Moments of Truth.

HB Antwerp was founded for a single purpose: to disrupt diamond manufacturing. Not to improve it at the edges, but to take a supply chain that had run on handshakes and paper for more than a century and make every step of it verifiable, from the mine to the finished stone. Nobody had done that end to end before. Microsoft, in parallel, was looking for a company digitising an unconventional industry, and there are few less conventional than this one, so they became the engineering partner. I was one of the first people assigned to the project from HB’s side, and I was there from the launch.

The mission sounded simple. Microsoft’s engineers knew nothing about diamond manufacturing, so someone had to turn everything the business did into requirements they could build against. That was us: about seven people across product and engineering on HB’s side, working with five to seven Microsoft engineers per module across finance, production and sales. I had come from an operations analyst role, which meant I had already spent time inside most of those workflows. That’s why I was on the project.

The hardest thing we hit was price. Our stones came from a mine in Botswana on consignment, ours for sixty days against an initial value. That value then changed when our own experts assessed the goods. It changed again once the stone was polished and its real quality was revealed, and the gap between what an expert predicted and what the stone turned out to be was wide and routine. Then it sold, usually for a different number again. Publicly, this is the part of HB’s model that made the news: the company pays for stones on their polished value rather than the rough, which is what sends more money back to Botswana. Internally it meant one physical object carried four different prices, and cashflow, insurance and the royalty calculation each depended on a different one.

Dynamics could not do this. When an asset enters Dynamics it carries one static value, and we needed an asset whose value was a history rather than a number. We tried to assemble it from what Dynamics already offered and it simply wasn’t flexible enough, so we ended up building custom on Azure alongside it rather than bending the ERP into a shape it was never designed for.

Most of my two years went on the gap between those two ways of thinking. I was translating rules that had never been written down, every department’s exceptions and every “we just know that”, for engineers who quite reasonably expected a supply chain to behave like a supply chain. The logic clashed constantly. I ran the syncs, took the escalations when reality didn’t match the spec, and owned the testing that checked what got built actually worked for the people on the floor.

The project ran for two years and I moved straight from it into Granular. That wasn’t a coincidence. Two years of specifying a system for someone else to build is a very effective way of finding out which problems it was never going to solve.