Blog
Notes on shipping AI where it has to hold up
AI governance, the EU AI Act, DORA, and the product decisions underneath them. Written from inside the work, not from the compliance briefing.
Articles
Start here
What an AI use-case register actually looks like
Everyone is talking about the EU AI Act. Almost nobody is showing what compliance looks like on a Tuesday afternoon, at the intake meeting, when someone wants to ship a feature. What I learned building the register — including why the most useful question isn’t about risk at all.
DORA doesn’t care that it’s AI
If you ship AI into financial services, there is a regulation that reaches your product before the EU AI Act does. It has applied since January 2025, it has no AI chapter — and that is exactly why it catches teams out.
“There’s a human in the loop” is usually a fig leaf
The most reassuring sentence in AI product development, and most of the time it means nothing. A reviewer approving a hundred decisions an hour isn’t oversight. The four conditions that make it real — all of them product decisions.
We were asked to cover two thousand templates
The number was in the deck and it sounded like ambition. The usage data said something else. Why coverage is easy to report and value isn’t — and how to argue for aimed scope with evidence.