Someone used AI to save twenty minutes. Three departments will spend the next month living with the answer.
THE DECK LOOKED GREAT.
Professional.
Concise.
Twenty-three slides.
Nice hierarchy.
Bullet points exactly the correct distance apart.
There was even a framework.
Nobody knew where the numbers came from.
Minor issue.
This is the workplace version of AI slop, and unlike the executive LinkedIn post, it does not merely embarrass the person whose photograph is attached.
It enters the bloodstream.
The report becomes a meeting.
The meeting becomes a decision.
The decision becomes a process.
The process becomes documentation.
The documentation becomes input for the next AI system.
Six months later the organization is citing something nobody remembers establishing in the first place.
Congratulations.
Your company has begun eating its own homework.
Corporate Slop Predates the Robot
Before blaming ChatGPT, we should acknowledge something.
Companies were already extremely good at generating useless material.
Colin Ellis points out that corporate slop existed long before generative AI: needless processes, bloated forms, unresolved behaviors, endless emails, pointless pet projects, unnecessary meetings and constant reaction to executive whims all consume employee time without adding meaningful value.
This is crucial.
AI did not introduce organizations to inefficiency.
Organizations are artisanal producers of inefficiency.
We made meetings that could have been emails.
Then emails that could have been nothing.
Then decks explaining the emails.
Then dashboards tracking whether employees read the decks.
Human beings did this ourselves.
What AI changes is production speed.
You can now make the unnecessary deck in thirty seconds.
Workslop Is Someone Else's Problem
The seductive thing about AI productivity is that the benefit is immediate.
You had to write a project brief.
AI wrote it.
Time saved:
45 minutes.
Fantastic.
Then Maria reads the brief.
Something seems wrong.
She checks two claims.
One source does not exist.
She messages you.
You check.
Then she has to rewrite the implementation assumptions.
Then finance discovers that one figure came from an outdated forecast.
Then somebody schedules a meeting.
Your 45-minute productivity gain has now become four people's Tuesday.
This is workslop.
Not simply bad AI output.
Bad output that transfers the cognitive burden from the person who generated it to the person who receives it.
Harvard Business Review's Matthias Holweg and Thomas Davenport argue that when this happens repeatedly across connected business processes, the problem can expand into deterioration of the organization's stored knowledge itself. They describe an organization-level form of decay in accuracy and quality when slop moves sequentially through workflows.
That is much more interesting than:
ChatGPT hallucinated.
The hallucination got promoted.
The Productivity Gain Is Hiding in Someone Else's Calendar
We tend to measure AI efficiency locally.
Did you finish faster?
Wonderful.
But companies do not operate locally.
Your output becomes somebody's input.
This means an AI workflow that saves the creator thirty minutes while requiring three downstream employees to verify it is not productive.
It is a time-transfer scheme.
Workslop does not eliminate labor. It relocates labor to whoever is unfortunate enough to receive it.
This is why measuring “hours saved” can become such corporate comedy.
Employee asks AI for competitive analysis.
Five minutes.
Analyst spends two hours checking it.
Manager spends an hour correcting assumptions.
Meeting spends forty minutes debating the wrong number.
Dashboard reports:
AI SAVED EMPLOYEE 55 MINUTES THIS WEEK.
Amazing quarter.
Knowledge Decay Is Worse Than a Bad Answer
The dangerous part is not one bad document.
Organizations survive bad documents constantly.
The dangerous part is repetition.
Imagine a strategy memo containing a plausible but inaccurate market assumption.
That memo gets summarized for leadership.
Leadership's summary goes into a planning document.
The planning document becomes the basis for a budget.
An AI assistant later retrieves the budget and explains why the company chose the strategy.
Now five artifacts agree.
This creates the appearance of corroboration.
Except all five share one ancestor.
A mistake.
Strategic Marketing Tribe, summarizing the concern raised in HBR, calls this knowledge decay: low-quality generated output gets fed into later work, mistakes consume verification time, trust falls, and an organization's accumulated knowledge can gradually become less reliable.
This is almost biological.
Bad information reproduces.
The Company Stops Remembering Why
Organizations have memory.
Not a brain.
People.
Documents.
Processes.
Emails.
Databases.
Stories.
Bob knows why the warehouse system was configured that weird way because Bob was there when the old one caught fire.
Then Bob retires.
So somebody documents it.
Excellent.
Now imagine an AI system summarizing Bob's documentation.
Then another AI system summarizes that summary.
Then an employee uses the second summary to rewrite the official process guide.
What disappears first?
Probably the annoying part.
The exception.
The historical reason.
The bizarre edge case.
The sentence that says:
Do not perform this step on international orders because of the 2017 tax ruling.
