Insights
Considered opinion on product strategy, AI in production, food technology, trading infrastructure, and building software that holds up in the real world.
You don't have an AI strategy. You have an AI feature.
Why companies keep shipping AI capabilities nobody uses — and what the ones who get it right do differently. The distinction between bolting AI on and building it in, and why the data model is almost always the real product.
The factory floor nobody told the software about
Most food safety software was designed to satisfy auditors, not to help the people working on the production line. The gap between compliance documentation and factory reality is where the actual risk lives.
Integrations are not features
Every vendor has a slide showing which systems they connect to. That slide is where honesty goes to die. On the ownership vacuum, silent failure modes, and what it actually takes to keep data flowing reliably between systems.
The SaaS layer nobody asked for
Best-of-breed procurement sounds rational until you count the systems someone quietly built to make all those best-of-breed products talk to each other. On the integration layer that grows by accident and becomes critical infrastructure.
Why migrations take three years
Legacy system replacements are scoped at six months and completed in three years, if they are completed at all. The reasons are almost never technical — they are organisational, structural, and almost entirely predictable in advance.