My view on artificial intelligence in 2026

Three years in, and it is a tradition. Here are my annual AI reflections, for the third time!

How 2025 went

Last year I ended with some thoughts on what could happen to e-commerce when more and more people use AI agents to assess products instead of visiting the websites themselves. Once again, things moved faster than I managed to predict. By May, more and more people had started to take an interest, and the terms AEO and GEO emerged. There is little doubt now that a growing share of website traffic goes through AI agents. My impression, though, is that the solutions for getting visible through AI agents still have a lot of room to improve.

OpenAI in particular seems to have ambitions of positioning itself within AI visibility, and of offering products that can promote visibility through ChatGPT. But how can that really be commercialised without hurting the trust people have in the language model?

2025 was the year vibe coding took off. I have “vibe coded” some fairly complicated applications myself. I get the most out of it in frontend development and simple API development, while others may get more out of other areas. Either way, the quality of the generated code improved steadily through the year, and the problem of getting lots of code that does not work is a thing of the past. With the right set of instructions, the generated code I use is on par with (or better than) what I would have written myself.

There has been no shortage of promises of huge productivity gains, or of warnings about what happens if you are left standing on the platform. We have read reports that most AI projects fail. Perhaps more surprising was the MIT report from this summer, saying that we get dumber from using AI all the time (no, not really, but we do need to keep the brain muscle in shape; the article is in Norwegian)!

So where do we go from here, now that it is 2026?

I have gained a more balanced view of AI development. I do not feel that things are moving as fast as before. It seems we are approaching a plateau, where things can settle down a bit. AI tools have become part of everyday life for most people, and what we need now is time to integrate the tools into processes and tasks where they actually make a difference.

Today AI tools are easy to get hold of, and most businesses can use AI to help answer enquiries and emails, or to generate content. It can be fully integrated into the company's systems, like Microsoft Copilot, or as simple as the employees having personal ChatGPT accounts.

If you use a general LLM in your work, you can get a productivity gain, but so can your competitors. How do you stand out when everyone has the same generic tools?

It is like having a super-skilled colleague who works part-time until lunch, but works for your biggest competitor afterwards.

So everything you create with an LLM is very easy to copy, while those who manage to build their own machine learning models will have something unique that is hard and expensive to copy. That is why I think machine learning will become more important going forward, and that generic AI will not be enough.

That brings us back to studying what data we have access to, and how we can use it to achieve deep learning and understanding in our own core area.

Through 2025 we have seen the models get better, but we have also seen the AI providers struggle to become profitable. The most thorough models are no longer available on the free tiers or the lower paid tiers. They typically require premium access, where ChatGPT Pro costs $200 a month. That says something about how expensive this technology is. At the same time the results are at a very high level, and often right on par with human expert level.

Let's start talking about the big questions

The very fact that AI has become part of everyday life for everyone makes me think about some big questions, for example: how does this affect the way we look at information?

Where information, skills and knowledge used to be limited resources, each of us now has access to something that resembles a team of experts in every conceivable field. Those who have made a living from offering information are under threat, and we have already seen several consulting firms cut staff. I have generated several comprehensive reports myself, particularly for some studies I have done on e-commerce and visibility, with impressive results!

I believe the way we look at information is changing fundamentally, and I think this could matter a great deal for us going forward. There are both threats and opportunities here. Many will have to learn to work differently, and competition in skilled professions and activities will increase.

This is tending towards the philosophical, but the knowledge revolution we are seeing now has the potential to challenge much of our understanding of the world, both at the individual level and at the level of society.

For 2026 I hope the “AI hype” will calm down, and that more people will start to take an interest in the bigger questions.

Also read last year's article and the first one, from the start of 2024.