The Soup Got a Promotion

The copy machine at the unemployment office sounded like it was digesting gravel.

A woman in a gray sweater stood beside it with a stack of forms and the patient dead-eyed look of someone who had already been punished by three other machines that morning. The page came out with half the words smeared into fog. The clerk looked at it, shrugged, and said, “good enough.”

Then the bad copy went into a file. Someone read the bad copy and made a bad decision. Everybody spent two days fixing what one lazy shrug had created.

This, apparently, is the future with better lighting.

The companies went all in on artificial intelligence because the men upstairs had begun to sweat through their shirts. They wanted productivity. They wanted fewer people asking for salaries, bathrooms, sick days, explanations. They wanted to tell investors they were riding the bright animal instead of being trampled by it.

So they brought in the machines and asked them to write the memos, summarize the meetings, screen the applicants, draft the reports, answer the customers, polish the garbage until it looked expensive under fluorescent lights.

And now the office is filling with soup.

They have a nice name for it, naturally. Workslop. Knowledge decay. The language is almost cute, like a children’s cereal for doomed middle managers. But the thing itself is older and uglier. Bad information goes in. Worse information comes out. People stop knowing what they know because the machine is always there with a confident little paragraph and the voice of a man wearing a borrowed suit.

The machine does not have to be right.

It only has to sound finished.

That is the dangerous part. A human fool usually leaves fingerprints. You can hear the doubt in him. You can see the sweat. He stammers, scratches his neck, says maybe, I think, I’m not sure, let me check. Those are useful noises. They are the little bells around the neck of stupidity, warning the rest of us not to build a bridge on it.

The machine arrives clean.

No cough. No shame. No memory of the last time it lied. It hands you a paragraph smooth enough to skate on, and some poor bastard with three deadlines and a manager named Bryce copies it into the bloodstream of the company.

Then the errors begin breeding in the walls.

A wrong number becomes a plan. A fake citation becomes policy. A summary that missed the one important sentence becomes the thing the vice president repeats at the next meeting with the calm authority of a man who has never had to clean his own mess. The workers notice. Of course they notice. Workers notice everything. They know when the floor tilts. They know when the new system is a shrine built to save money while costing everybody their sanity.

But noticing does not mean escaping.

The same bosses who could not be bothered to listen to people before suddenly need those same people to verify the output, correct the hallucinations, chase the ghosts, scrape the mold off the new miracle. The miracle writes fast. The human cleans slow. This is called efficiency by people who do not do either job.

I have known men like this.

Not machines. Men. The kind who made work for everyone else by being quick in the wrong direction. At the post office there was always some supervisor who wanted the numbers clean by noon, so he moved the mess behind a curtain and declared victory. The mail did not vanish. It sat there. Bags of it. Tubs of it. Human trouble with labels on it. But the report looked pretty for a few hours, and pretty reports can carry a mediocre man a long way in this country.

AI has become the perfect pretty report.

It gives management the feeling of motion. It lets them point at a dashboard and say transformation without choking. It lets them confuse activity with work and speed with thought. The machine produces something, and something is the favorite drug of the modern office. Nobody asks whether the something is any good until it has already been forwarded six times and used to justify a decision that will land on somebody else’s neck.

That is how trust dies.

Not all at once. Not with a gunshot. Trust dies like a bar tab grows. One small ugly surprise at a time. A colleague sends you a report and you wonder if he wrote it or if he merely babysat the machine. A recruiter rejects you and you wonder if a person ever saw your name. A manager praises an AI-generated plan and you wonder whether the whole building has quietly agreed to pretend soup is steak.

After a while, everyone starts checking everyone else’s pockets for dice.

The companies wanted workers to use AI. Then the workers used it badly, or grudgingly, or too much, or just enough to survive the ridiculous expectations created by the people demanding it. Now the same companies are shocked to discover that a workplace built on mistrust produces mistrustful work.

Beautiful.

Tell people the machine is smarter than they are, then blame them when they stop thinking. Tell them AI is mandatory, then complain when the mandatory tool leaves oil on the carpet. Tell them to move faster, faster, faster, and act surprised when they stop looking where they are going.

The word decay is right.

Decay is not dramatic. It does not kick down the door. It works in the dark. A beam softens. A tooth loosens. A file gets copied with the mistake intact. The knowledge leaves the room quietly because nobody practiced keeping it alive. Why remember how the thing works when the chatbot can produce five bullet points about it? Why train the kid when the system can generate the onboarding document? Why ask the old woman in accounting who knows where the bodies are buried when a glossy answer arrives in six seconds?

Because the old woman knows which body is still breathing.

Because she remembers the exception from 2014.

Because she can smell a bad number before the spreadsheet finishes loading.

Because human knowledge is not just information. It is scar tissue. It is embarrassment. It is the look on Eddie’s face the last time someone tried that clever shortcut and nearly set the place on fire. It is knowing which rule exists because of a lawsuit and which rule exists because Jerry in compliance enjoys suffering as a hobby.

The machine has none of that.

It has language. Mountains of it. Oceans. A flood so smooth it makes silence look guilty. But knowledge is not the same as language, and work is not the same as output, and a company is not smart because every fool inside it can now produce a confident memo before lunch.

Maybe some of this will settle down. Maybe the adults will come back into the room, tired and underpaid, and put the machine where it belongs: on the bench with the stapler, the calculator, the coffee maker, the other useful objects that do not get to run the building.

I am not holding my breath.

There is too much money invested in pretending the soup is getting better. Too many careers tied to the ladle. Too many men with clean shoes explaining that the next model will fix the mess made by the last model, which was also supposed to fix the mess before that. The future is always one update away from not smelling like a wet basement.

Meanwhile the workers sit under the lights, checking the machine’s homework, repairing the damage, trying to remember the old ways before the new ways make idiots of everyone.

The copy machine jams again.

Somebody says good enough.

And somewhere in the file, the gray smear becomes official.

Source: Companies That Embraced AI Are Now Rotting Away in a Very Specific Way