Nobody Trusts Anyone's AI Work Anymore

Watch what happens in a meeting when an executive reads out a plan that obviously came from ChatGPT. The team goes quiet in a particular way. Then watch the same executive receive a report a junior person generated with AI. The face does the same thing.

Both sides are having the identical reaction to each other, and until recently there was no data on it. Now there is, and it is worse than the anecdotes.

Zety surveyed 1,000 US employees in January 2026 and found that 55% receive workslop, AI output that looks polished but has nothing inside it, from their own managers or supervisors. Not from interns. From the people above them. Meanwhile Founder Reports' 2026 compilation found the mirror image: 57% of managers have had to fix AI-created work from their teams, and above the senior-manager level the rate passes 60%. At companies that mandate AI use, 73% of workers have had to fix a coworker's AI output.

Everyone is handing everyone else hollow work, and everyone can tell.

The trust gradient

SHRM's 2026 workplace AI study found something that should worry anyone running a company. Asked whether they trust senior leaders when those leaders talk about AI, 80% of directors said yes. Among managers, 65%. Among individual contributors, 47%.

The people closest to the actual work trust the AI story least, and heavy AI use by leadership makes it worse: 26% of individual contributors said their trust in a leader drops when that leader leans hard on AI. A leader who thinks their ChatGPT-drafted memo reads as decisive should sit with that number. To a meaningful share of the room, it reads as absence.

The Stanford and BetterUp research that coined "workslop" measured what receiving it feels like: 53% annoyed, 38% confused, 22% offended. Offended is the interesting one. People experience hollow AI work from a colleague as a small insult, because it is one. It says your time is worth less than the thirty seconds this took me.

The cost is measurable too: roughly two hours of rework per incident, about $186 a month per affected worker, an estimated $9 million a year for a 10,000-person organization. But the money is the small loss. The large one is what the rework does to how people see each other.

The resentment goes underground

Here is the part that makes this dangerous rather than just annoying. Business.com surveyed 1,055 workers at AI-using companies about the AI behaviors they hate most. The top five were all the same behavior in different clothes: shipping unchecked work, passing AI output off as your own, trusting a chatbot over a colleague's expertise, sending unedited AI text, outsourcing your own thinking. People have not rejected AI. They have rejected AI used as a substitute for effort, which is a judgment about the person, not the tool.

And then the finding that should stop you: 67% of workers never said anything to the colleague whose AI habit annoyed them. Two thirds. The resentment doesn't surface. It accumulates.

Where it goes instead is now documented. Writer and Workplace Intelligence surveyed 2,400 knowledge workers in 2026 and found open sabotage of AI rollouts: employees avoiding the tools, ignoring the new processes, pretending to use systems they never touch. Reporting by Aftermath found workers building private defenses, disclosing "this draft was not generated by AI" and asking colleagues not to run their work through a model, and a games-studio team that decided, in their own words, to remain human, typos and all.

When people start certifying their work as human-made, the way food gets certified organic, the trust problem is no longer a rough patch. It is the environment.

What the flinch is actually detecting

It would be easy to read all this as resistance to change, and some leaders do. The data says otherwise. The same surveys show most workers think AI helps them personally. What they cannot stand is other people's unreviewed AI, and there is a precise reason.

AI output carries confidence with no relationship to the thinking behind it. A generated plan sounds equally assured whether it was checked by someone with judgment or pasted straight from the chat window. Humans turn out to be very good at sensing that gap, better than any dashboard. The flinch in the meeting is a detection event. It is the organization's immune system noticing confidence without substance, months before the missed quarter makes it official.

Which means the fix is not less AI, and it is certainly not mandating more of it, the mandate companies had the highest rework rates. The missing thing is visible judgment. Every piece of the resentment data points at the same absence: nobody can see whether a human actually reviewed, compared, chose, and stood behind the work. When that step is invisible, every polished document is suspect, and everyone quietly re-checks everything, which is where the two hours per incident goes.

Make the judgment visible

This is the part we built ALLO around. AI output lands on a shared canvas next to the brief, the references, and the other options, where a team can see it get compared and see a person choose. The reasoning stays attached to the work. When something ships, it carries the visible history of the judgment that shipped it, who looked, what was rejected, why this one.

That changes what a deliverable means. A document arriving out of nowhere with AI polish on it invites the flinch. The same document arriving with its comparison and its decision attached answers the only question anyone was really asking: did a human actually think about this.

The trust that AI burned this year will not be restored by better models. The models will get better and the flinch will get sharper, because the gap being detected was never about quality. It is about whether anyone stood behind the work. Show that, and the rest of it, the rework, the silence, the quiet certification of humanity, starts to unwind.

FAQ

What is workslop? AI-generated work that looks finished but lacks the substance to move work forward, forcing someone downstream to redo it. The term comes from Stanford and BetterUp Labs research, which found 41% of desk workers encountered it in the past month, costing about two hours of rework per incident.

Do managers produce workslop too? Yes, and at scale. Zety's January 2026 survey of 1,000 US employees found 55% receive workslop from their own managers or supervisors, and 19% of workers say heavy AI reliance by leadership lowers their trust in those leaders.

Why do people resent AI-generated work from colleagues? Surveys show the hated behaviors all involve AI replacing effort rather than supporting it: unchecked work, unedited text, passing AI output off as one's own. Receiving it reads as a statement that your time matters less than the seconds it took to generate. 22% of recipients report feeling offended.

Does mandating AI use fix this? The opposite. At companies requiring AI use, 73% of workers have had to fix a coworker's AI output, the highest rework rate measured, and separate research documents employees actively sabotaging mandated rollouts.

How do teams rebuild trust around AI work? By making human judgment visible: showing that AI output was compared, reviewed, and chosen by a person, with the reasoning attached. ALLO puts that process on a shared canvas so a deliverable carries the history of the judgment behind it.

--
Workslop is AI-generated work that looks finished but lacks the substance to move anything forward, so someone downstream has to redo it. In a January 2026 Zety survey of 1,000 US employees, 55% said they receive workslop from their own managers or supervisors, and 45% said it has made them more cautious about AI at work. The Stanford and BetterUp research that named the term found 41% of desk workers encountered it in the past month, at a cost of roughly two hours of rework per incident. The damage is not the wasted time. It is trust, in both directions.