Too specific.
Clean it up.
Simplify.
Make concise.
Now the documentation is more readable.
It is also wrong.
Corporate knowledge often lives in the ugly exception. Slop loves removing ugly exceptions.
This is the difference between information and institutional knowledge.
Information says what normally happens.
Institutional knowledge remembers why the weird thing exists.
Nobody Wants to Verify the Nice-Looking Thing
Presentation quality makes this worse.
Humans trust polish.
A messy spreadsheet announces:
Please inspect me.
A beautiful AI-generated report announces:
Already handled.
The headings are correct.
The writing is confident.
The citations look citation-shaped.
The chart is attractive.
Why ruin everyone's day by asking whether the chart has any relationship to Earth?
HBR's adjacent guidance on AI at work increasingly emphasizes judgment as the scarce skill: as polished output becomes easier to generate, employees need to know what to trust, challenge and refine rather than merely how to produce faster.
This is exactly right.
The future knowledge worker may spend less time producing first drafts.
Fine.
Then the job moves toward:
Is this true?
Is this useful?
What is missing?
What does the machine not understand about this company?
Those are not AI skills.
They are judgment.
Annoyingly human again.
Your AI Read the AI That Read the AI
Eventually the recursion gets funny.
A salesperson uses AI to summarize customer calls.
A manager uses AI to summarize the summaries.
Marketing asks AI to identify themes in the manager's summary.
Strategy asks another AI to produce a presentation from marketing's themes.
The CEO asks an assistant for the three most important insights.
At the end of this chain, the customer said:
The button is confusing.
The CEO receives:
Customers increasingly prioritize intuitive digital experiences, presenting a strategic opportunity to accelerate human-centered transformation.
Nobody is technically lying.
Somehow truth has still died.
This is corporate slop in its purest form.
Every step improves the language.
Every step removes the event.
Automation Loves Formal Processes
There is another irony.
Companies often deploy AI into dysfunctional processes without fixing the processes first.
Ellis's argument about old-fashioned corporate slop is useful here: organizations already have bloated meetings, unnecessary approvals, outdated procedures and low-value work.
Then AI arrives.
Instead of asking:
Why do we do this?
the company asks:
Can AI do this faster?
This is how you end up automating a report nobody should produce.
A form nobody should fill out.
A meeting nobody should attend.
A workflow whose original employee retired during the Bush administration.
Automation does not make a stupid process intelligent.
It makes stupidity scalable.
The Slopification Loop
The full loop looks something like this:
Someone needs output quickly.
AI generates output.
Nobody verifies it properly.
The output enters the organization.
Someone else treats it as source material.
AI summarizes it again.
Context falls away.
Error survives.
Employees become less confident in internal documents.
More verification is required.
Trust declines.
AI is introduced to help employees deal with the growing information burden.
Please admire the machine.
This is a beautifully engineered trap.
The cure produces demand for more cure.
People Stop Knowing Things
The long-term danger is deskilling.
Not because AI automatically makes people stupid.
Because skills decay when they are not practiced.
If nobody writes the first analysis, fewer people learn how to structure the analysis.
If nobody does the first research pass, fewer people learn which sources are suspicious.
If nobody builds the model, fewer people understand what the model assumes.
Then when the AI output looks wrong, there is nobody left in the room who can explain why.
This is when the organization's relationship with AI changes.
It stops being a tool.
It becomes infrastructure.
Then dependency.
The company knows things because the system says it knows them.
The Good Company Will Be Slower in Weird Places
This sounds anti-productivity.
It isn't.
Use the tools.
Automate the genuinely repetitive work.
Draft.
Summarize.
Translate.
Explore.
Search.
Calculate.
But keep human friction wherever mistakes compound.
High-consequence decisions.
Core strategy.
Institutional memory.
Financial assumptions.
Medical decisions.
Legal interpretation.
The weird exception.
The information somebody will cite twelve months from now after everyone in the original meeting has forgotten what happened.
Maybe the best AI organization will not be the company that automates the most.
It will be the company that knows where not to.
Efficiency is not producing more output per employee. It is reducing the amount of work everyone else has to undo.
The Deck Still Looks Great
This is what makes corporate AI slop difficult.
It does not necessarily look broken.
It looks better than the human version.
Cleaner.
Faster.
More consistent.
More professional.
That is why it survives.
The organization keeps rewarding the visible output while the invisible verification work accumulates underneath.
Eventually everyone becomes incredibly productive at producing things nobody trusts.
And somewhere in the company, buried inside twelve summaries, four decks and an AI-generated quarterly strategy document, the original mistake is still waiting.
Beautifully formatted.
