When AI promises productivity but delivers more work
Artificial intelligence has become the buzzword across companies everywhere.
It doesn’t matter the industry, the size of the company, or the type of work being done — the pressure to adopt generative AI tools has fully arrived in the corporate world, promising to transform how teams operate and, of course, boost productivity.
But there’s a detail nobody ran by the people actually doing the work.
While executives celebrate the gains this technology promises to deliver, a large portion of employees are living a very different reality, drowning in a phenomenon that Stanford researchers have dubbed workslop.
Never heard the term?
Think of it this way: it’s what happens when someone uses AI to generate work that looks fine at first glance but is actually full of errors, inaccuracies, or just doesn’t make sense. The result? Instead of saving time, the person next in line has to fix, rewrite, or even redo everything from scratch.
This is the picture that a survey of 5,000 American workers has started to reveal, and the contrast between what executives think and what employees actually experience day to day is, to put it mildly, surprising. 😬
Below, we dig into the data, the real stories, and what this gap tells us about how AI is being rolled out in companies today.
What the numbers say about productivity and AI
A recent survey, reported by the Wall Street Journal, polled more than 5,000 office workers in the United States and painted a pretty revealing portrait of how artificial intelligence is being perceived across different levels of the corporate hierarchy. The results show that 92% of senior executives say AI makes them more productive. Sounds great, right? Well, it really depends on where you’re sitting in this story.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
When the mic gets passed to the employees actually doing the day-to-day work, the tone shifts completely. No less than 40% of non-managerial workers say AI simply doesn’t save them any time at all. A significant chunk of these professionals report that, in practice, AI adoption has created an entirely new layer of work that didn’t exist before. They spend time checking, correcting, and even rewriting content generated by AI tools that reached them riddled with factual errors, nonsensical passages, or decontextualized information. In other words, instead of gaining time, these workers are burning extra energy fixing what the machine got wrong — and often without getting any credit for the additional effort.
This disconnect between leadership’s perspective and the experience of people on the ground isn’t small, and it has a technical name that’s gaining more and more traction in discussions about the future of work. Workslop, a concept coined by Stanford researcher Jeff Hancock, who co-authored the study behind the term, describes exactly this kind of output: work generated with AI assistance that looks passable on the surface but hides structural flaws that only show up when someone more experienced or more attentive takes a careful look. And the problem is that in fast-paced corporate environments, that careful look often doesn’t happen before the content gets passed along to the next person in the chain.
Workslop: the problem nobody wants to admit
The term workslop is a blend of work and slop — as in something done carelessly, without thought. And as much as it might sound like a direct dig at the technology, the issue goes way beyond that. The phenomenon isn’t AI’s fault per se, but rather a result of how it’s being integrated into company workflows.
According to Hancock himself, who in addition to being a Stanford researcher serves as a scientific advisor at BetterUp, people are being told to use AI often without any direction or support. When an organization adopts an artificial intelligence tool without proper training, without clear usage guidelines, and without well-defined review processes, the almost inevitable result is a flood of auto-generated content that arrives at the next step in the chain with serious embedded problems. And who foots the bill, every time, are the employees caught in the middle of that pipeline.
Hancock’s study, which has not yet undergone peer review, analyzed a subgroup of 1,150 office workers within the total sample of 5,000 and arrived at some very concrete numbers: 40% of these professionals had encountered workslop in the previous month, and they spent an average of 3.4 hours per month dealing with it. That might not sound like much, but when you scale that number to an organization with 10,000 employees, the estimate comes out to $8.1 million in lost productivity. That’s a lot of money going down the drain because of rushed implementations.
A real-world example helps illustrate the scope of the problem. Ken, a copywriter at a large cybersecurity company in Miami, shared (under a pseudonym, out of fear of losing his job) that the company’s CEO laid off several colleagues and instructed those who remained to start using AI chatbots, claiming it would increase productivity. In practice, while initial drafts appeared quickly, Ken and his coworkers ended up spending more time rewriting text, correcting errors, and resolving conflicts between outputs generated by different people’s chatbots than they would have if they’d never used AI at all.
Quality dropped significantly, the time to produce content actually increased, and worst of all, team morale tanked, according to Ken. When employees tried to report the productivity declines caused by AI, the company’s executives turned the blame back on the workers. This cycle creates a very serious transparency problem inside organizations. 😓
Real stories from people living with workslop
Ken’s case is far from an isolated one. Kelly Cashin, a freelance product designer, shared that she runs into workslop regularly in her day-to-day work. According to her, it’s become common for people to simply copy and paste a bot’s output directly into chats or emails. And when she questions a confusing passage that a colleague sent, the response is often something like: yeah, I don’t really know what the AI meant by that either. In other words, people are outsourcing their own judgment to the chatbot.
Cashin acknowledges that she understands why this happens. There’s enormous pressure to boost productivity, combined with real uncertainty in the job market. In that context, nobody wants to be seen as the professional who resists innovation, even if the tool is creating more problems than it solves.
The issue isn’t limited to the traditional corporate world, either. Philip Barrison, an MD-PhD student at the University of Michigan, conducted observations in primary care clinics where healthcare professionals were encouraged to use AI to generate email responses to patient questions. The idea was to save doctors and nurses time. But according to Barrison, both the reports he gathered and his own observations indicate that the time savings never materialized. Many professionals described considerable editing work, frustration, and concerns about data security and the possibility of patients receiving AI-assisted emails containing errors.
