The Hidden Dangers of Shadow AI

Enterprises hold growing volumes of connected-device data, yet many are still stuck in early experimentation. The gap isn’t the technology, it’s the readiness of the workflows, processes, and skills that determine whether AI can turn IoT data into meaningful outcomes.

This episode explores:

• Why shadow AI is creating unseen risk
• How internal processes block AI-led progress
• What teams need before scaling automation
• Where IoT data adds unique value to AI models
• How leaders can move from experiments to results

Tune in to hear from Nassia Skoulikariti at Apiro Data about the shift from selling raw data to delivering actionable insights and outcomes.

Transcript

Intro: You are listening to IoT Leaders, a podcast from Eseye that shares real IoT stories from the field about digital transformation, lessons learned, success stories, and innovation strategies that work.

Nick Earle: So, in this week's episode of IoT Leaders, I'm talking once again to Nassia Skoulikariti. And, yes, that is a Greek name. She is a very interesting and talented lady. She is the founder of a company called Apiro Data; you'll hear about that.

And also, has created, free content on the internet to train people about AI. The theme, for this is where we are in the collision, the merger, the coming together of IoT and AI, which is a very common subject being talked about a lot.

The practical implications of all of it, in terms of where people are in their process, what mistakes they're making. Nassia does a very good job of a three-stage framework to think about. And then we finish by talking about what implications this is going to have to all of us out there who have IoT in some way as part of our responsibilities.

So, it's very educational, it's from somebody who's right at the front line, somebody who has worked very closely with IoT over many, many years, and is active at a lot of the workshops, a lot of the teams that are working on the standards for IoT at Mobile World Congress. Many of you will know her, and,

It's an eye-opener on the world that's coming towards us as AI and IoT start to blend together. So, here it is, and I hope you enjoy it.

Nick Earle: So, Nassia, welcome to the IoT Leaders Podcast.

Nassia Skoulikariti: Nick, it's a pleasure to be back. I'm really happy to be here and having this chat with you.

Nick Earle: Oh my, and we've known each other for many years. Typically, we bump into each other at conferences in Barcelona, which are often running the breakout groups and things like that, but today, it's a little different. Today, we're going to talk about your business, and what you're seeing in the AI space and the merging together of AI and IoT. And the company, that you're the founder and CEO of is Apiro Data, and, because you're Greek. I suspect “Apiro” means something in Greek, so maybe we start there. What does Apiro mean?

Nassia Skoulikariti: Well, Apiro means infinity. And, when we, you know, set up the company going to seven years ago now, we were heavy into the IoT space. So, infinity, infinite possibilities, connecting everything to everywhere seemed appropriate. And then I came up with that name, actually, during a holiday in Ibiza, by the pool, a few cocktails in, thinking, what am I going to do? I want to start a company. And my fiancé challenged me, and he said, well, stop talking about it and do something! So, I grabbed my phone, and I registered the name, the domain name of apirodata.com.

Nick Earle: There you go. So, a quick GoDaddy.com later, and you had it. Exactly. And of course, as Eseye, our platform, of course, is the Infinity platform, so… But I don't think our founders were around a pool in Ibiza with a cocktail. That's another story, and another image entirely.

Nick Earle: So let's move swiftly on. Let me start this one off, by recapping a theme, something that's reflected very heavily in our, Eseye predictions report for 2026, which is just coming out. And that is that IoT is going to expand sideways into AI and IoT. And just to recap it, and then I want to ask for your opinion on that. There are many people now commenting on something called Sentient IoT. It's got some other names as well. But essentially, it's based on the premise that 80% of all the data that's feeding the LLMs, the large language models, is currently, you know, content, written content, spoken content, video content. And it's been scraped or stolen, depending on your point of view, whether you're a content creator, and is being fed into these models. But the point is that we've got 80% of it already.

And if the effectiveness of an AI model in the LLM, in order for it to be generative, is based on the data that you… you know, it's only as good as the data you feed it. Then the next big frontier is data around… about things, and that directly brings us to IoT, or in other words. The next big wave of data that's going to train the models for AI is going to come from things, and that's not public data, it's going to be things that are owned by companies. So it's private data, there will be some public, but it will be data from things, and some people estimate that that if we could connect everything in the world, the data from things even if it's just a minimal amount of data per thing, could be a hundred times bigger than the amount of data that we've already trained the LLMs on. So that's the… that's the big vision. That's the theory. As someone deeply involved in this area, which we'll get into, what's your view on that?

