Let’s talk about the person who only ever sees the work when it goes wrong.
Picture a woman at a freight company. Forty trucks, maybe sixty. Three years ago her desk was a pile of paper, a monitor, and a phone that did not stop.
Then the company bought the software. Invoices come in, the machine reads them, matches them, posts them. Ninety-six out of a hundred go through with no human involved at all.
The other four land on her screen.
That’s the job now. Four in a hundred. The smudged scan. The vendor who billed in dollars. The duplicate. The one where the delivery note says eight pallets and the invoice says nine.
She is the most experienced person in that department. And she has not seen a clean invoice in two years.
Everybody calls this a win, and on the spreadsheet it is. Headcount is down. Processing time is down. The vendor’s dashboard shows a ninety-six percent straight-through rate, which is the number that got somebody promoted.
But look at what actually happened to her.
The machine took the easy cases. The boring ones. The same invoice from the same supplier, over and over, until you know in your gut that the petrol levy looks wrong this month. That repetition was not waste. It was how she learned the shape of normal.
What’s left is the tail. The strange stuff. And here is the trap: you cannot spot strange if strange is all you ever see.
There’s an old word for this job in operations. Exception handling. Every automated system has one, because every automated system has cases it cannot close. The design is always the same. Let the machine run the volume. Push the leftovers to a person.
And in the slide deck, that person gets a nice name. Human oversight.
She is not oversight. Oversight means you can see the whole thing. She sees the four percent that broke. That’s not oversight. That’s a complaint desk with a salary.
Two things rot from there.
The first is her judgment. Judgment is built on the ordinary. You learn what a good invoice looks like by seeing ten thousand good invoices. Take those away and the skill does not stay. It fades. Six months in, she is slower. A year in, she is guessing. And the strange cases are exactly where the money leaks, which means the company put its least informed person at the door where the money walks out.
The second thing that rots is her will.
Think about what her day is now. Nothing finishes clean. There is no pile that empties. Every single thing that reaches her is somebody else’s problem, and when she fixes it, it disappears and another broken one takes its place. No wins. No rhythm. Just other people’s mess, all day.
People do not quit because work is hard. They quit because the work stops making sense.
And here is the part that should make you angry.
When a duplicate payment goes out, nobody calls the model. Nobody asks the vendor what its accuracy was on that batch. They walk to her desk. Because her name is on the approval, and her name is the only name that can be walked to.
She signs for a system she only ever sees when it fails.
The dashboard stays beautiful because she is the one holding the mop. Every mistake she catches makes the machine look better. Every mistake she misses becomes her mistake. That is not a partnership. That is a person standing under a leak with a bucket, being told the roof is working.
So what has to survive here?
This. Responsibility has to come with a view. If you answer for a system, you must be allowed to see it working, not just watch it break. Not a monthly report. Not a summary somebody else wrote. The actual ordinary flow, the clean cases, the boring ones, in front of your eyes, often enough that your gut still knows the difference.
That is not a training perk. It is the condition that makes the signature mean anything.
If you run a business and you are automating a process, here is what to do with it.
Keep the accountable person’s hands in the clean flow. A sample every day, not a slide every quarter. Rotate the people on the exception queue back through the normal work, so the skill does not dry up. Test them on both. Can she still tell a good case from a bad one, cold, without the machine’s hint? And if the answer is no, do not promote her to oversight. Fix the job.
Because the alternative is worse than waste. It is a signature that means nothing, held by a person who cannot see, blamed for an outcome they had no way to judge.
The machine should carry the volume. That is what it is good at, and there is no shame in letting it. The human should carry the judgment. That is what we are for.
But judgment needs to see more than the wreckage.
Give her back the boring ones. Not because she needs the work. Because the boring ones are where she learned what right looks like.