In the day-to-day life of companies, many terms are used to define those “facade” works — well formatted, but without relevant content: sausage stuffing, nonsense, verbiage, rhetorical embellishment, verbal pyrotechnics and even aesthetic sophistry.
Now, researchers from , in the USA, have coined a neologism for the phenomenon — workslop — whose official description is: “AI-generated content that masquerades as good workbut it lacks the substance necessary to actually drive a task forward.”
In a recent article, the authors warn that, to meet goals for routine use of AI — determined by managers concerned with showing a quick return on investments — employees have turned to technology for tasks that they could better solve without it.
The result has been an exponential increase in workslop, that is, superficial, poorly directed content, which, at first, may even seem productive, but which only increases the rework of the recipient colleagues. Ultimately, it works like a “handbrake on” for productivity.
Research carried out by teams from Stanford and BetterUp Labs revealed that 41% of workers have already encountered this type of production “made up” by AI, “costing almost two hours of rework per instance and creating subsequent productivity, trust and collaboration problems”, says the article.
Deeply investigating the workslop
It is not possible to say that the current article is inventing the workslop: the practice is well known in corporate jargon. Who has never seen memos full of buzzwords like “empowerment” and “disruption”, or a report flooded with bullet points, which, in essence, say nothing?
Because generative AI has, it’s common to receive a generic AI-created email, or a vendor order along those lines. However, things change when the message comes from a colleague or a boss.
Recipients of these “empty” messages gave different testimonies to the researchers: a department head was torn between redoing a subordinate’s project or sending it back for revisions, while others spent much of their time worrying about justifying subpar work.
In one of these cases, the director of the Stanford Social Media Lab, Jeffrey Hancock, reports the reaction of a project manager who received a robotic email from his supervisor: “I had to scramble to do something that should have been her responsibility, which hindered my other ongoing projects,” the executive reportedly said.
After surveying employees from 1,150 companies, 40% of respondents said they had received a workslop sample from a colleague in the last month. The worst thing is that the phenomenon doesn’t just spread between peers; Managers also send these empty reports in all directions.
So while companies may be spending hundreds of millions on creating efficiencies and increasing productivity, and encouraging employees to use it liberally, they may also be creating rework, confusion and wasted timeas if they were planting seeds of inefficiency.
Tips to avoid workslop
Since, in a highly competitive market, companies cannot afford to give up a technology that, according to the authors, “”, the solution to the workslop problem is to set up protections and train employees, says the study.
The first step is to define when AI is appropriate or not, a kind of “literacy” in which employees are instructed to treat the production of artificial intelligence as that of an untrained intern. Only by knowing the limitations of the tools can people avoid blindly trusting the answers.
Another recommendation from the anti-workslop playbook is to be specific about when AI is appropriate. For the authors: “When organizational leaders advocate AI everywhere all the time, they model a lack of judgment in how to apply the technology,” which translates into copy-and-paste.
The idea is to use AI to polish work, not create it, the study says. This means employees should continue writing drafts with relevant information and context before using technology as a thought partner. The researchers suggest that the human and AI partnership should follow the same rules.
The final tip is to invest in clear communication. The authors recommend that, before using a large language modelemployees master interpersonal communication. The spread of workslop not only delays processes, but destroys team trust, with colleagues losing respect for each other.
In addition to the emotional cost, each workslop occurrence generates a real cost for companies of US$186 per month, according to the authors’ calculation, based on the average salary of those interviewed. In an organization with 10,000 employees, this is equivalent to an annual loss of US$9 million (approximately R$48 million).