Nassia Skoulikariti: Well, I agree with it. I think that IoT service providers, connectivity providers, even the enterprises with connected things are sitting in a gold mine, in my opinion, and that's real-time data that can be contextualized, right? Because up until now, unless we were looking at predictive analytics through IoT, you know, to provide that data some context, and that was mainly through machine learning, not so much through generative AI as we see it today. That data gave information, gave feedback. But now, taking all that real-time data, the majority of it will be real-time, but also we have historical data there, and training these large language models, it could be either feeding the larger, models, or our own large, small or large models. And that would bring a different… I would say, entity together, which is the collision between IoT and AI, and that's the sweet spot right there.

Nick Earle: Okay, all right, good. Well, as I say, there's a growing number of people who, believe that and are predicting that. However, there's also a big gap between a belief and the reality of what that will mean and what you have to do, which brings us onto why you're on the podcast. So, Apiro Data, tell the viewers and the listeners, what the company is all about, and what value you give to them.

Nassia Skoulikariti: Absolutely. So, last time I was on the podcast, it's been what a couple years ago, Apiro Data was all about IoT readiness, IoT integration platform, helping enterprises connect devices, looking at data flows end-to-end. But IoT has matured. And the real challenge now is how do we take all those connections and make them intelligent, right? So, we've evolved as well. We started as an IoT company, and now we are an executions company. Apiro Data today is where IoT and AI really collide, and we call that the execution intelligence, the systems that turn connected data into coordinated action. And we really focus, Nick, in three core areas.

Helping organizations, first and foremost, to fix what might be broken internally. It might not be broken, but it needs to be organized differently when we are wanting to take the IoT real-time data, or internal data that every company has, and feed it to AI. So, we look at the operational side of the business, the processes, the workflows, the data. And we make sure that it's ready for what's coming next to create those intelligent systems.

Then that logically brings us to AI readiness. And that's, you know, frameworks and use cases built and designed to provide real value. And this includes, you know, going all the way from seeing what can take AI, where we can use AI, is it needed or not, to training the organization and the individuals to know how to use and operate in the AI world. And of course, we're still in the IoT space, but I like to call it now modern IoT acceleration, and that's where we infuse IoT and AI together to make automation explainable and adaptive. So, when we put all those things together. Companies then start chasing technology, and we start making it work.

So, you prefaced earlier, and I will just put it into the way that I see it of where IoT is going, but I see that IoT gives us the data. AI will give us, and is already giving us, the speed. But without the intelligent wrapper and execution around it, that's where we see the chaos, and the pilots that are failing, and everything else in between.

Nick Earle: That's a nice framework, and as you were saying it, I was reminded I was doing my background reading, and I came across, in fact, I would recommend it to anyone who's listening, there's free reports from McKinsey. And McKinsey have just this month, November 2025, released a big report on the state of AI adoption. And it's not IoT-specific at all, but what it shows to your first point, the first of those three, are the, what they call, you know, internal efficiency.

What they show is that it's still in the very early phases. There's a lot of, experimentation. In fact, if I'm recalling it right, most companies, something like 70% of companies are experimenting, that's the good news. Interesting that 30% aren't. That's another subject. Which direction are those guys looking at? But the 70% that are, they're typically only experimenting, and the data, I think, was just a few months old, because it was a big global survey.

They're only experimenting in one department. On AI. You know, can we make this department more efficient. And then… then when you say, well, how many are… there's some great charts in there, and how many people are experimenting in two departments, in three departments, in four departments, in all departments, and of course, the number tails off. But I think that that's consistent with what you say, is that people are sticking their toe in the water, I'm coming back to your Ibiza analogy here, I don't mean… They're sticking their toe in the water, and they're experimenting, they're learning.

And they're really looking for, sort of efficiencies, which I guess raises the question. They're not doing the second and the third stages. They're not looking at the workflow, they're not looking at creating new capabilities, they're not looking at the external view of what it means. Really just experiment… they're almost like early adopters on the Gartner adoption curve. Would you agree with that?

