Goodbye 2025, Welcome 2026

With the end of the year just around the corner, it is time to look back and take a brief look ahead to 2026.

AI is everywhere

As expected, AI is now in all products, and its benefit is questionable at best. The downside of all those futile attempts is an explosion in licensing costs. Even when you do not have any upside of these new capabilities (if they exist at all), the downside gets you. Are we really 3x more productive with the AI assistant to justify the 3x explosion of licensing costs for Docker Desktop or Octopus Deploy? I do not think so.

It shows once again that truly good AI use cases are hard to find. It is much easier to integrate a chatbot to meet managers' demands for AI.

Finding helpful AI use cases

In this turbulent year at work, we also faced a gigantic demand for AI. It was only topped by the number of self-claimed AI experts that wanted to help us to reach our full potential.

But the problem for us is not missing knowledge, or tools, or agents. The problem is understanding the business well enough to see what takes time and where AI can help. We got a lot of room to explore AI solutions, what allowed us to cut through the marketing promises at a high pace. Most of our experiments resulted in not pursuing an idea, but the few we followed through had a massive impact.

As it turns out, LLMs are great to check scanned documents. They help us where OCR does not work and check signatures in a reproduceable manner. What once was a tiresome and error-prone manual task is now automated.

Where LLMs fail terribly is with complex rule sets. Sometimes the thinking part considered all relevant rules, but then the LLM went on and did it differently. For use-cases like these, we are back at building everything by ourselves. That takes time, but it gives us a correct result we all can understand.

If you are interested in the building blocks of AI, you can find the code samples over at PythonFriday.dev.

Challenges ahead

With a lot of luck, the AI hype gets gently pushed into a new phase in which we finally can talk about solving problems and not just about using AI for the sake of using AI. That would give us all much more productive discussions, and we could build solutions that have a positive impact.

Much more likely is that the AI hype needs a lot more oxygen in the form of cash to keep burning. In that case more and more experts will pop up and demand massive investments by the government to not fall behind. Details are usually sparse, especially as the requested sums increase.

2026 has the potential to be even more of a mess than 2025 was. I expect plenty of opportunities to learn new things. Not necessarily because we want to, but more because we must.

Nevertheless, let us take one challenge after another and do not miss out of the nice things that happen. They exist too and deserve our attention that much more.

I wish you a happy 2026 - may the odds be ever in your favour…