Thinking
Writing on delivery, AI, and the craft of consulting. Occasional and worth reading.
5 min read
Most AI proofs-of-concept fail not because the model is wrong, but because the delivery infrastructure around it is missing. Here is the framework we use to fix that.
4 min read
Two decades of regulated delivery in financial services teaches you things about execution that most AI teams haven't learned yet. The lessons are not about technology.
6 min read
Non-determinism does not mean untestable. It means the test strategy needs to change. Here is what that actually looks like in practice.
Agentic AI is arriving in financial services, insurance, and professional services. The governance frameworks are not ready. Here is what to design for.
3 min read
Most model cards are written for AI researchers. Legal and compliance teams need something different. Here's what to include.
The decision is almost always made on the wrong criteria. Here's what to look for instead — including questions to ask in the first meeting.
Everyone talks about foundation models and use cases. Nobody talks about the integration layer in between — the part where most AI projects actually break.
Testing a system whose outputs are non-deterministic is genuinely different from testing traditional software. The difference is not primarily technical — it's philosophical.
AI makes bad requirements faster. The discipline of business analysis is more important now than it was five years ago — not less.
Transformation programmes fail in predictable ways. Most of the signals are visible from day one — if you know what to look for.
Most organisations practice Agile-the-ceremony without Agile-the-outcome. The framework isn't the problem. Here's what is.
The gap between a working prototype and a production AI system is almost always a delivery problem, not a model problem. Here's what that gap actually looks like.
I'd worked inside large consulting firms long enough to see the pattern clearly. Here's what I built instead.