There is a scene that repeats in many companies now that generative AI has become a daily tool. Someone asks for a report, a proposal, or a summary. It arrives fast, well formatted, professional in tone. But read closely, it says nothing. The missing data has to be filled in, the errors corrected, and the work that looked done has to actually be done.

That content has a recent label: workslop. And behind the term is something more uncomfortable than an office anecdote.

What workslop is

Workslop is AI-generated content that looks polished but lacks real substance. It looks finished but it isn't.

The concept comes from research by BetterUp Labs together with Stanford University, reported by Harvard Business Review in 2025.

Its central finding is counterintuitive: misused AI does not save work. On the contrary, it relocates it.

Because the empty document doesn't disappear. It moves from one person to another. Whoever generates it feels they made progress; whoever receives it inherits the task of making it mean something. The work isn't eliminated — it changes desks.

The number worth looking at

The research puts figures to that intuition.

41% of workers said they had received workslop. And each instance costs, on average, close to two hours of rework: reading, spotting what's missing, correcting, redoing.

The figure gets more revealing when crossed with another. According to a 2025 MIT Media Lab report, 95% of organizations see no measurable return on their AI investment. Two numbers that, together, tell a single story: a lot of AI is being used, and much of that use generates activity without generating value.

It is not a technology problem. It is a problem of how it's used.

Why it happens

Workslop doesn't appear because the tool fails. It appears because it's easy to confuse speed with progress.

Generating a text takes seconds. Thinking about what that text should say, with what judgment and for whom, takes longer. When the pressure to "show progress" wins, the temptation is to hand over the first thing that comes out well formatted. The format hides the emptiness. For a while.

The problem is that the emptiness doesn't stay put. It travels. And each time it crosses a desk, someone pays the difference between what looks done and what is done.

The invisible cost: trust

There is damage that isn't measured in hours.

When someone repeatedly receives workslop, they start to distrust — not only the tool, but the colleague who used it without judgment. The silent question — "did they think about this or generate it and pass it to me?" — erodes something harder to rebuild than time: trust within the team.

And an organization that distrusts its own work becomes slow for other reasons. It over-reviews. It duplicates checks. It loses the very agility AI promised to deliver.

How to avoid it: judgment over volume

The antidote to workslop is not using less AI. It's using it with purpose. Lumen frames it with a simple distinction: AI is meant to amplify judgment, not replace it.

A few practices that bring order to that use:

1. Define what AI is used for — and what it isn't. Not everything that can be generated is worth generating. The first decision is where it adds value and where it just adds noise.

2. Set an explicit quality standard. A team that knows what "finished" means doesn't hand over drafts dressed as deliverables.

3. Put a person in charge of the output. AI proposes; someone decides whether it's ready to circulate. That checkpoint changes everything.

4. Measure value, not activity. The question isn't how many documents were produced, but how many solved something.

None of these practices is sophisticated. They all start from the same idea: the tool accelerates, but responsibility stays human.

Frequently asked questions

• What is workslop? It is AI-generated content that looks professional and finished but lacks real substance. Instead of solving a task, it shifts the effort to whoever receives it, who has to correct or complete it.

• How much does workslop cost? According to research by BetterUp Labs and Stanford reported by Harvard Business Review (2025), 41% of workers received workslop, and each instance involves roughly two hours of rework. In parallel, the MIT Media Lab (2025) reported that 95% of organizations see no measurable return on their AI investment.

• How can a company avoid workslop? By defining what AI is and isn't used for, setting a clear quality standard, assigning a human owner to each output, and measuring real value instead of production volume.

• Is the solution to use less AI? No. The problem isn't the amount of AI, but its use without judgment. AI applied well amplifies human work; applied poorly, it disguises it. The difference is oversight and purpose.

The underlying point

Generative AI made producing content almost free. That is an enormous advantage and, at the same time, a trap. When generating costs little, the judgment to decide what deserves to exist matters more than ever.

Workslop is what happens when that part is forgotten. It's the proof that efficiency without judgment isn't efficiency — it's well-formatted waste.

The good news is that the solution requires no new technology. It requires something older and harder to automate: thinking before sending.