Since the tools were optional at these clinics, Barrison observed an interesting pattern: once the initial novelty wore off, the professionals simply stopped using AI. The tool was abandoned not out of resistance to technology, but based on a practical assessment that it wasn’t delivering on its promise.
The money that never came back: the AI investment dilemma
One of the reasons companies are pushing so hard for generative AI adoption in the workplace is the need to justify billions of dollars in investments already made. Aiha Nguyen, who leads the Labor Futures program at the nonprofit research institute Data & Society, explains that many companies are trying to cut labor costs after investing heavily in technology. But those investments haven’t paid off yet.
The numbers are telling. A frequently cited MIT report found that 95% of companies are not seeing a return on their AI investments. More recent assessments from giants like SAP and Deloitte show a larger fraction of companies generating returns, but they’re still in the minority. According to the Deloitte report, companies expect — or hope — that better returns will materialize in two to four years, which is considered slow for technology investments.
Meanwhile, companies like Block (Jack Dorsey’s company), Amazon, Dow, UPS, Pinterest, and Target have laid off workers while attributing the cuts to AI’s productivity potential. In other words, employees are already paying the price, even before the technology has proven it works as advertised. This creates an atmosphere of fear and insecurity that, paradoxically, worsens the quality of AI-assisted work, feeding the workslop cycle even further.
As Nguyen points out, the problem is that generative AI is being presented as a general-purpose tool capable of doing anything, but reality doesn’t work that way. And a good chunk of workslop stems precisely from this lack of clarity about when and what AI should actually be used for.
Why executives and employees see such different things
This perception gap between leadership and the operational workforce has pretty understandable roots when you take a step back to analyze it. Executives generally assess the impact of artificial intelligence through macro-level indicators: production volume, delivery speed, operational cost reduction, and engagement with new technologies. Those numbers can indeed show real improvements in certain contexts. Generative AI tools are genuinely useful for automating repetitive tasks, creating initial drafts, organizing information, and speeding up processes that used to take hours. The problem is that these metrics rarely capture the extra effort happening behind the scenes to make those deliverables come out with any real quality.
Employees, on the other hand, have a microscopic view of the problem. They deal daily with the limitations of the tools, they know when AI-generated text is wrong, when an automated analysis doesn’t make sense for the company’s actual context, and when a machine’s suggestion is going to create more problems than it solves. But they often don’t feel safe or encouraged to report these issues, whether out of fear of appearing resistant to innovation or because the company culture hasn’t created a safe space for that kind of feedback. The result is a silence that further feeds the illusion that everything is running smoothly at the top.
There’s also an element of implicit pressure that can’t be ignored in this equation. At many companies, AI adoption came paired with an unspoken message: those who don’t use the tools will be left behind. This kind of pressure leads employees to use the technology without adequate training, without understanding the limits of what it can and can’t do, and without criteria for evaluating when the machine’s output is reliable enough to move forward with. When AI use becomes a mandate disguised as a trend, the ground is perfectly set for workslop to flourish. And the productivity everyone wanted ends up existing only in the slide deck for the next board meeting.
The role of unions and regulation in this conversation
The workslop issue and the forced adoption of AI have already reached labor negotiation tables. Dan Reynolds, a research economist at the Communications Workers of America union, says AI has become a point of friction when unionized workers negotiate the terms of new contracts. Unions are demanding clearer mandates for the use of the technology, as well as greater worker participation and control over how AI is deployed in the workplace.
Sarah Fox, director of the Tech Solidarity Lab at Carnegie Mellon University, reinforces this point with a healthy dose of skepticism. According to her, when companies say they’re implementing AI to improve productivity, efficiency, and to help workers be better at their jobs, that actually obscures larger shifts in workplace dynamics. In practice, what often happens is a reduction in worker autonomy — not the empowerment that was promised.
Fox points out that companies are quite open about using AI to optimize operations, and the natural response from workers and researchers is to question what these tools can actually do and the power dynamics surrounding their use. This is a conversation that’s only just getting started and is likely to gain momentum as more data about AI’s real impact on work continues to emerge.
What could change in this equation
The good news is that this gap between executives and employees isn’t inevitable, and some organizations have already started to realize that how AI is implemented matters just as much as which tool is chosen. Companies that invested in real training for their teams, that created clear guidelines about when and how to use artificial intelligence tools, and that established robust human review processes before any final deliverable are seeing much more consistent results. The technology works best when it complements human judgment, not when it tries to replace it without any criteria.
Jeff Hancock himself, despite having documented the problem, believes that generative AI may eventually fuel tools that help workers improve efficiency. The key word is eventually, which depends on implementations that are more careful and less rushed than what we’ve been seeing so far.
Another point gaining attention in future-of-work discussions is the need to create more honest metrics for measuring AI’s impact on organizations. Measuring only delivery volume or production speed doesn’t tell the full story. It’s necessary to include quality indicators, review time, satisfaction of the professionals involved, and the real cost of rework. When that data enters the equation, the picture that emerges is far more complex than what executives typically see on their monthly dashboards. And it’s from this more complete view that better decisions about technology adoption and expansion can actually be made.
Ultimately, the conversation about workslop and the perception gap inside companies points to something that goes beyond technology: the need for more genuine dialogue between those who make decisions and those who carry them out. Artificial intelligence has real potential to transform team productivity — that’s not up for debate. What is up for debate is whether companies can implement this transformation in a way that works for everyone, not just for those watching the numbers from a distance. Because at the end of the day, it’s the employees who make the work happen, with or without AI. 💡
