System integration consulting

It broke on their side, and it is our fault.

The external system answered the wrong thing, answered late, or did not answer at all. Your customer sees your error. Working out what happened takes a day and ends in a thread where both sides are right.

A seam is where assumptions live. What each side actually promises gets written down, and separately, what it actually does. All the future work lies between those two lists.

Failure gets designed, not success. The question is not how this works when all is well, but what happens when the neighbour is unavailable, answers slowly, answers twice, or answers untruthfully. For a supplier, a payment rail and a state reporting platform those are three different behaviours.

Whose side gave way is visible from the system, not from the correspondence. This is not about blame; it is about the investigation taking minutes.

A wrong assumption on a seam is usually not a bug. Where a regulator is nearby, it is an incident with your name on it.

A seam that survives the neighbour's behaviour, and a record that shows what actually happened. Plus a boundary of responsibility named in advance — before it is needed in an argument.

Trading equipment under state traceability for precious metals. The machine, the payment, the state reporting and the warehouse all have to reconcile, and every operation has to be producible on demand. The courier meanwhile drives through country where there may be no signal — the app has to survive that rather than count on a connection.

Another case: a state programme subsidising mortgage rates. The expensive part was not selecting applicants but the eight-plus systems that had to be held together behind one funnel.

Responsibility for the other side staying up. We cannot stop your neighbour falling over; we can stop their fall looking like yours, and stop it taking all of you down.

Name the seam that gets investigated by hand most often. And one number: how many times a month.