Nassia Skoulikariti: I see a lot of it happening, like you explained, a lot of the same or similar signals, pretty much in every discussion that I'm having. Somebody comes to me and says Nassia, we need to train our people on how to use AI, right? And through the conversation, you know, then we come to the realization that's not really what they need, where they are at the moment. Rather, what they need is to fix their operational side, and I'm generalizing, I'm putting it into a box there. By operationalizing, I mean, look, do they have processes in place? What happens to the data? Who does what? What is, like, the current KPIs, and what do they want to achieve? Do they have a use case that makes sense for them to test and put out?

Because, right, what I see at the moment, and this is happening across the board, is what we call shadow AI. You know, individuals inside a company are playing with AI, they're using their own tools, they are experimenting, they're learning…

Nick Earle: Outside of the corporate network, in many cases. ChatGPT. Right.

Nassia Skoulikariti: And guidelines and guardrails, which is quite dangerous. From one hand, it's good, because they're learning, they're experimenting, they're trying to be innovative, because at the end of the day, why all this matters? It matters because we all want to make our work life a bit easier, a bit more efficient, right? Save money or make money. It usually boils down to one of those two things.

So, from one hand, it's good, they're experimenting. From the other hand, from the organizational side of things, it's extremely dangerous, because imagine, you know, they are using data, company data, to do their analysis with external systems. A lot of them, they don't even pay for the AI tools that they're using, which means that data that they are using is going to train the models. So, it's extremely dangerous for the obvious reasons.

And at the same time, you know, even if people say, and they have, like, guidelines within the organization of how they're going to use AI, what they're going to do, what tools they're going to be using, then you have the silent approach. Nassia is creating her own agents and workflows for the job that she's doing. Nick is doing the same for the job that, you know, the job that he's doing. Instead of a cohesive organizational approach to AI, we have individual approach to AI.

So, and this is one of the key reasons why we see this discrepancy from, you know, adoption, from trial to adoption. And there is a lot of different things. Even if somebody's trying to do a pilot and put it into a sandbox, then what happens? If you don't have the processes and the operations and the data in line? You put it into live environment? More chaos. More problems. So, I see a lot of that.

Nick Earle: I'm reminded, and I'm old enough to be able to say this, I can say or maybe mature enough to be able to say this. When you were talking, I was reminded very much of when I was based out in Silicon Valley in the early days of the internet, when suddenly you know the idea of creating your own website, putting a website, a company instance out on the web, and letting users interact with your data. And it was very similar in that there was a lot of expectations, because it was such low cost, and AWS was just forming, and, you know, people could do it on their own credit cards, and they didn't even have to put it through the company expenses. And then… and then what happened was, you suddenly were getting issues around governance, security, liability, or risk.

And the point you made about, if they're using public models with company data. They're probably not thinking about the fact that they are training the model, which can be accessed by anybody, including competitors, and so they are opening the door on the data, and you very rarely hear people say that in this… in this sort of, not quite a wild west, but in this mass experimentation that we see right now, where lots of people are experimenting because they feel they can, they feel they should from a career point of view, they feel they can save money, improve processes, they're talking about the fact they won't be hiring as many people. It seems like it can do everything, but the light that you're shining on is that there are implications without a strategy and a policy and a governance model.

Nassia Skoulikariti: Absolutely.

Nick Earle: There are some pretty significant implications of all of this.

Nassia Skoulikariti: Oh, absolutely, absolutely. And it's exactly what you saw back in the day, and it hasn't changed for any technology adoption. It is just a little bit scarier with AI, because it learns from that information, so data leaks can become… actually, it's… they're happening. If you think about it, even, something that, that is not me putting information out there. Google changed their policy, and they automatically allowed, unless you go and you turn it off, data from our emails, and our Google Drive, and everything else to, train Gemini!

We have to go into the settings, into our Google email, and turn off, you know, for our emails, the data, and you know that a lot of companies using Google as their main company email service, so that's a lot of private information back and forth. It's simple things like that that we need to stay on top of in order to safeguard our organizations. And it can go to the opposite extreme, Nick. We've seen it, right? We've seen it that putting guardrails in that are so intense that just the innovation goes out of the window, so we have to find the right balance.

Nick Earle: And the innovation goes out the window, and the hallucination… and the hallucination rushes in, but it doesn't have enough data. It is… it is difficult. So that's the… let's keep the structure going. So, the first area, the early adopters, in fact, arguably, they're the innovators, I think we'll look back. The early adopters and the innovators, there's mass experimentation, not enough thought, being applied to governance policy, security policy, company image, data protection, et cetera, et cetera. So really, really difficult for, as it always has been, for the IT departments to keep control of the experimentation.

And so your company, and we should perhaps have explained this earlier, your company provides consulting advice to people. You do training, and in fact, you, you also deliver this framework strategy as a… one of your key offerings is, you know, let's go in, do… Yeah. Do… tell me… tell me where you are. “Oh, hang on, stop what you're doing, start first with this strategy overall roadmap framework.”

Nassia Skoulikariti: That's the first step.

Nick Earle: We start by looking in of where the company, the organization is at the moment. You know, we have a framework where I call the three Ds. First, we do the discovery thing. Very simple, right? It's self-explanatory. I try to always keep anything to do with technology.

Nassia Skoulikariti: Stupidly simple for people to understand. We cannot assume that everybody understands technology, so… Discovery is the key thing, and within the discovery, that involves looking at systems, so the complete technology stack, looking at operations, looking at processes, looking at how people using what systems they're using, looking at the skill set of the people, and we do that, also interview all the key stakeholders to understand where they are, and from there, you know, then we can go into the design phase.

Then we understand, we need to understand, okay, the CEO wants to add AI to the contact center, as an example. Is this the number one priority the team is working on? You find, usually, that any given team in an organization, they have anywhere from 5 to multiple, you know, 15, 20 different projects. So, what is the priority? Then we put the projects into priorities, and we look, can AI actually help with this workflow? So, we're looking at the different workflows, and we're starting with one thing. It can be into the product side, or it could be on the internal side.

And this actually reminds me that it's very key to also explain, you know, I simplified, again, the way that I have scaled now look at AI. It's the internal AI. You know, this is the AI inside. That's where we start using AI to simplify our own operations. That could be predicting maintenance, for instance, for IoT people, intelligent routing, we're looking at anomaly detection, and this is where we're looking at margin protection and reducing costs and making the workflows and how people are using and working and operating on a day-to-day basis, you know, more efficient.

You know, that's not to get rid of people, per se, it's just to optimize what they're doing, so that we can free up their time, so they can do more strategic, more innovative work, right? So that's the inside AI, that's the internal operational AI. Then we have the AI as a differentiator. That's the product augmentation.

Nick Earle: Sorry to interrupt you, so this is on your three-step process. The step one is the internal. Exactly. Step two, which is the product creating a new capability.

Nassia Skoulikariti: Exactly.

Nick Earle: On the product, what you build, what you stand for, what you sell, what your customers need, we're into this column now.

Nassia Skoulikariti: Exactly, we are on that, and this is where we're taking, you know, we augmenting the products that we currently have. If we… we stay with the theme, we are IoT leaders, so if we were to think of what an IoT connectivity, company would do, is… it would be to embed intelligence inside those connectivity products, offer adaptive pricing, for instance, or look at energy-efficient routing, or provide service SLAs powered by AI. This is just, like, a couple examples of what I mean by augmented the current product line, right? And so we're making the connectivity, the IoT connectivity, since I'm using that example, smarter, but also value-based, and that's the AI differentiator right there.

Nick Earle: Regular listeners to the pod, and there are a lot of regular listeners to the pod, may have had their memories tweaked then when you said that, because if I think back. Two podcasts jump out… two or three podcasts in the IoT Leaders series jump out at me. One of them was Amazon, which was, you know, a fantastic podcast. Where they're trying to tune, the connectivity to get, you know, five nines, 99.999999, connectivity per device by looking at additional data, as in the signal strength, the latency of the environment surrounding the device, so collect a lot more rich data, and then do AI… edge AI analysis, and have that drive the orchestration engine, the switch for AI. So it's sort of from the device in, not from the platform out.

Another one was Volvo, which was about 5 episodes ago, of course they're all available online, but 5 episodes ago. Where he said, a French guy at Volvo, who's responsible for IoT for Europe, said, oh, my job is to connect 500 million things across 140 factories, and you mentioned predictive maintenance, so that we can keep the software up-to-date on them all, so the production lines on those 140 factories, like Volvo trucks, for instance. The production lines never stop, and… there's an example of the intersection, the collision, as we talked about it on those two.

