there, data is there. So that’s really an issue again, and the outcome would be also to have digital initiatives delayed, not only AI initiatives. And then we have two interconnected components, so security and networking again. We are seeing that security is a key challenge for companies that are accelerating on AI investments, such as, you know, the transformative companies that I mentioned before, the minority that are already accelerating on AI, actually 100% of them have seen a rise in security incidents in the last year. 

So that’s a key concern again. So the faster you move AI, the more important it becomes, of course, to build security by designing to your journey. And networking is again there. So, you know, networking is often overlooked. But again, I was mentioning before foundations, and when you have an inadequate network, this is again a common reason for AI programs to fail. 

And, you know, those companies, again, that are succeeding are crediting external network performance, again, as one of their components of success. But overall, only 5% of CIOs mentioned that their networks are ready and fully prepared for future business needs, which is quite concerning, I have to say. And let me just conclude very quickly. It’s not only about bandwidth. It’s also especially related and linked to flexibility and resilience again. 

So foundations first, talent skills, data, governance, and then you can scale a lot. 

Louise Walden 

Thank you. I are the Horizons report will clearly go into that in more detail, but there’s some really interesting insights around investment, return on investment and cost and visibility, I think. So thank you. And you can click through and download the report that we’re referring to and that’s in the chat. Gerardo, can I move to you? 

What happens on the business side when things go wrong? 

Gerardo de Geest 

That’s actually an interesting question. I think maybe let me first say, people always think about loop when things go wrong, it’s a waste of money. But that’s actually the smallest issue here. I think the real damage is much bigger. And I would say, look, better ever clean failure, that’s actually cheap. 

That’s stored success. That’s what’s going to be very expensive. And let me maybe say a little bit more about what I mean by that. I think, first of all, there’s loss of credibility. If you’re the CFO of a stocklisted company or something, you have a churn model that you funded and it failed. 

First of all, you look stupid. But second, a much more important, right? You are the CFO that is funding all these future AI products for the full business. So the next 10 or 100 are going to all suffer from that one failure that is still in the back of your mind and really raising the bar. And probably a lot of them are going to be stopped for not the right reasons. 

The second point I would make is maintenance track, and that’s what I mean with that stole success. Are you gonna have it sit down there for another couple of years? It’s gonna waste more AI tokens, it’s gonna waste cloud, it’s gonna waste maintenance, it’s gonna waste people’s energy, time, complexity. You end up with a graveyard of field projects later on. Better cool with a fail fast than actually letting it drag on. 

And the last point I would make is, let’s say, refer the trust. So if you have some automation, let’s take a mobile network, right? And you’re maintaining looking into that. And you’ve built trust in all kinds of AR use cases. But then you have one that misfires, and it takes the whole of, let me say, Amsterdam down on a Saturday afternoon with a lot of people, right? 

As a network engineer, within one second, you lose all the trust in AI and everything. And that’s really going to basically take the hurdle very high for the next couple of years to deploy anything in AI on a live network. So it’s very tricky there, and that’s very important that you’ll be careful of. 

Louise Walden 

Thank you. There’s some really wise words in them. I love some of your examples. Thank you. Julian, where do you think the readiness gap shows up first, though? 

Julian Skeels 

Yeah. I mean, I really love Gerardo’s point, but, you know, Gerardo’s point that it’s better for something to fail than it is for it to kind of, it’s still success. And I think I really agree with that. And I think it can have even more sort of layer on top of that. There’s a bit of an uncomfortable truth here that most of these developments, most of these pilots were actually designed to pass in the first place. 

They were put on rails. They were set up for success. They had the best data. They had the best, most connected users, hyper-care support all around those. They were put on rails to be successful. 

They proved that the model worked. did the pilot ever consider that the enterprise would be able to operate it? Was that part of the success criteria? And I think if we’re very, very critical, and I speak to myself as a former software engineer and things like that, we love things to be successful. And perhaps when we’re very quick to ascribe the failures to It’s somewhere else. 

It’s not me. It’s somewhere else. But I think sometimes we have to question with this, what the measures for success really look like and make sure our pilots reflect that because when we say, where does it actually show up? It tends to happen when it gets out that when the thing gets out of the lab. And the symptom is typically variation. 

It’s not, again, it’s not complete failure. It’s that annoying a flaky kind of failure, which is it works well. Some places it doesn’t work well. Other places, you know, from a simple point, you know, a copilot or some other agent, you know, takes two seconds in the HQ and it takes 10 seconds in a remote branch. For example, or data is arriving late and transactions are timing out. 

