12/08/2026
Claude for everyone, mandatory.
In the first quarter of 2026, we wrapped up our “testing” of various AI assistants. From IDE plugins to Coding assistants and Frontier models like OpenAI and similar. And made a decision: Claude, for everyone, mandatory. Not as a bonus tool, but as a strategic direction.
Then everyone documented three tasks where AI helped, and three where it didn't.
But the bigger shift is somewhere else. In what's now possible.
The question used to be: "Can we afford a thorough analysis and a couple of ex*****on scenarios?" Now it's: "Which bold ideas do we want to pressure-test before we commit?"
More depth. More iterations. More testing that used to »we don’t have time/budget«. New types of solutions that simply weren't on the table before.
We also have tasks where AI complicated the solution. Where it lost context or spiraled on a problem. Where it took more time than we'd have spent ourselves. We document those cases just as seriously as the wins. Because those are exactly what teach us when to use AI and when not to.
Just as important is the safe and responsible use of AI. Something we've put real effort into. Client and end-user data stays local. It doesn't go into AI models.
We are continuing with more AI being the center of our work. Because it’s the only way to thrive in this fast-paced world.
Thanks to all the enthusiasts on the team who are really pulling this forward. Without you, none of this would be moving. 🙏