And then there was a recent one from Turkey, which I think goes into your third area, but it's more to do with, AI agents, autonomous agents, talking to each other and… data from patients, things in the hospital, and staff, doctors and nurses, so that's more like a number three. But number two is definitely where I see some of the bigger companies doing now, which is the predictive… it's not cost-saving as such, it's avoiding cost expense and avoiding process delay. By actually doing things like predictive maintenance, like at your electric car. I used to have a Tesla, it would… I'd get a software update overnight, and I'm wondering why I got it. Well, the reason why I got it was because they fixed problems I didn't know I was going to have in the future, because somebody else had had those problems, so they created a software patch, like your iPhone or whatever it does.

And so, the predictive maintenance means everybody is happier, and therefore you don't have to raise a ticket, your customer experience is better, and that's this middle column.

Nassia Skoulikariti: Data reduces downtime and all those good things that we've seen, just with the new age of AI, because we've been doing predictive maintenance for at least a decade, right? But now, we've made it more conversational, and we made it more agentic and more autonomous, where before, predictive maintenance will signal a potential problem in the future. Or in the near future, because let's just say a cog might be down to its last life cycle, and we will need to change it.

So, the new age of AI implementation will be that some agent — by agent, I mean AI agent, because not human agent, rare — will take the initiative. I call it initiative, but it's actually… we'll do it autonomously. We'll order the right new cog to replace it, schedule the right people to go in to do the actual work, if there's people who will be doing the engineering work to replace the cog, or robots, you know, those need scheduling as well. So, it becomes a lot more fluent. We're not there yet, Nick.

Nick Earle: I was going to ask you that I think a lot of people are...

Nassia Skoulikariti: Scared.

Nick Earle: Scared is a good one. Well, they're really scared… they're scared for their jobs. They're also scared… when we talked about governance and security and company image and whatever, the more the agents make decisions without those things in place. And the less the humans check the output, before, before it goes out, the more likely you could have a major, reputational brand… issue on your hands.

Nassia Skoulikariti: But we need the human in the middle, you know, checking at some… when we're looking at the workflow, right, whether it's internal or external, there is certain points, key points, intersections, that on the workflow, and by workflow, it's any given process. You know, even setting up this podcast and having this discussion, it was a particular workflow that, you know, was followed. And, you know, there's the human interceptions in those key points where the human chats.

So, I don't think, you know, we were scared about AI taking people's jobs, and yes, some jobs, the ones that are repetitive, will be replaced by automated systems. However, I think we are more into humans being augmented, freeing up time in order for us to actually do what matters. I mean, a CEO, a leader — do we rather spend time checking slides and creating slides, or thinking about the strategy and making the next quarter's numbers? What's more important? Things like that, you know, that AI can help us, and it's more constructive that way.

Nick Earle: Let me ask you a question which I didn't think I was going to ask you, but it's been prompted by what you just said. Again, another pod we had recently was, a guy who was in Czech Republic, we've gone very international lately, Czech Republic, that was doing load cells, which are these pieces of equipment that weigh things, typically things that are really, really heavy.

And he was talking about the… and how he's using AI to go from a company selling and servicing load cells to a company that is offering financial services. Where people, for instance, like, who have grain silos on farms, were telling them when the best is… how much grain they've got inside the silo, which is an IoT problem, and then when's the best time to sell that grain based on the futures market pricing of grain. It was very, very interesting, but one of the comments he made right at the end was what you just said.

He said, the problem we've got is that as AI, the agents get better and better and better, and they start talking to each other and training each other. He said, yes, he said, I know that a human has to intervene and have, but in order to have the knowledge to intervene, you kind of had to have had your apprenticeship of doing the basic work from, you know, your 20s until you're in your 40s, and you have enough experience to be able to say, no, that's the wrong decision, you don't understand. I'm going to augment the process by changing what the AI recommendation is.

What's your view on, whether or not… how do we get the skills into the people that are going to augment the AI models if the basic work that they're doing in the first few years of their career is going to be done by the models themselves? It's almost like a catch-22.

Nassia Skoulikariti: It's absolutely a catch-22, and it's affecting, I think, mostly the younger generations, the people that are coming into the workforce, now, or in the near future, rather than, you know, you and me, who've been in the industry, and we've had various experiences for a few decades, so we can take that knowledge and work with AI, and I think it's very important that we nurture the people that don't have the knowledge in order to learn the skills.

