The Museum of Old Work, Issue 01: New AI. Old Work.

A LETTER FROM THE EDITOR

Why a museum, not a report

I have spent the past few years in offices that had just bought the future. New licences, faster models, launch-day excitement. Then I looked at what happens on Monday morning.

The digital document gets printed, signed with a pen, scanned, and returns to email as a photo. The monthly report is rebuilt from zero although the data already exists. The follow-up stops after the second message. The licence sits activated and unused, like a precious object under glass.

Nobody is lazy in this story. Serious people carry processes that no one ever designed. They settled layer upon layer, until improvisation became the standard. So familiar that we no longer see them.

That is why this magazine is a museum, not a report. A report gets filed. An exhibit forces you to look. We took old work out of the office, placed it under gallery light, and marked one thing in every frame with orange: the way out.

This is not a museum of outdated technology. It is about processes that are still running. You may recognise them. One of them may be yours.

The visit starts on the next page. The exit, as in any good museum, is through the last room.

Răzvan

ACT I

THE EXHIBITS

Old work is invisible. That is its first line of defence. Copy-paste does not appear on the org chart. The print-sign-scan loop has no owner. The monthly rebuilt report has no budget line. They all live in the space between systems, paid for in small daily hours that never reach any report. To see them, you have to take them out of context. Put them on a plinth, under glass, with gallery light and an inventory label. Only then does the familiar object become what it really is: an artifact. A piece from an era we claim has ended. Eight exhibits open the collection. Every one of them still runs today, in companies that already have AI.

EXHIBIT 01 · ACT I

CTRL+C. CTRL+V. UNTIL THE WORK WORE OUT.

Three worn keys say more than any process audit.

Nobody ever decided that data should travel through fingertips. The systems do not talk to each other, so a human became the cable. The wear on CTRL, C and V is not carelessness. It is physical evidence of a process that was never designed, only copied from one day to the next until it became the standard.

EXHIBIT 02 · ACT I

THE DOCUMENT WAS ALREADY DIGITAL.

We printed it, signed it, scanned it. So it could become digital again.

Nobody designed this loop. It survives from a time when a signature meant paper, and paper meant proof. Today the document leaves as a file, passes through a printer, receives a pen stroke, and returns to email as a photograph. Four steps, zero added value. We call it a process because it repeats, not because it makes sense. Its printer has already earned the display case we put it in.

EXHIBIT 03 · ACT I

THE APPROVAL IS WAITING FOR APPROVAL.

Every step demands one more signature. The risk does not shrink. It hides.

Nobody built the barricade of stamps, and everybody defends it. Each extra approval promises safety and delivers delay. When everything requires a signature, nobody actually decides. Responsibility just circulates. One stamp has left the system. The rest are waiting to be approved.

When everyone signs, no one decides.

EXHIBIT 04 · ACT I

THE MONTHLY REPORT, RESTORED FROM ZERO.

Same sources. Same tables. Same reconstruction, every single month.

The data exists. The sources have not changed. And yet, every month, someone reassembles the report fragment by fragment, with the patience of a manuscript restorer. We treat a recurring file like an artifact dug out of ruins. The care is admirable. The fact that it is necessary is not.

EXHIBIT 05 · ACT I

THE ANSWER EXISTS. IT IS BURIED IN THE INBOX.

We do not answer messages. We excavate context, attachments and versions.

The inbox became an archive without ever being designed as one. The same questions receive the same answers, written from scratch, because nobody turned them into a working practice. Every repetitive email is one more drawer. The decision is in there, underneath. It just has to be dug out, daily.

EXHIBIT 06 · ACT I

THE FIRST DRAFT. AGAIN.

Old work forgets what it did yesterday.

The most expensive blank page is the one that ignores everything already paid for: the versions, the decisions, the comments, the corrections. Every document started from zero throws away the work of the one before it. The piles around the plinth are not failures. They are invoices. We paid every one of them.

EXHIBIT 07 · ACT I

THE FILE EXISTS. SOMEWHERE.

The documents are not missing. The path from question to answer is.

When information has no shared structure, searching becomes work in itself. Every open drawer is a hypothesis. The person searching is not disorganised. They work inside a maze that everyone built and nobody manages. There is exactly one direct route through.

EXHIBIT 08 · ACT I

THE TRAINING ENDED. THE PRACTICE NEVER STARTED.

Without scheduled time and real tasks, the old way of working returns first.

The kit is complete: notebook, headset, materials, laptop. Everything clean, arranged, untouched. Knowledge does not turn itself into routine. It needs official time, one concrete task, and a person who owns the result. Otherwise the enthusiasm settles neatly under glass. And stays there.

