Paid to copy-paste
August 22, 2026
The first elephant in the data room is not a technology problem. It is a competent person, on a Tuesday, reconciling four systems in a spreadsheet because nobody ever gave them a better way.
The Monday report
Somebody in your organization rebuilds a report by hand every Monday. They export from one system, paste into a second, fix the date formats, reconcile the totals against a third, and send a spreadsheet to eleven people. It takes three hours. It has taken three hours every Monday for four years.
Nobody treats this as a problem. Nobody experiences it as one. The report arrives. The eleven people are satisfied. The person doing it is good at it. Genuinely good: fast, careful, the only one who knows which of the four systems to trust when they disagree. That competence is the disguise. The work looks like expertise because someone skilled is doing it.
This is the elephant we call Robot: repetitive work and the preparation tax, where people do copy-paste instead of thinking. It appears on no org chart. It has no budget line. Its cost is paid in the analysis that never happened.
The evidence
Nobody acts like the ceiling is this high
57%
of U.S. work hours could technically be automated with technologies that already exist
McKinsey Global Institute, Agents, robots, and us, 2025.
Does 57% mean half of all jobs are going away?
No. And the difference matters more than the number does. McKinsey Global Institute measured hours, not people, and technical potential, not economic likelihood. The finding is that with technology already in existence, the activities filling 57% of U.S. working hours could in principle be performed by a machine. That is a ceiling, not a forecast. Read it as a headcount prediction and you get it wrong in both directions at once.
Ceilings are still useful. A very high one tells you something specific: when a repetitive task survives untouched year after year, the reason is almost never that it cannot be automated. Nobody looked. Or somebody looked and picked the wrong task. Or the person doing it is too competent to complain.
The copy-paste is a choice. An unexamined one, made by default and renewed every Monday.
What does the Robot actually cost?
Start with the hours, because the hours are the easy part. Three hours every Monday is about four working weeks a year, spent by one person, on one report. That arithmetic is uncomfortable enough that most organizations stop there and call it a productivity problem. Stopping there is what keeps the elephant fed.
The expensive part is second-order. The person best placed to ask why two systems disagree spends their Monday making them agree instead. A process attached to a name, not a role, becomes a single point of failure that takes its vacation with them. A number assembled by hand is a number nobody can audit. Decisions get made slowly, or confidently and wrongly.
There is a quieter cost still. Ask someone to do mechanical work for four years and they will become very good at mechanical work. Goodhart’s law says that when a measure becomes a target, it stops being a good measure. Here the measure is that the report goes out on time. So the skill you were actually paying for atrophies in plain sight, while every performance review confirms the measure was met.
How do you know the Robot is in your room?
It hides in plain sight, so look for the artifacts rather than the activity. Any one of these is normal. Three of them in the same process is an elephant.
- A report that lives as a spreadsheet on somebody's desktop, rebuilt rather than refreshed.
- A process attached to a named person, which quietly stalls when they take a week off.
- A step whose only justification is that one system cannot talk to another.
- Totals reconciled by hand, plus an unwritten convention about which source wins when they differ.
- A deadline set by how long the preparation takes, rather than by when the decision is needed.
Why does automation so often land on the wrong workflow?
Because the wrong workflow is the one you can see. It has a vendor, a demo and a slide. Generative AI has made this sharper rather than softer: it is now trivial to automate something impressive and irrelevant. The result demos beautifully to people who never have to live with it on a Monday.
The workflows that actually cost you are boring, undocumented, and defended, often sincerely, by the person who owns them, because that ownership is where their standing in the organization comes from. Nobody volunteers their own three hours as the pilot candidate.
From the workflow you can see. To the workflow you can price. Choosing well is not a technology decision at all. It is an honesty exercise: what does this workflow cost us today, in hours, in delay, and in the risk that one person leaves? Answer that for the top handful of candidates and the ranking usually surprises everybody, including the person who was sure they already knew.
How do you get the Robot out of the room?
The same three moves we bring to any elephant, pointed at this one.
Map
A few weeks, not a few quarters. Follow the hours rather than the org chart: where does preparation actually happen, who does it, and what is waiting on it? The output is a short ranked list. The ranking is the valuable part, not the inventory.
Prove
Take one workflow, the smallest one whose disappearance somebody would notice, and automate it end to end, run by your people, with us alongside them. The pilot is the training. If it does not survive a real Monday, it did not work. Finding that out in three weeks is the point.
Scale
Extend whatever earned it, and leave your own people able to run it, govern it and change it. An automation that nobody in-house can modify is not a solved problem. It is a new elephant in better clothes.
Find the workflow that costs you most
The Elephant Safari names your herd in ten questions and ranks them by what they cost you. Or start from the business problem instead: needing to do more with less. Get the Robot out of the room. Get the thinking back.
Every figure in this article traces to a primary source. See it in Knowledge