Because what I see happening, Nick, is that we — and I'm generalizing when I say we mean people in general, majority — is like, we query something on ChatGPT, Claude, Gemini, and we get an answer, and we're like, oh, that's good enough, but do we understand the context? Unless you had that experience, you don't really know what it means. So, in the meantime, if we don't have the knowledge, but we are trying to acquire the knowledge, is to check with multiple sources.

You know, each area is like, okay, AI's telling me XYZ, but why? So, it's that iterative communication with the different AIs, and taking the output, and putting it as an input to another model to see what you're getting, and then cross-referencing, going to reputable sources, working with mentors. Mentorship is going to become more and more important.

Nick Earle: That's interesting. Yeah, mentorship, skills, training, skills, retraining. And also different processes whereby, for something critical, maybe you need two people independently. Yes, before something, something, happens, because people make… in fact, that's what the airlines do, of course, is that there are certain things when they're flying a plane that are completely automated. Both the pilot and the co-pilot have to make independent decisions in order for something critical, so… Yeah. So as the technology rises up, they're still relying on two people, but it… it… skills training, which again, Apiro Data, of course, you're in the right place, is really, really part of this. Let's keep the structure going, and…

Nassia Skoulikariti: And if I may intersect there, is, like, I wanted to make sure that we're all aware about AI, and we all have the basic skills around AI, and make it approachable. That's why, you know, we created TechStage Hub. TechStage Hub is a community-based learning and support system, and opened it up for free for everyone, and especially, you know, for students and everything. I put a lot of information about, you know, AI-specific, training courses, all in there for free, because I cannot tell you how important it is for us to have at least a basic understanding, from somebody who's very young to somebody who's very old, because it will affect… it is affecting our lives already, so we need the basic awareness of where it's going, but most importantly, why? And what can we do about it?

Nick Earle: So let's underline that, because I was trying to guide us towards that. So… given that the skills, and we'll redraw the dots, given that AI is growing in capability exponentially, moving up the stack, skills, retraining, awareness, understanding, practical experience is going to be hugely, hugely important as the… water level, or the capability rises for AI, it's just going to be more important.

What that means is that there… and we see it in other areas as well, a lot of the AI companies are now giving the content away for free, and that's what you're doing. You've created with TechStage Hub. Which is linked to Apiro Data, it's a website, isn't it?

Nassia Skoulikariti: A community, well, think of it as our own social… media kind of, like, interface, right? Where people can talk to each other, they can watch, podcasts, they can watch webinars, they can take different classes, because information, quite frankly, if somebody has the time, the information, they can search YouTube, you know, they can find a lot of information, they can talk to ChatGPT or Claude or Gemini and learn all those things, but that's all information.

The difference is on the experience and operationalizing that intelligence. What do they do with it? Because what I found out is, like, you can give people a lot of things for free, you know, but they consume it, but unless they practice and, you know, they do something with it that makes sense for them. Doesn't work, it's just information.

Nick Earle: Okay, so that's a really important point is there'll be huge demand for information, but… and then what to do with that information, as opposed to just information overload. I want to get back to your structure.

Nassia Skoulikariti: Yes, we have one more element to finish.

Nick Earle: One more element, and I think that's, well, why don't you describe the element 3, the third leg of…

Nassia Skoulikariti: That's AI as a service, and actually, it kind of blends a lot of what we've been talking about. That's the external usability of AI for external systems, customer experience, and all that stuff that we've been talking. I mean, even how people training themselves could be put into that pillar.

And for an IoT company is where a company, and the examples that you mentioned, goes beyond connectivity, and then they help their customers, their enterprise customers, to interpret all the real-time data from the devices, you know, to do something with it and contextualize, which is what, you know, I think is the goldmine for connectivity companies, connectivity IoT companies, and where it's all going to go is creating that additional value. That's, you know, the advisory, that's the automation, that's, you know, providing the insights.

And where we stop selling data, because information is commoditized. Just as connectivity is commoditized, and we start selling outcomes. You know, that's the key thing, and that's where AI can help us move from something that is a commodity, that is a feature, to something that creates value, and it creates different revenue streams as well for the company, and for the individual, it creates different ways of understanding.

