The AI Operating System Is Coming - and most businesses aren't ready

In December 2025, we sat down with my co-founder Agis and our guest Sotiris Karagiannis, Chief Innovation Officer at Space Hellas Group, AI pioneer and early AI entrepreneur who was among the very first to study artificial intelligence at the University of Edinburgh, to talk about the strategic direction of AI.

Not the hype. Not the shiny tools. The big picture.

What followed was a wide-ranging conversation about where AI is genuinely heading, what's holding businesses back, and where the real opportunities lie. Here's what we unpacked.

AI Is Becoming an Operating System, Not Just a Tool

We've gone from chatbots to agentic systems that browse the web, analyse documents, trigger workflows, and even transact on our behalf. The direction of travel is clear: AI is quietly becoming an operating system sitting between people, data and business processes.

For business leaders, especially in the mid-market, this raises uncomfortable questions. Where is the technology going in the next couple of years? What will change in how we build products, serve customers and organise work? And what happens to the companies still on the sidelines while their competitors are deploying these systems?

The Internet Analogy, But Accelerated

Sotiris drew a sharp parallel between the early internet and today's AI landscape.

"It's amazing how I see the analogy between the early internet days and today's AI. There was a hype, but everybody was trying to understand what the internet would provide. Now we know - it took us a lot of years to understand the full value of the internet. AI is something like this, but with a huge, huge difference in terms of acceleration."

The lesson? Just as businesses that moved early on the internet captured lasting advantage, the same is happening now with AI. The difference is that the window is narrower. The acceleration is real.

Voice Will Be the Next Big Shift

One of the most striking takeaways was Sotiris's conviction about voice AI. Not as a gimmick, but as the primary interface through which people, and then businesses, will interact with AI.

"I think that voice will go faster, and in one or two years, we all will be talking to some kind of our own AI assistant, not only for our work, but also for doing things in our everyday lives."

The implications go deeper than convenience. As Sotiris explained, these fast voice interactions will need to be supported by "huge systems behind that - data systems, backends, cloud services, even local edge compute, on-premise computing for companies that do not want to share proprietary information."

Agis reinforced the point from personal experience: "Even today, the voice models are not that good. But when I'm driving, I can call the voice agent and have a discussion. I often do my research on various topics, even projects. I will discuss with the model the project."

The shift from typing to talking isn't trivial. It changes how we interact with machines entirely, and it lowers the barrier for people who might otherwise resist AI adoption.

Why Businesses Aren't Getting ROI Yet

We've all seen the reports: companies are enthusiastic about AI, but the results aren't materialising. Sotiris was blunt about why.

"Companies haven't understood yet how they can do that, and there's not enough people to actually consult them. So they're making very small steps. Usually they just want to jump into the bandwagon of AI. They're doing it in the wrong way, and they get disappointed."

He pointed to a noisy market where "everybody comes out there and says, 'Oh, I'm an AI expert, I'm an AI consultant,'" without the depth of experience to guide businesses through complex implementation. His advice was clear: find a reliable person to consult you, then take small steps. Define your early wins. Find your lowest-hanging fruit. Build from there.

Agis added a critical insight about data readiness, a reality check many businesses need to hear:

"I have been in this space for almost 10 years, and I always remember: clients and companies think they have good data. Before looking at the data, I can tell you - you don't have data. And the reason is very simple. You have not collected the data having in mind what you want to do with it."

The 1:10 Cost Ratio That Should Keep Leaders Awake

One of the most concrete points came from Sotiris on the economics of automation:

"Based on what I'm seeing already, the ratio of the costs - the cost of the previous process to the cost of the new one - is about one to 10. So if you don't automate fast enough, you're going to operate at 10 times more cost than your competition."

That's not a theoretical risk. It's already happening in areas with simple, repetitive work that can be automated, from report generation to media assets to routine processes. The businesses that move first don't just save money. They structurally change their cost base.

Two Types of Opportunity: Productivity and Innovation

We identified two broad categories of AI opportunity for businesses.

Productivity: Doing what you already do, cheaper, faster and better. Sotiris urged companies to "examine which processes they can make faster and cheaper" as a starting point.

Innovation: Doing things you couldn't do before. This is where it gets exciting. From AI-powered research at scale to agents that actively discover new customers, the possibilities are genuinely new.

"Imagine that you can leave your agents to go out and find your new customers. You could not do that with the internet. You had a good page, they had to come to you with SEO. But now agents can discover your new customers out there. It's amazing." - Sotiris

The Future of Search, and Commerce, Is AI-Native

Sotiris flagged something that should be on every business leader's radar: the future of search itself is changing.

"We will be searching through AI. A lot of things, including e-commerce, will change. Also the way you present your company will change, because the information is going to be gathered from the AI. So you will want to optimise the way AI processes this information. This alone is like a huge industry."

This is a fundamental shift. If customers find you through AI agents rather than Google, your entire digital presence strategy needs rethinking.

Are We Overreacting on Regulation?

The conversation turned to data governance and regulation, always a sensitive topic. Sotiris posed a provocative question:

"Are we overreacting? People going out there throwing their personal data in every kind of social media. And then they come back and say, 'I don't want you to get this data.' The European Union is, I would say, decelerating things with regulation."

Agis drew a sharp analogy with EU agricultural regulation: "We have very rigid regulations for our own produce, but we import the produce of other countries which has no regulation whatsoever. If we want to do the same with technology, just import innovation that someone else built, it's a very good strategy to feel better, but we will just import technology that someone else has, with the data."

Their shared view: regulation has a role, but fear-mongering is counterproductive. The real risk isn't in using data. It's in standing still while competitors don't.

The Public Sector and Smart Cities: A Massive Untapped Opportunity

Agis shared a personal passion that ties directly to AI's potential for public good:

"The public sector holds a lot of information about everything - taxes, revenues, companies, health records. I'm excited about fixing cities, making them safer with AI and data. There's data out there, open data and closed data, that can be collected easily without any personal information. Anonymous information that can make the city safer."

His vision of safer cities, starting with women's safety, using publicly available data and AI is the kind of use case that demonstrates what becomes possible when data meets purpose.

Small and Global: The New Shape of Business?

I asked whether we're heading towards a future where small, mighty teams with fully integrated AI agents can operate globally from anywhere. The answer was nuanced.

Sotiris was clear: "On the services side, there's an amazing possibility. You can produce knowledge or services that can work with 25 customers, delivering your services without you being awake, or even being alive."

And then he offered a reality check: "If you're producing a product and you don't have the logistics, even if you have beautiful smart agents selling all over the place, you have to deliver." Digital services? Absolutely. Physical products? AI won't solve your supply chain.

The Bottom Line

The message from this conversation is clear. AI is moving from a tool you use to an operating system you work within. Voice interfaces will accelerate adoption. The businesses that start small, get their data strategy right, and find reliable guidance will pull ahead, and the cost gap between movers and waiters is already widening.

As Sotiris put it: "Time is of the essence here." Businesses don't need to do everything at once. But they do need to start.

Sotiris Karayannis is Chief Innovation Officer at Space Hellas Group, an AI pioneer and early AI entrepreneur. He studied artificial intelligence at the University of Edinburgh.

This article is based on a two-part live conversation recorded in December 2025 between Sotiris Karagiannis, Celia Rizothanasi (co-founder and AI Strategist, iA: Intelligent Acceleration) and Agis Georgiou (co-founder and AI & Data Science Expert, iA: Intelligent Acceleration).