And it’s that kind of flaky performance that is hard to really nail down because under all of this, is a very complex network of circuits, routing, overlay technologies, all of which have got their own version of the truth, all of which are trying to point that the problem sits somewhere else, all of which make it really, really hard, unless you have that singular view of everything, to understand where is data flowing, how well is it flowing and under what policies Is it flowing and do all of those things when we engineer them as a complete system? Give that pilot look brilliant in the lab a chance of being successful. 

Louise Walden 

So if we go back to our CIO then how can they tell if their own ambition then has outrun their network? 

Julian Skeels 

Yeah, I mean it’s it’s it’s a good question. I think You know, there’s, you know, as we often in our teams talk about that, you know, as long as we talk about, you know, slow is smooth, smooth is fast. And sometimes we may have to ask ourselves, yeah, these tools, and even myself, you know, for the first time in 20 years, I’m writing code. assisted by tools, and I’m churning out all this sort of stuff at a pace I can barely believe. But probably none of that stands a chance of ever really being successful, unless I really engineer and think through the whole thing. 

And AI doesn’t get us away from having to really think about the full scope of the impact of the change we’re trying to generate, which extends beyond models. extends into networks, extends into people, and as well, and the changes that happen to think about that as an entire engineered system. So I think there are questions that we as leaders can ask of people trying to drive this change. You know, business cases, these things tend to be, dare I say, a little bit thin. It’s going to be, you know, if we do this, we can do some kind of cost avoidance over here. 

For example, we’re not really, you know, does that business case really identify where the users, where the data, where the dependent systems are? Does it really think about, you know, things like the response time, the reliability that’s going to need to be in place in production? in production where it really matters. You know, and if you don’t then, you know, from a business case, you’ve just got a bit of an unaccosted infrastructure to sit, you’ve got an unaccosted infrastructure decision. I’d be asking CO2 hospitals, can they really seek, do they have the visibility into their networks to be able to see whether some is actually working or not, and will they be able to see that in real production systems? 

Can you have one view of where your data is flowing and how it’s performing? And then finally, and I think this is probably most importantly, it’s ownership, who’s responsible for this thing? In any system where you have, a complex combination of models and data and networks and people and locations and other technologies. Who’s responsible for maintaining the service? Who’s responsible for restoring it when it fails? 

Or are we going to get in a world where you’re going to get a conference call and which every provider is going to claim it, their stuff is fine and it’s somebody else’s stuff that’s gone wrong? And so if your plan just assumes all this stuff will just magically work, just because you can magically say, you magically produce code. If you believe that, you know, you’re just going to magically make all this change happen. It’s not a plan. It’s just a really big, unpriced risk that you’re taking on. 

Louise Walden 

So, I mean, Gerardo, is this a technology problem or a planning problem? Do you think? 

Gerardo de Geest 

All right. Maybe first what I’m thinking is look The sign that your ambition has outrun your network is really that you’re negotiating with your own infrine instead of building on top of it. And that’s usually not a good sign. And then you get all these great new AI use cases. But they can only be built if you have some kind of bespoke network project coming on and it takes another six or 12 months, right? 

And you get some massive roadmap against it. And it just takes forever. And you basically build up this whole massive wall to prohibit you from actually creating value with AI. So I think that’s just a dangerous thing to do, right? And you end up with basically great pilots, but then once you start scaling them and want to deploy them in a big enterprise, they don’t work anymore. 

So it’s a trickery or a slippery slope. Let me put it like that. So maybe back to your question, because I’m dwelling out a little bit, right? Technology versus planning, right? I think my experience is first of all, if you think it’s a tech problem, it’s usually actually a planning problem. 

It was just that symptom wise, it feels like a tech challenge, but you need to buy something and you need to buy it now and whatever. But in reality, You just need to plan better or you need to change your thinking and actually build on top of the infrastructure that yet you have already today. 

Louise Walden 

OK, thank you. I’m going to just go back to our CIO in this situation, Julian. If they recognize the gap, what should they fix first? 

Julian Skeels 

Yeah, I think to Gerardo’s point, it’s very easy to boil the ocean on this. I feel that you’ve got to start with a complete across the board, you know, network upgrade, and you can sink a lot of cash into doing something like that. Chances are by the time you get to where you want to get to, the world has moved on, the requirements are different and, yeah, you start over. So, you know, I think And again, you know, earlier, start with one or two really high value business processes, start things all deliver real value and identify those problems that, that frankly, carry the fewest dependencies, the fewest pieces, because it is all about, you know, as discussed earlier, you know, the cost of a failed deployment. is not always necessarily in the value. 

It is the opportunity cost of what’s going to come afterwards. So building up that momentum and doing that by really engineering the solution, mapping where the users, where the data’s are, and then focus ruthlessly on those pieces. Because a lot of this is about momentum. We’re in a very nascent space here where we’re learning all the time. And the minute we believe we understand everything is in the minute we actually prove that we don’t. 