Nick Earle: In the years that I've been in tech, which is many, over 40, I've heard many times people talk about going from cost-based pricing to value-based pricing. And, and also saying that a few brave companies offering outcome-based pricing. I remember, again, going back to my time in Silicon Valley, there were some very innovative, companies that were doing supply chain management. I believe i2 was the name of one of these companies. I might have got that wrong, but I think I'm right. And i2 was saying, we can save you… 10% of your total supply chain costs with our… what were then, basically application software with workflows pre-built in.

And actually, we're so, confident that we can, that we'll only charge you X percent, or this much of your 10%. And then they advertise themselves on their website as, we've saved, you know, 400 companies 15 billion dollars, and that became their big… their big thing. But it was very hard for them to do that with using the technology that we're doing right now. It's interesting that it pops up again, and you talked about going to outcomes, and where companies then, this third leg.

And I'd like to come back to whether you have to do the first two first before you do the third, but park that for the moment. If they go to this third leg, what you're saying is you can actually start selling AI as a service, and what you're actually selling is an outcome. Right. Which has always been needed, but it's been very hard… there's huge demand for it.

Nassia Skoulikariti: How do we account for it?

Nick Earle: Absolutely.

Nassia Skoulikariti: Prices.

Nick Earle: But how do you price it? How do you limit your risk? How do you limit your exposure? You know, every customer environment is different. You seem to be saying that you do believe that outcome-based of value-added capabilities are going to be where we… the AI as a service will enable.

Nassia Skoulikariti: Yes, because if information is commoditized, is available, and intelligence is also available to everybody, when somebody can take their data and have AI analyze it for them. What is the value beyond the pipeline, you know, the data pipe that we provide as connectivity providers, as technology players, as telco companies, right? What is it? It's taking that data that we've been talking about and selling contextualized insights from that data. And that means that… because, first, we have to understand the data and what is needed ourselves in order to sell it as a service. So we package it for those companies, and we sell that intelligence as a service, that data intelligence from all the real-time data that we get from all the IoT devices. And why would somebody come to a company to take that instead of doing it themselves? It comes back to the same thing, you know? Do they have the time and the resources and the know-how?

Nick Earle: Well… well, maybe you'll be on the pod for the third time in the future, but for the moment, let me finish with two questions. One I've already mentioned briefly and I skipped over. So, we talked about the three phases, which is very useful as a structure. They're probably not completely discrete, I'm sure there's overlap, but is it your view and your recommendation of somebody who's at the cutting edge of this and makes their money from advising people on what to do. Is it your view that you have to start with the first one, which we called efficiency, and then go to the product one, and then go to the AI as a service? Is that broadly… although they may overlap as projects, is that broadly what.

Nassia Skoulikariti: They overlap. It's like, for instance, at least the first two, when we're looking at the operational side, the inside AI is like fixing our own home. It's like cleaning our own home, right? And making sure our house is in order. Then you've got the product side of the AI adding it to our products, one doesn't exclude the other. So, you can do both at the same time, or you can do…

Nick Earle: First one.

Nassia Skoulikariti: Exactly! So, what I'm seeing happening at the moment is companies adding, bolting on AI capabilities to their current products, more so before they look at AI internally to see how AI can help them operationally, and with their workflows, and so on and so forth. So, it's… and… but the third one, you know, before we go out and selling AI as a service, or intelligence as a service for that matter. We cannot do if we don't understand it, and if we haven't done it internally ourselves. How can you stake something and tell your customers you're going to achieve XYZ cost savings if you haven't done it internally?

Nick Earle: You've got to drink your own Kool-Aid.

Nassia Skoulikariti: Exactly.

Nick Earle: Okay, let's finish with, getting personal for the listeners. And let me explain what I mean by that. So, we have, quite an amazing, demographic of listeners on IoT Leaders after 4 years. And I was reviewing the data recently. About 55%, 60% of the people who listen are either departmental heads with budget and or… or CXOs, which is… an amazingly high percentage, and I think it's testament to the guests that we've had on and the approach we take to make IoT a business issue rather than a technical issue. But what I'm trying to say is there's a lot of… on the assumption that this podcast gets the same demographic as the previous ones, there's a lot of very senior people all over the… listeners in more than 90 countries, who listen to this.

So, what I mean by getting personal is that… let's look forward. Today, most of them have got IoT in their business title. They are in some way related to IoT, which is why, of course, they listen to the podcast. Going forward. How will their roles change? What do you think the future will be of a manager responsible for IoT? What does all this mean for them?

