There is a pattern I keep running into when I talk to business owners in this region. They know AI is real. They have seen the demos, read the articles, maybe even sat through a presentation from a vendor. And then they went back to work and waited.

Not from ignorance. From a specific logic: let someone else take the risk, prove the thing works, and then follow when the path is clear.

I understand that logic. I grew up in London, spent years as a business analyst and systems engineer inside large organisations where careful, consensus-based decision-making was the operating culture. Then a decade in Sweden watching Scandinavian businesses apply the same instinct at a regional scale. Watch-then-move is genuinely baked in. It is not irrational. It works for a lot of things.

It does not work for this.

The reason is compound effects. When you wait for someone else to prove AI works, you are not just saving yourself the risk of being wrong. You are also giving up all the learning that comes from running real systems. The person who deployed their first AI workflow in 2024 is now on their third iteration. They have trained staff. They have data. They have made mistakes and fixed them. They have a workflow that took six hours and now takes forty minutes, and they have had eighteen months to reinvest that time.

You cannot catch that up by moving faster later. You can close the gap, but you cannot erase it.

I chose to build IPRESTANDA in Frøya, in Trøndelag, because I think this region has something worth preserving: real businesses, real industries, people who actually know how to operate things. Aquaculture, construction, professional services, logistics. These are not glamour industries, but they are complex, and they are full of operational problems that AI is genuinely good at solving.

The risk tolerance gap here is not because Norwegian SMEs are behind. It is because the available path has looked too abstract, too global, too designed for someone else's context. Most AI tools are built for English-speaking enterprise buyers. The case studies are from Silicon Valley or from the City of London, environments I know well and that bear little resemblance to a 40-person company in Hitra. There is no obvious on-ramp for that business, and the people building the tools have never needed to find one.

That is the gap I am trying to close. Not with theory, but with systems that actually run.

MIT Technology Review's current AI assessment is worth reading if you track this space. The clearest signal in it is that AI has moved from pilots to operations. The companies that ran experiments in 2024 are running infrastructure in 2026. The pricing model has shifted to outcomes, not seats. That does not happen until something is reliable enough to stake commercial terms on.

The window is not closed. But it is narrowing. The businesses that move in the next twelve months will compound into a real operational advantage. The ones that wait for the market to mature further will find the gap harder and harder to close.

My first step when I work with a new client is always the same: forget about AI for a minute and tell me where you are losing time. What is the thing that happens every week that should not require a human? We start there. Small, defined, measurable. Get one system running well, then build the next one.

That is all early moving is. Not a transformation programme. One workflow, working properly.

Murphy Alex writes about building operational AI systems from Trøndelag at murphyalex.net. IPRESTANDA is at iprestanda.com.