Back office

Work with written rules, verifiable output and little dependence on context that is not documented. Form processing, validating data against sources, classifying and routing requests, preparing documentation and routine reconciliations all fit this description, and they represent most of the administrative volume in any mid-sized organisation.

What travels badly is anything depending on judgement about unprecedented exceptions, on relationships with specific people, or on knowledge that exists only in the head of whoever has done the work for years. That knowledge can be written down, and writing it down is a phase of the project rather than an assumption of it.

The middle zone is where decisions are won or lost: work that looks routine and whose exceptions consume most of the time. We measure that proportion in discovery, before proposing any model, because sizing to the normal case is the mistake that makes cheap operations expensive.

The most reliable gain is in cycle time rather than cost per hour. Internal administrative processes frequently run in batches, when someone has time, and a dedicated team handles them continuously. The difference between two days and two hours changes what the business can promise its end customer.

The cost gain depends on the model and is measurable before deciding. With Lisbon at one hundred, Porto sits at ninety-two, Poland at seventy-eight, Brazil at around forty-five and Uzbekistan at thirty-five, on fully loaded cost rather than on salary.

And there is the gain that does not appear in the budget: qualified people stop spending part of their day on administrative work. That time has a real cost and is usually larger than the direct saving, even though no accounting line records it.

With discovery, lasting one to two weeks, observing the work as it is rather than as it is documented. We measure real volumes by case type and by hour, identify the exceptions, and produce a design both sides recognise as accurate before any model is proposed.

Then comes knowledge capture, the phase that decides the outcome and the one most frequently compressed. We need protected time from the people doing the work today, scheduled and recognised as work, because what is not written down leaves through the door with whoever knew it.

Launch happens as a pilot with ten to twenty per cent of volume over six to twelve weeks, with the current team still standing. The criterion to advance is a number agreed beforehand, met for two consecutive weeks. One good week is noise; two are a signal.