Nassia Skoulikariti: So, IoT and AI, we talk about the intersection, we talk about those two technologies colliding and working very well with one another. Can one exist without the other? Of course it can, and it has for many years. But together, pun intended, that's where the magic happens.

So, what will happen to the roles, right? Right now, we have very few people who understand the technology holistically, IoT, AI, and everything in between. But on the leader side of things, we need to understand the context and what it means, what it goes into the future, the strategy side of the things. But more so, like, is it for IoT or is it for AI? I would put it as, like, the middle bit, intelligence. Where is intelligence going? Because in my view, we're moving away as an IoT industry from, like we said, you know, charging for data usage. A lot of IoT companies, they're charging for data usage, or device costs, and so on and so forth, to what we talked about earlier, data value, right? That's the intelligence, the data value. So, and that's where connectivity becomes intelligence.

So, what is that all going to evolve to? We're going to want to be intelligent agents. For a lack of a better word, you know, it sounds such a cliche.

Nick Earle: We're going to… those of us who are in IoT are going to become super intelligent leaders.

Nassia Skoulikariti: Exactly!

Nick Earle: A bunch of minions, a bunch of minions, oompa Lumpers around us, which are actually the AI agents for the work. But we'll actually become the super-agent and selling the value. And your point about the data, you're absolutely right, and it's been a theme of these podcasts, you know, data prices, IoT data have been declining by 23% a year for 6 straight years. What that means. That's a ball that's bouncing down the stairs, as I've said many times. You cannot catch it. If you're growing anything less than 23% a year, you're going backwards.

That data, as the amount of data grows, that data, and the granularity of the data, you know, from small devices and sensors increases, you can never catch that ball, so you cannot build a strategy based on that. But what you can do is build a strategy based on the value of the richness of the data to turn it into insight. Correct. Information and insight for your internal processes of compliance. So that is a major, transformation, which will take many years. And take many years to play out, but… but I do believe it's consistent with the theme we opened the podcast with, which is the… I think we called it the collision. Collision, the merger, or the collision's a better word, of, AI and IoT being the next big phase, and what it will look like, and the need for people to reskill and retrain.

Linked, because of the hundred times more data than there is in things, not yet connected, but things in time. Versus, what we've done so far with, pilfering the content. Let me, let me, use it that way. Nassia, we could talk for hours, as always, but I'm conscious of the time. Thank you so much for your, insight.

Nassia Skoulikariti: Pleasure.

Nick Earle: As a reminder, APIRO, A-P-I-R-O, New Word Data.

Nassia Skoulikariti: Data.

Nick Earle: Yep, all one word or two words?

Nassia Skoulikariti: One word.

Nick Earle: Oh, there we go.

Nassia Skoulikariti: Apirodata. So it's, apiroData.com or .io.

Nick Earle: Yep.

Nassia Skoulikariti: And you can find us there. TechStage Hub as well, so you can find me on LinkedIn, I live in there. I propagate information and knowledge and, try to be as disruptive as possible, saying what people don't want to hear, so…

Nick Earle: Oh, that makes two of us! And if anybody goes to Mobile World Congress, it's almost impossible to go to MWC Barcelona, at least, without bumping into you. I know you're very active, doing a lot of things there. So, Nassia, as always, thank you very much. Thank you for being a guest on the pod, and thanks for explaining what it's like to be at the leader of this collision as IoT and AI start to blend.

Nassia Skoulikariti: Nick, always my pleasure.

Nick Earle: Okay, talk to you again soon. Thanks.

And before I sign off, I'd like to share some exciting news about our plans for FY26. Now, for the past four years, we've been known as IoT leaders but starting in January next year, we're evolving the podcast to IoT and AI Leaders. Why are we doing this? Because this change reflects the growing opportunity for AI to be enriched by the massive amounts of IoT data that's out there, which by some estimates is anywhere from 50 to a hundred times greater than the existing text, audio, and video data that's been used to train today's large language models. So, be sure to tune in and find out what IoT can become when AI moves to the centre. Thank you for listening throughout 2025 and I wish you all a very happy New Year.

Outro: You’ve been listening to IoT Leaders, featuring top digitization leadership on the frontlines of IoT. We hope today’s episode has sparked new ideas and inspired your IoT and digital transformation plans. Thanks for listening. Until next time!