ACT II

THE PEOPLE STILL CARRYING IT

An old process does not carry itself. Someone hauls it, every day, usually without ever appearing in a report. The lawyer comparing contracts line by line. The sales director watching follow-ups die on day three. The assistant keeping one decision synchronised across three places, from memory. The controller hunting for the one exception that matters inside a wall of identical cells. The problem is not a lack of intelligence. Quite the opposite. These systems keep working precisely because very good people manually compensate for what was never built. Their daily heroism is invisible, and that is exactly why nothing changes. Eight portraits. Different roles, the same weight carried by hand.

EXHIBIT 09 · ACT II

THE SPEED IS NEW. THE LIABILITY IS OLD.

The differences could be found in seconds. They are still found by finger.

The verification stays human, and it should. Her signature carries the liability, not the system’s. But between judgement and searching lies a gap of hours. A system can surface exactly the lines that changed. She decides what they mean. Today she does both. Only one of them uses her title.

EXHIBIT 10 · ACT II

THIS IS WHERE THE PIPELINE LEAKS.

The first message goes out. The second, maybe. The third waits for later.

The pipeline does not die from a lack of leads. It dies in the small places where the follow-up never leaves, because the person meant to send it had nine other things to do. The cards do not fall out of bad faith. They fall through the crack between intention and calendar.

EXHIBIT 11 · ACT II

AI PURCHASED. THE WORK STAYED IN THE BOX.

Access to AI is not the same thing as using it.

The licence was activated. The process was not. The object under the dome is expensive, admired and perfectly sealed, while the paper stacks around it keep growing. The company bought access to speed and kept the instructions for moving slowly. The orange seal is waiting to be pulled. It has been waiting for months.

A licence activates in a day. The work does not.

EXHIBIT 12 · ACT II

THE SHARED SYSTEM IS, IN FACT, A PERSON.

The same decision lives in the inbox, the calendar and a notebook.

The company has no single place where the truth lives, so the truth is held together with string, memory and care. She knows which version is current, who confirmed, and what changed since. This connective work appears in no job description. It only disappears when the person leaves.

EXHIBIT 13 · ACT II

ATTENDANCE GETS COUNTED. CHANGE STAYS INVISIBLE.

The stack of certificates grows. The folder on the desk stays identical.

It is easy to report how many people sat in the room. It is hard to show which task is done differently since Monday. The tall stack confirms attendance. The thick, untouched folder confirms the rest. Between the two sits the difference between training delivered and practice installed. Only one of them changes what Tuesday looks like.

EXHIBIT 14 · ACT II

THOUSANDS OF CELLS. ONE EXCEPTION.

Finance does not need more data. It needs the decision.

The wall of tables demands equal attention for every row, although only one row matters. The exception that requires judgement gets lost inside the volume that requires none. Manual reconciliation is not rigour. It is the cost of not knowing where to look.

EXHIBIT 15 · ACT II

THE FACTORY LIVES IN 2026. THE BACK OFFICE DOES NOT.

The line produces to the millimetre. The office approves at the speed of paper.

On the floor, tolerances are measured with a caliper. Behind the glass, the same parts wait for a stack of documents tied with rope. The precision of production stops at the administration’s door. The company has two speeds, and without meaning to, it uses the slower one.

EXHIBIT 16 · ACT II

THE GOODS ARE MOVING. THE APPROVAL BLOCKS THE DOOR.

The system knows where every box is. Nobody knows who signs next.

The belt runs, the boxes flow, everything is traceable to the second. Then the flow meets a column of binders that does not move at all. Modern operations stop politely in front of a habit. The habit has never once shown up for an audit.

ACT III

THE MOMENT THE PROMISE BREAKS

A LAWYER, A SALES DIRECTOR AND A GM

Walk into the same AI training.

This is where the punchline should land. It does not. Three different roles, the same folders, the same closed laptop, the same conversation about AI held above papers that have not changed. The role does not predict who will be using AI three months from now. The chosen task, the allocated time and the shared way of working say far more than the title on the business card.

The promise does not break when the contract is signed. It breaks three weeks later, quietly, with no announcement email. The launch went well. The training took place. The access works. Then the calendar springs back to its old shape. The old tasks return first, because they already know the way. The room stays empty, the materials stay beautiful, the licence stays active. Nobody decided to abandon it. It is just that nobody decided the opposite. Three scenes about the distance between having AI and working with it.

EXHIBIT 17 · ACT III

A LAWYER, A SALES DIRECTOR AND A GM

Different roles. The same folders, the same closed laptop, the same old work.

