The Cost of Manual Work

Manual work is usually costed as time. Someone estimates how long a step takes, multiplies by how often it happens, and treats the product as the price of doing it by hand. That number is real, and it is also the smallest part of the bill. The expensive parts of manual work are the ones that do not appear on anyone’s timesheet.

The cost is mostly waiting

A step that takes ten minutes of effort rarely takes ten minutes of calendar. It sits in an inbox until someone opens the inbox. It waits for the person who knows the answer to come back from a site visit. It queues behind whatever else that person was already holding. The effort is minutes; the elapsed time is days, and every downstream step inherits the delay.

This matters most where the manual step sits on a path that releases money or unblocks other people. A slow approval does not cost the approver’s ten minutes. It costs whatever was waiting on the approval, for as long as it waited.

Two people, two answers

A human process is not one process. It is as many processes as there are people running it, each with their own reading of the rule, their own idea of what counts as close enough, and their own set of exceptions they have quietly been making for years. None of that is bad faith. It is what happens when a rule lives in people’s heads instead of somewhere it can be read.

The cost of that variance is not the variance itself. It is that nobody can tell which version was applied to any given case, so any question about a past decision turns into an investigation rather than a lookup.

Rework is the compounding part

Manual steps produce errors, and errors found late cost more than errors found early, because everything built on top of the error has to be unwound with it. A wrong figure caught at entry is a correction. The same wrong figure caught after it has been summarised, reported and relied upon is a retraction, an explanation, and a loss of confidence in every other figure that came through the same door.

That last part is the one that compounds. Once a set of records has been wrong once, the reasonable response from anyone reviewing them is to check all of them — which is a new, larger piece of manual work created by the original one.

The record that was never kept

The deepest cost is the one that only appears when someone asks. Manual work produces an outcome but usually not an account of itself: who decided, when, on what basis, against which version of the rule, and what evidence they were looking at. In the moment nobody misses that account, because everyone involved remembers. Months later, when a lender, an auditor or a counterparty asks how a number was arrived at, the memory is the only source and it is not good enough.

Reconstructing a record after the fact is the most expensive work in this whole essay, and it is entirely avoidable. The point of automating a step is not that a machine is faster than a person. It is that a machine leaves a trail as a side effect of doing the work, and a person does not.

Which steps are worth it

Not every manual step should be automated. The ones that should are the ones where being slow blocks money, where being inconsistent creates dispute, or where being unable to explain the decision later is a real exposure. Steps that are slow and harmless can stay manual for a long time without hurting anyone.

Related reading: why lender-financed construction runs on evidence, not paperwork. Or see how we work.