The Repetition Is Not the Job

I spent years inside enterprise operations watching work get automated, and doing a fair amount of the automating. The pattern was reliable. Someone found a task that ate hours: a report assembled by hand every week, a request routed through four inboxes, a spreadsheet reconciled against another spreadsheet. The task got automated, the hours came back, and everyone agreed it should have happened sooner. Then, every so often, the same instinct pointed at a different kind of task, and the result was quietly a disaster. The interesting question is what made those tasks different.

The difference was never technical. The tasks that automated beautifully were motion: work whose steps could be written down completely, whose success was checkable, and whose failure was cheap and visible. The tasks that went wrong looked identical from a distance but had something folded inside them: a judgment call, a relationship, a moment where a person was quietly deciding something on behalf of the organization. Automating the first kind removes drudgery. Automating the second kind removes an owner, and nobody notices until the day the decision goes wrong and no one can say whose it was.

The repetition is not the job. The decisions inside it are. Every recurring task is a wrapper of motion around a small number of judgments. The invoice run is motion around "is this bill correct and should we pay it." The scheduling is motion around "who is entitled to this hour." The customer reply is motion around "what are we promising." Automation succeeds when it strips away the motion and leaves those judgments visible, faster to reach, easier to make well. It fails, expensively, when it buries a judgment inside the machinery, where it keeps being made, by default, with nobody's name on it.

Sort by accountability, not by frequency. Firms inventory their work by how often it happens and how annoying it is, which is exactly backwards. The useful sort is by answerability: for each step, ask who would have to explain it if it went wrong, and to whom. If the honest answer is nobody, you have found motion. Automate it without sentiment. If the answer is a name, because money moved, a person was told something, or a commitment was made, you have found a decision, and the standard changes: a tool may prepare the decision, gather what it needs, even recommend, but a person approves it and owns the outcome. That line is the difference between a firm that uses tools and a firm that has quietly delegated away pieces of itself.

A tool cannot be answerable, so it cannot be responsible. Responsibility is not a workload to be redistributed; it is the capacity to explain, to be corrected, and to bear the consequence. A system can log what it did. It cannot answer for it. When an organization says the system decided, responsibility has not been removed; it has been hidden, and hidden responsibility does not stay hidden. It resurfaces at the worst moment, attached to whoever is standing closest. The practical rule follows: every automated flow carries a named owner, the way every department carries a manager. If nobody's name fits, the process has not been automated. It has been abandoned in motion.

Good automation knows when to stop. The routine cases are where machines earn their keep, and the unusual cases are where organizations earn their reputations. Design for the boundary between them: the moment a flow meets something it was not built to judge, an amount out of range, a customer who is angry rather than mistaken, a document that almost matches, it should halt and present the situation to its owner, plainly, rather than improvise. A system that is confident about the routine and humble at the edges builds trust with every quiet week. A system that improvises at the edges spends that trust in one bad afternoon. The measure of mature automation was never how much it handles. It is how cleanly it hands back.

None of this is an argument against the tools. I have spent too many hours doing work a machine should have been doing to romanticize drudgery. It is an argument about which layer you are allowed to delegate. Repetition is a cost, and costs should fall. Judgment is the firm: the accumulated record of what you will promise, pay, tolerate, and stand behind. Hand the first to the machines with gratitude. Keep the second where it has always belonged, with people who can be asked why. The repetition was never the job. The job is the part with your name on it.