So, you know, start with that, test it from, you know, representative sites, not just from the headquarter, try and remove all the biases, everything that you’ve got that leads you into these decisions of building something that doesn’t scale. And then focus, I say focus ruthlessly on that, because this is a long game. This is technology that, you know, the people who follow us will be working with. We will never have a time. It’s not possible to see a future where this is not a fundamental part of how everybody will work. 

So this is a long game we’re playing here. So the goal is not to solve it all the one go. The goal is to get the momentum, build the muscle memory, identify the areas that you can focus on. You can have impact and real impact and relentlessly prioritize that. 

Louise Walden 

So it’s a marathon or sprint then. 

Julian Skeels 

It is a marathon not a sprint. 

Louise Walden 

Okay, so I’m just going to go back to Gerardo. How do you help clients balance the AI investment against everything else that’s competing for network investment? 

Gerardo de Geest 

Maybe, let me first of all say that I think my first advice is always there’s no such thing as AI investments. It’s about business impact and in use cases, right? So if you talk about look, the machine has broken down. I need investment for a new machine that will always take priority over AI investments. Because I don’t even know what AI investments means if I’m just a CEO of a company or a CEO of a company, right? 

So if I continue that talk, I would say, look, first of all, built on a use case. So I talk about something like predictive maintenance, voice driven call center, or in whatever business you are. But what does that say, the business impact of it? Because then it’s much easier to prioritize an investment, right? Second, I would say, is tie it to the lifecycle that you’re in. 

Because if you’re doing, let’s say, bespoke upgrades or investments, it’s always much more expensive, much more difficult, much more tricky than when you’re doing something already, and you’re just piggybacking on. And the third thing I would say is be honest with yourself, but also be honest about the horizon. Because I see a lot of investments like we need to invest now in this big thing. It will come over two years and we actually need it in five years. And as Julian was saying, especially at the speed the world is moving now in one year from now, the world will look very differently. 

So folks can actually focus on the near-term successes built on that and get from that actually the money that he needs for the future in this month. 

Louise Walden 

Thank you. I’m just mindful of time, everybody. So, Martina, I’d love to come to you for your close discussion a little bit here. What’s your final thought given everything that you’ve researched and what you’ve heard today? 

Martina Longo 

I think I’d be quick because a lot of items were covered already by Gerardo. I enjoyed it. I agree with that. Like, AI is not really about AI investments. It’s about business and business investments, of course. 

such as business use cases. You need to prioritize your business use cases that could be linked to AI investments, for instance. And then build what we call, of course, these execution discipline, of course. And you need to be worried about everything around these investments before committing to that. So you need to be deliberate about which use cases to scale and how to engineer for the real operations behind it. 

networking again. So yes, an important part of this equation, you need to consider that the working should be part of the infrastructure should be treated as part of the architecture. And of course, we have highlighted how that work is as critical to AI. And you know, if you look at what we call the AI fuel businesses of today, they rely entirely on high-performance network on secure and reliable networks. But beside that, I would just close off with thinking about the connection with the business and the C-suite. 

So again, in the research, we saw how, and that should be common sense, but not always, how a closer alignment between the CIO and the CFO, for instance, improves, of course, the access to the funding, This piece is the investment decision and also helps balance the innovation with course cost management. And so again, we are seeing that if you think about governance, governance is a continuous discipline as a competitive advantage as well. So most of the companies that are ahead in closing these AI and working gaps, they have a center of excellence composed by business and technology leaders where they actually focus on a defined set of priority, business priority, to define the technology initiatives needed. They have a shared ownership of funding. They have defined value metrics and also risk assessments and governance. 

So that’s more of the recipe for success that we are seeing right now on the market. 

Louise Walden 

Well, thank you very much. We’ve actually got a couple of questions. So I’m going to go to Gerardo for the first one. And I’m going to read it out loud. So it would be great to hear the infrastructure and reality and how customers must improve, I suppose, to to receive that impact? 

Gerardo, can you answer that one? 

Gerardo de Geest 

I can give a perspective. I think first of all, the infrastructure reality is widely different across enterprises. I’m not sure if it’s the biggest driver for impact first, but let me first say, look, I’ve worked for enterprises where I was just flabbergasted by the infrastructure. The spend was massive. Everybody agreed, even the CIO agreed that it was ridiculous. 

But the idea was the business is asking for it. And then if you push back and it goes wrong, I mean, it’s a good way to get fired. So you don’t and you just spend the money and you put that infrastructure down. That’s great. And yes, that enables great AI use cases. 

