The Archaeology
Field Note 04 · The Schema
They wanted AI insights. The AI was ready. The data wasn’t.
A London-based trading company had built a dashboard used by FX dealers. Leadership wanted the next step — automated observations, AI-powered pattern recognition, the kind of intelligence that helps a dealer spot an opportunity three seconds before someone else does.
It should have been straightforward. The data existed. The dashboard worked. They just needed someone to help with the prompts.
Within the first week, we found the problem wasn’t the AI. It was everything underneath it.
The dashboard had been built in C# over several years by developers who had since moved on. There was no documentation. No saved formulas. No record of how any calculation was derived. The numbers on the screen were correct — probably — but nobody could explain how they got there.
If we’d simply written prompts against this data, the AI would have generated confident insights based on calculations no one could verify. In FX trading, that’s not just a mistake. It’s a compliance risk.
So we went into the SQL databases and did something deeply unglamorous: we reverse-engineered every calculation. Documented every formula. Mapped every data structure. Created, for the first time, a clear record of what the company’s own numbers actually meant.
Three months. Three production-ready AI prompts. One documented system.
The most valuable thing we delivered wasn’t the AI. It was the archaeology — the patient, painstaking work of understanding what was already there before building anything new on top of it.
Every organisation has code nobody can read, formulas nobody documented, institutional knowledge that walked out the door with the last developer. AI doesn’t fix that. It builds on top of it. And what you build on a foundation you don’t understand is a building you can’t trust.
Before your next AI initiative, try this:
pick any number on your main dashboard and ask the team to explain how it’s calculated. Not what it shows — how it gets there. Who built the formula. What’s included and excluded. If nobody can answer that cleanly, you’ve just found your first project.
The Schema is a publication about taste, curation, and building beautifully in an age of noise. Written by Bhavneet Saini.