What the Tasks Were Hiding

Every month, for six years, the same person reconciled the customer numbers from sales against the customer numbers from finance. It took most of three days.

The two figures never matched. Sales counted a customer as active from the day a contract was signed, and finance counted one from the day the first invoice was paid. She knew this, adjusted for it, and moved on.

The reconciliation is now automated and runs in eleven minutes. It matches perfectly every month, because nobody told it about the difference, and the report that reaches the board no longer carries any trace that there is one.

Where this sits

The first two essays in this series argued that organisations do not solve for structure, and that the person standing inside one cannot see it. Together they describe how structural drift happens: an arrangement nobody chose, assembled one reasonable decision at a time.

The next three watched the instruments an organisation owns reach for that arrangement and stop short.

A values programme works on intention. An engagement survey stops at the feeling. The one to one gets closest of the three, and cannot carry what it finds out of the room.

This essay is the turn. It asks what happens to drift when the tasks are taken away.

Three failures, one cause

The three instruments look as though they fail for three separate reasons. They fail for one.

Management orthodoxy runs on a single sequence: measure the thing, then manage it. A cost is tracked and then reduced, a risk is scored and then mitigated, a target is set and then met.

Structural drift cannot enter that sequence, because there is nothing to measure. An arrangement is not a quantity. It has no unit, no baseline and no owner who would report it.

So the orthodoxy offers two substitutes, and only two. It can state an intention and hope the arrangement follows, which is what a values programme does. Or it can measure something next to the arrangement and treat the reading as though it were the arrangement, which is what an engagement survey does.

The one to one is the exception that proves the point. It is the only instrument that actually hears about the arrangement, and it produces nothing the measuring system can take in.

None of this is negligence. An organisation that has not acted on drift has usually done exactly what its management system allows, which is to act on what it can see.

Structure has not been neglected. It has not been seen.

The work was the shock absorber

For as long as people did the work, that did not matter as much as it should have.

Every defect in an arrangement has to be dealt with somewhere, and almost all of them were dealt with inside the tasks. Two definitions of a customer were reconciled by the person doing the reconciliation. A report that needed three systems was assembled by the person who knew which one was wrong on Mondays.

A handover was made to work by someone who remembered what the process had left out. A deadline set before the work was scoped was met by someone who absorbed the difference in their evenings.

None of that appeared in any description of the job. It was not in the task, it was in the person doing the task, and the organisation experienced it as the task simply getting done.

That is how drift stays invisible for years. Each increment is paid for in small amounts of human adjustment, made by people who treat the adjustment as part of the work, and the payments are never itemised.

The organisation looks as though it runs on its processes. For the most part it runs on the corrections people are quietly making to them.

What automation removes

Automation is sold, reasonably, as the removal of tasks. What it also removes is the adjustment that was travelling inside them.

A system executes the process as written. It does not know that the written process is wrong, because the knowledge that it was wrong lived in a person and was never written down.

So one of two things happens. Either the automated task fails visibly, and the defect that had been absorbed for years surfaces at once. Or it succeeds, faithfully and at speed, and carries the defect forward into every report, decision and system that depends on it.

The first outcome is inconvenient and useful. The second is the one that matters, because it looks exactly like success.

An organisation's structure is what is left when the tasks are taken away.

For most of the history of management that was a thought experiment. It is becoming an operating condition, one task at a time.

Load bearing, and still unseen

An earlier essay suggested that structure was about to become close to the only thing an organisation is made of. This is what that means in practice.

While people did the work, the structure only had to be good enough for them to compensate for. Once they stop doing it, the structure has to be right, because nothing is left inside the process to correct it.

For the first time the arrangement is carrying the full load itself. It is doing so at the same moment it remains the one part of the organisation nobody can see.

That is the turn in this series. Drift was a slow cost, paid in effort, spread across people who never itemised it. As the tasks go, it stops being a cost carried in the background and becomes a risk carried in the output.

The gap between how much structure matters and how much of it is visible was always there. It is widening, and it widens fastest wherever the most tasks are being handed over.

If the people who were quietly holding the arrangement together are the ones being automated out of it, who is left who knows what they were holding?

What cannot be seen, and what can

It would be easy to conclude from this that the answer is a better map. It is not, and an earlier essay has already said why.

Nobody can see the whole structure of an organisation of any size. The accurate accounts that exist were nearly all written afterwards, from outside, and paid for by damage. That has not changed.

But the reconciliation that opened this essay contains a smaller observation. Nobody in that organisation could have drawn its structure. The person doing the reconciliation could say, precisely and without hesitation, which two definitions did not agree and what she did about it every month.

Not the web. One strand of it, named exactly, by the person whose hands were on it.

The second half of this series starts from that distinction.

Opinion: The Automation Is Not the Risk. It Is the Audit Nobody Commissioned.

Our position is that artificial intelligence does not create structural risk in organisations. It removes the human work that was concealing the risk they already had.

That is a less comfortable claim than the familiar one, and a more useful one. The familiar version treats automation as a new hazard to be governed. This version says the hazard is old, it was drift all along, and the tasks were where it was hidden.

It follows that the organisations best placed are not the ones with the most advanced systems. They are the ones that know what their people were correcting before the systems arrived.

Very few do. Knowledge of that kind was never collected, because it was never needed: the correction was made, the work was done, and nobody asked what it had cost to make it work.

That is also why the tools the demolition half examined cannot help here. A values programme, a survey score and a well run one to one would each have missed the reconciliation. It was not a belief, it was not a feeling, and nobody thought to raise it, because it was simply part of the job.

The question we would leave open is practical, and it belongs in the next business case for automating anything. Before the task goes, ask the person doing it what they correct that is written down nowhere.

If the answer is nothing, the task is ready. If the answer takes more than a minute, the organisation has just found out something about its own structure that no instrument it owns would have told it.

The contents of this article are for informational purposes only and do not constitute professional, legal, or financial advice.