Also amazing. At the same time, every infrastructure, also this one, has its limitations. And we’re always able to find the use case that’s got its limitations. So for me, the much bigger catch is how do you go around with these limitations and how you create a culture in your organization to think about what is possible? And what can I do with that infrastructure today? 

And even the worst infrastructure that I have seen, we were able to deliver great impact through changing that mindset. So for me, the real question is really, what can you do with that infrastructure? And how do you help the your business owners to get the most out of it and drive to write priorities? 

Julian Skeels 

Yeah, I mean, if I’m jumping through, I really agree with that as well. I mean, I think historically IT teams, network teams, things like that are being kind of seen as being this kind of like isolated groups, unaccountable groups there, I say. And in some ways, actually, some ways they quite like that, you know, with the, probably the lack of understanding, the lack of inspection, the lack of accountability. all of that, it’s kind of almost like an unwritten agreement, right? But I think this is a time now for when we’re talking about such interconnected systems where the success of one impacts the success of another. 

This is the time where the great businesses, the great leaders of infrastructure, IT services, things like that, it’s the time to get out of the chair and get into the business and to really understand what’s happening and to really engage in a very, very authentic way. Moving away from a team who’s what they deploy isn’t assumed but not understood capability, which is tightly integrated into the success of the business with shared skin in the game and things like that. Those are the organizations that will be successful. Those are the leaders that will be successful. I think it involves a very different type of engagement to make that successful. 

Louise Walden 

Thank you. Our second question is for the network manager that reports into a CIO. What advice would you guys give them? to be on top of the impact of AI on the network. And I’m going to come to Martina first, because the research certainly helped provide that advice. 

And then probably Julian, if you can probably having real life experience also, could help with that question too. 

Martina Longo 

Sure. But I think that Gerardo would be most technical than me in this case. What I can say is that, of course, the CIO has many competing priorities right now. So they’re asking him to be those business orchestrator, implementing KIA scale and delivering new value and so on and so forth. So if you’re a network manager, you need to be aware that networks, as we saw from the research, you have to be on top of resiliency. 

You need to be aware of cybersecurity issues. But most importantly, you need to be able to communicate effectively to your CIO that needs to talk with the CEOs and the boards of what are the potential risks behind having a network that would fail. I think that’s something people thought. So you need to, again, be able to speak the business language. When you’re implementing AI with all the different increasing workloads and so on and so forth, what are the risks behind a network that would fail? 

that I, something that I would advise a network manager. So be technical, but also be more focused on the impact on the business. Thank you. Julian? 

Julian Skeels 

Yeah, totally agree. Lead with the business outcomes. Work with, again, get out of the chair, work with the application of business owners, understand what response times need to learn, what reliability needs to look like, what the cost of failure is, and then work back from that start, start with the use case, start with the value, and then work back to what is needed as an engineered system to be able to deliver that. And take ownership, demonstrate clear ownership. So how do we get out of this world of multi-provider blame loops and things like that? 

Kind of agree how these things get monitored, diagnosed, things like that. This is where This technology has the impact to have huge effect on business outcomes, but it involves, you know, one of the challenges of, you know, for those of us in a, you know, a technical discipline is that we have to speak in different ways to different people. And certainly when we’re working, you know, remembering a CIO is a business function. So we need to speak in a business language about business outcomes and impacts. 

Louise Walden 

Thank you. And Gerardo, you would agree? 

Gerardo de Geest 

Yeah, let me take also a different angle that might be helpful for this question. I’m always thinking, let’s say, in a framework of three angles to look at AI on a network. The first one is AI use cases that simply run over the network. And it can be as simple as you opening your favorite AI solution that’s a cloth and asking you the question and expecting you to come back with an answer in a certain timeframe, right? And you’re using the network for that. 

That’s the first and more of more of your network capacity will go to these things, okay? That’s number one. The number two is AR use cases that actually are built on top of your infrastructure. You can think about actual use cases that use some of your sites. You can be thinking about models that you really train for your enterprise that require certain server capacity or whatever you know it’s right. 

So that’s a second element to think of. And then there’s a third network element which I call AI Food Network. So what AI use cases can you as a network manager think of that actually help you drive the cost down or improve the quality or any other metric on the network. And then of course, taking what Julian and Martina said as well, along these three buckets, think through what are the business outcomes for that. 

Louise Walden 

Thank you very much. I think that’s a really good place to stop. So I think it’s time to wrap up. Thank you so much to Martina and Gerardo and also to Julian. Like I said before, we’ve added the link to the Horizons report that Martina authored. 

So please take a look and we’d love to hear what you think about that. And also please get in touch if you’d like to discuss anything that you’ve heard on the chat today. And if you want to talk about anything to do with AI or infrastructure. Thank you and goodbye.