21 min read 1/11/2026

Scenario play - Everybody builds software

subtitle: Who builds agents?

You can literally play with the future. Some call it scenarios. I like the term too. It implies to me that this is just one of the possible futures.

One central tenet of Risto Siilasmaa’s book - the paranoid optimist - was that future can be anticipated. That anticipation simply entails making a model of the possible future. This can be a mental model, calculations or just imagined scenes in the life. The scenario can be of customers, the market or our own internal operations.

But why does Risto propose we look into the future? So that we can notice the world unfold. It will either follow parts of what we imagined and anticipated. Or then the world just does something completely different. Nothing fully works out as anticipated.

The thing that happens is that you are primed to observe the right things. What could look like big chaos actually contains some shapes of what you have foreseen. You have developed a concept of what the key stuff in future means.

I take this as the same as what Louis Pasteur posited: “Dans les champs de l’observation le hasard ne favorise que les esprits préparés.” (In the field of observation, chance favours only the prepared mind.)

I have very much come to appreciate this! Let’s play a scenario: everybody builds agents.

Code generating agents digitalise each and every business action

All software is now up for grabs. If you can connect to something with software, then you can start to build for it. Phone: start making apps. Take on the latest Claude/ChatGPT/Gemini and go for it. Getting going is a matter of prompting.

Getting going is a matter of daring and awareness. These both perceptions will shift in months not years. The tools get even simpler. More people will know of the possibility. I know the latest Claude Code and ChatGPT Codex are quite nerdy. The early adopters are already moving.

CISO makes Ai agents. CTO makes Ai agents. Executive assistants make Ai agents. These agents handle anything that is repetitive and worth spending 30 minutes to automate. These automation opportunities don’t arrive at software people only. The world of agents is not at all for software engineers only.

I play the game that Claude Code and ChatGPT Codex will be the real agent workbenches where innovation will both democratise and scale. The Claude Code is an agent by itself - mostly known for software creation capability. But try it out for agent generation. The Claude Code cracks that nut just as easy.

Then you see one implication, the role of the central software factory changes radically. Especially IT will see a big shift from being the place where automations are created.

Take a guess of the other shifts. Even if you get it wrong, you’ve prepared.

From LLM-driven to software

The modern code-generating agents are quite wild. They on-the-fly create software scripts and even software systems to handle things. They don’t just plainly dump everything to be processed into an LLM.

These agents gradually shift workflows that everyone automates into software.

See how the digitalisation really picks up manyfold. It is not just the nerds creating software. Everyone in the business creates software.

Makes total sense. What can be made to work with LLMs is formal enough. Most of that process can be turned into software.

Play to the thought game. What do you see?

Customers speak out agents into existence

Stretch the thought just a bit more. If you have B2B customers, they are certainly on the same trajectory.

Consumers…. They can speak things into existence. Just prompt. The person has an agent. It likely will be valuable. Some fail. Some succeed.

Consumers will create their own software.

Missing enablers? Your arena of innovation

Epilogue after checking in with Claude - Thinking only is not enough

I used Ai to publish the article on my own site. I was not certain I had all the angles yet in my mind. My hunch turned out right. I was missing acting on the future.

I have had the fortune of acting on the Ai-future over the last months. I have acted on the trends. I have tried my best at affecting the future as it has unfolded. I have learned more than half what this is all about by my mistakes and attempts.

Acting creates the visceral proof whether or not something is real. You can read up on the internet all you like. You can follow the academia. You might develop anticipation. You don't develop data in your own domain and your own world with that.

Acting based on the scenario is the test whether or not what you predicted is real.

Predict, act accordingly, observe, understand. Keep teaching your brain.

Further reading: