One of the mantras most repeated by the apostles of AI automation is that AI-assisted work was going to boost productivity in companies that apply it, but the data shows that the reality is very different, depending on how that productivity is measured.
For example, data from the study ‘Forrester Consulting’s Total Economic Impact 2023′ from IBM, highlight an increase in productivity based on a 30% reduction in the time it takes to manage an incident, but it does not measure the quality of that management. It is at that point where AI, rather than boosting productivity, is sinking it.
The effect “Workslop“. As they say in Harvard Business Reviewin many companies, the mass adoption of AI tools translates into enthusiasm and apparent advances, but behind these figures lies an increasingly evident problem: the proliferation of mediocre content generated by AI, known as “workslop” or work garbage. The phenomenon occurs mainly when AI is used to produce documents, reports and materials that are very apparent at first glance, but are superficial at their core and end up generating more work in reviewing and correcting them than it would have taken a person to do it from the beginning.
A recent study conducted by BetterUp Labs together with Stanford Social Media Lab reveals that 40% of American employees reported having received content “workslop” in the last month. Data indicates that 15.4% of all the content they received at work falls into this “Workslop” category. BetterUp Labs estimates put the cost of reviewing AI-generated work at $186 per employee, or about $9 million per year for a large corporation with 10,000 employees.
It is useful to get rid of “paperwork”. AI is proving useful in routine tasks, such as email automation, simple summaries or basic content generation, allowing the employee to relieve cognitive load. That is, freeing your brain from work that, although necessary, does not really represent progress in tasks or projects. The report Genes in Divide (MIT, 2025) confirms that 70% of employees prefer to use AI to compose quick communications and perform simple analyses, noting that “AI has already won the battle of easy work.”
However, for complex projects and jobs that require memory, continuous adaptation and deeper analysis, 90% still prefer to turn to human professionals. Research from Carnegie Mellon University (CMU) and Duke University indicates that AI can serve as a starting point for developing an idea, but fails in 70% of cases in which tasks are asked to complete unattended.
The invisible “tax” of AI. Every time an employee receives AI-generated workslop content, the process requires an additional investment of time and resources to unravel and correct any errors or inaccuracies that come with it.
The aforementioned BetterUp Labs study estimates that each employee wastes an average of one hour and 56 minutes analyzing or reviewing this “junk” content. So much so that it has even given rise to the birth of a new professional niche in which professionals who previously carried out this work now charge for analyzing it and fixing their mistakes.
The biases of AI. The study also analyzes the social and labor impact of this type of content. His conclusion: it is as harmful as the economic impact. 53% of employees say they feel upset after receiving these texts and 38% say they are confused. As published Forbesapproximately half of those surveyed consider their colleagues who submit workslop work to be less creative, less capable. Furthermore, 42% say they see them as less trustworthy, generating a deterioration in reputation and collaboration within the team.
This social impact does not originate from using AI to generate documents, code or graphics, but rather from not having taken the time to check whether the AI-generated content is correct before sending it or using it in your work.
Use AI with common sense. Researchers at MIT and BetterUp Labs agree that using AI indiscriminately, only following the mandate to adopt the technology, as some big technologies want to do at all costs, is not a good idea to increase productivity.
According to what was published by CIOdespite the fact that the CEO of Google beats his chest by ensuring that 25% of his code is already generated with AI, this work is not free nor does it result in notable improvements in the productivity of his engineers. Before they were dedicated to generating code, and now they use that time to review it or repair the errors produced by the new integrated code. Therefore, using AI on complex tasks that must then be overseen by engineers does not improve productivity, but displaces it at best and may even reduce it depending on the use case.
In WorldOfSoftware | We believed that AI was going to take our jobs. At the moment he has started whispering to your boss who he should fire
Imagen | Unsplash (Sigmund)