Three roles meet the same gap between having AI and working with it. The role does not predict what changes. The chosen task, allocated time and shared practice do.

EXHIBIT 18 · ACT III

THE ROLLOUT STARTED PERFECTLY.

Then the calendar went back to normal.

There was a launch, a training, and access for everyone. The room looks immaculate. That is exactly the problem: nothing is worn, nothing is moved, nothing is used. The old way of working already knew the road back. It found it without a map.

EXHIBIT 19 · ACT III

THIS IS NOT AN AUDIT. IT IS A MIRROR.

The tools you bought show up in reports. The real work shows up here.

Management sees the licences, the dashboards, the training budget. The honest mirror shows something else: the corridor of files the team walks through every day to deliver, the copy-paste, the stamps, the rebuilt reports. We are not looking for someone to blame. We are making visible the place where time, context and decisions quietly leak. There is an orange line on the floor. It leads out. It only needs to be followed.

The mirror does not show who you are. It shows how you work.

EXHIBIT 20 · ACT III

START WITH THE TASK EVERYBODY HATES.

Do not automate job titles. Pick one clear, repetitive task.

The first candidate for AI is not anybody’s job. It is the task everyone postpones: repetitive, verifiable, without glory. The one you hand over with relief, not with fear. That is where the first trust is earned and the first hours come back. The rest of adoption is built on that small proof.

ACT IV

THE WAY OUT OF THE MUSEUM

The way out of the museum is not another tool. There are plenty of tools, some already paid for and sealed under glass, as you have seen. The way out is a practice. One real task, chosen with care. A result that can be verified by a person who puts their name on it. A way of working the team repeats until it becomes a reflex, exactly the way the old reflex became one. It is not spectacular. It does not announce itself with fireworks. It shows up weeks later, in small things: a report that no longer gets rebuilt, a follow-up that leaves on its own, a search that takes one minute. Four pieces on what the beginning looks like.

EXHIBIT 21 · ACT IV

FIVE QUESTIONS BEFORE THE INVOICE.

Define the task before you buy the tool.

What goes in. What must come out. Who verifies the result. Which data is allowed to travel. Who decides the output can be used. Five simple questions, asked before any purchase, make the difference between a working system and one more beautiful object under glass.

EXHIBIT 22 · ACT IV

ONE CLEAR TASK. ONE VERIFIABLE RESULT.

An AI worker is not a digital colleague. It is a configured system.

Disordered volume enters on the left. A structured, sealed package exits on the right, straight into a human hand. That is everything an AI worker needs to be: a known input, a controllable output, a person who validates. If you cannot say what it processed and what it left aside, you do not have a system. You have a black box.

EXHIBIT 23 · ACT IV

SUBMIT THE MOST ABSURD MANUAL PROCESS IN YOUR COMPANY.

We will exhibit it anonymously. Then we will rebuild it.

The museum grows by donation. Real processes, submitted anonymously by the people who endure them daily. We display them as collection pieces, we show their mechanics, and then we rebuild them, so they no longer depend on somebody’s daily heroism. The box is open.

EXHIBIT 24 · ACT IV

THE LICENCE SITS IN INVENTORY. ADOPTION LEAVES MARKS.

Wear is the only evidence that does not lie.

The objects in the case are immaculate because nobody has touched them. The orange disc is scratched, worn at the edges, and sits next to a finished piece of work. The inventory shows what you bought. The wear marks show what changed.

THE WAY OUT IS MONDAY MORNING

THE LICENCE SITS IN INVENTORY. ADOPTION LEAVES MARKS.

Wear is the only evidence that does not lie.

The Museum of Old Work is not a place to laugh at. Every exhibit in here worked, once. The print-sign-scan loop was, in its day, a guarantee. The chain of approvals was prudence. The manually rebuilt report was rigour. They worked well enough to become reflexes. Then the world changed, and the reflexes stayed. AI does not change the work because you bought it. You have seen in these display cases what happens to access without practice: it stays new. Perfectly preserved. Untouched. The work changes when one real task receives a new way of working, a verifiable result, a person who validates it, and a practice the team repeats until it becomes the new reflex. One task, then another. No fireworks, just wear marks. At the unlearning school, teams unlearn the old reflex on real tasks, not on slides. The change happens inside everyday work. If you recognised one of your own exhibits in these pages, do not leave it in the museum. Write to me at hello@unlearning.ro with the most absurd manual process you carry. We will exhibit it anonymously. Or better, we will rebuild it together. The way out of the museum starts on Monday morning.