AI CASES
Every case follows the same shape: the problem, the approach, the result, and what went wrong along the way. That last block is not decoration. A project where nothing ever broke tells you nothing about whether someone actually built it.
The site you are reading is the case itself. It started as a Lovable template, design and code now come out of Claude Code, the logo was generated, and work only reaches production after review. The admin I use to update it is part of the case too.
An organisation does not adopt AI by buying tooling but by changing the way it works. I build the layer around that: working sessions, a knowledge base that stays current, and a recurring update showing what has become possible.
The admin around a gym is largely repetition: assembling programming, keeping track of what has already been done, and recording it in several places. That now runs as one system that composes a full week and submits it for approval.
The admin around A/B testing was entirely manual: title, hypothesis, description, and then tracking the status yourself. It became a form with a workflow underneath, and in two years the number of tests went from 19 to 163. This was my own work that I automated, which explains why adoption was the easy part.
At an organisation where I introduced AI, I built an automated newsletter about AI for my colleagues. Because almost nobody read it, I added a feedback module, and when that went unused too I stopped the newsletter.
An AI model remembers nothing between two conversations, and I work across domains that must not mix. I built a workplace in which Claude knows those domains, picks up work where the previous session stopped, and follows rules I do not have to explain again every time.