Everyone can produce a lot more now. That’s the promise, and in one narrow sense it’s true. You can generate a slide deck, a first-draft memo, or a chunk of code in the time it used to take to open the file. The part that got left out of the pitch is that most of that output is mediocre, and someone still has to notice.
The word for it is workslop
An HBR and Gartner piece on the trends shaping work in 2026 gave this a name: workslop. It’s AI output that looks done, gets passed along, and turns out to need real cleanup once a competent person actually reads it. The email that’s technically coherent but says nothing. The analysis with a confident conclusion built on a number that’s wrong. The code that compiles and does the wrong thing.
The seductive part is that the person generating it feels productive. They finished the task in two minutes. The cost didn’t vanish, though. It moved downstream to whoever has to catch the problem, and catching a subtle error in someone else’s plausible-looking draft takes longer than writing it correctly the first time would have. Multiply that across a team and you get a lot of motion and not much output that’s actually good.
The returns aren’t showing up yet
If AI were quietly making everyone’s work better, you’d expect the money to show it. Mostly it doesn’t. The same HBR and Gartner analysis reports that only about one in 50 AI investments deliver transformational value, and only one in five deliver any measurable return at all. That’s a lot of spending for a lot of “we’re using it, we think it’s helping.”
The teams that did see results had something in common, and it wasn’t generating more. The ones that redesigned how they actually work around the tools were roughly twice as likely to beat their revenue goals. That tracks with what the flood of workslop implies. Volume was never the constraint. The constraint is judgment about what’s worth producing and whether the thing you produced is any good, and you can’t buy your way past that by generating faster.
Taste becomes the scarce thing
When output is cheap and abundant, the bottleneck shifts to the part that’s still hard. Someone has to decide what’s worth making, and someone has to tell whether the finished thing holds up. Both of those are judgment, and judgment doesn’t come out of a model.
This is quietly good news for a certain kind of worker. The person who reads the draft closely, notices the reasoning is thin, and knows how to fix it is more valuable in a world drowning in first drafts than in one where drafts were expensive. Their skill used to be somewhat invisible, folded into the general work of doing the job. Now it’s the specific thing that separates output people trust from output people have to double-check.
The same goes for producing good work in the first place. Anyone can get to eighty percent now. Getting to the version that’s actually right, that anticipates the objection and handles the messy edge case, still takes someone who understands the problem well enough to know when the easy answer is wrong. That gap between eighty percent and correct is where the value moved.
What this means for you
If you’re early in a career and worried the tools make you replaceable, this is the part to lean into. The skill that pays is being the person whose work doesn’t need to be redone, and the person who can look at a stack of AI-produced material and quickly tell what’s solid and what’s junk. Generating the most has stopped being the thing anyone’s short on.
That’s a learnable thing, and it’s mostly about caring enough to check. Read the output as if you’re the one who’ll be blamed when it’s wrong, because increasingly you are. Trace the number back to where it came from. Ask whether the confident conclusion actually follows. The habit of vetting, which used to feel like a slow tax on getting things done, is now one of the more defensible skills you can have.
There’s a hiring angle here too. Employers are starting to screen for exactly this. The ability to catch a mistake in AI output, to look at something plausible and explain why part of it is wrong, is becoming a thing interviewers probe for directly. Being able to walk through how you’d check a shaky draft, under a little pressure and in a real conversation, is worth practicing before it comes up for real.
The honest read
AI made output cheap, and cheap output is mostly mediocre. That sounds like a problem, and for teams that measure activity instead of quality, it is. For a careful worker, it’s the opposite. The value in your work has migrated toward the parts a model can’t do: deciding what’s worth making, and knowing when the finished thing is quietly broken.
The people who beat their goals didn’t win by generating more. They won by being deliberate about what they let out the door. That’s a standard you can hold yourself to starting now, and it’ll matter more as the slop piles up.
Questions, answered.
What is workslop?+
It's a term from an HBR and Gartner piece on 2026 work trends for low-quality AI output that looks finished but isn't, so colleagues end up spending hours reworking it. The cost doesn't disappear when someone generates a rough draft in seconds; it shifts to whoever has to clean it up.
Is AI actually paying off for most companies?+
Not yet, by most measures. The same HBR and Gartner analysis reports that only about one in 50 AI investments deliver transformational value and only one in five deliver any measurable return. Teams that redesigned their workflows around AI were roughly twice as likely to beat their revenue goals.
What skill becomes more valuable as AI output spreads?+
Judgment and taste. The ability to produce work that's actually good, and to spot when something is quietly wrong, is harder to automate and more scarce than the ability to generate a lot of plausible-looking material quickly.