Different Industries. The Same Obsession: Solve the Problem.

01Four buildings, one first hour

A semiconductor fab and an assurance practice do not feel like neighbours. In the first, a tool that stops costs more in an hour than most software costs in a year, and the people who work there think in tolerances. In the second, nothing exists until it can be traced, and a conclusion nobody can evidence is worth precisely nothing. A certification body lives on independence. A government programme lives on accountability to people who never asked to be its users.

None of that transfers. The vocabulary does not, the regulation does not, and an instinct learned in one of those rooms will quietly mislead you in the next. What does transfer is how the first hour is spent — finding out what is actually going wrong, and refusing to propose anything until we know.

02The temptation is always to lead with the technology

It is easy to arrive with an answer. We work in a field with a great many of them, and every one arrives with a demonstration attached. The demonstration is the problem: it is built on the use case that shows best, which is almost never the use case that hurts most.

So we ask the unglamorous questions first. Who does this today, and how long does it take them? What happens when it goes wrong — who finds out, and how quickly? Which part of this would you keep even if the technology worked perfectly? And which part would you delete tomorrow if you were allowed to?

The last question is the one that changes the room. People are rarely asked which part of their own process they would delete.

03What actually crosses between industries

After fourteen years of this, the transferable part turns out to be short — and it is never the technology.

• 01The problem is almost never the one described in the first meeting.
• 02The data is always messier than the system diagram admits.
• 03The people doing the work already know where it breaks; nobody has asked them in a form they could answer.
• 04Whatever gets built will be judged on its worst day, not on its demo.
• 05Somebody has to run this after we leave, and they were not in the room.

04Which is not an argument for treating industries as interchangeable

The opposite, in fact. A model that flags a defect on a wafer and a model that flags an anomaly in a ledger may share a shape. What they must never share is a threshold, an escalation path or a definition of an acceptable miss. In one of those worlds a false negative is scrap; in the other it is a finding that should have been raised and was not.

That is why the people we send are people who have stood on the floor — engineers who have worked on equipment, analysts who have sat through an audit cycle, delivery leads who have run a system that other people depend on at three in the morning. Domain fluency is not a slide in a proposal. It is the difference between a system that is trusted and one that is worked around.

The domain changes every week. The obsession does not.