Like an insidious virus, “workslop” has quickly infected almost every company that uses artificial intelligence tools.
Workslop is AI content that looks good on the surface but lacks depth, contains subtle errors and/or hasn’t been properly vetted. In a work culture that prioritizes speed over quality, workslop has become widespread: 40 percent of employees have received some form of it in the last month alone.
This substandard work carries real consequences. The additional workload added onto colleagues to revise this material costs $186 per employee per month. For professionals running their own companies independently, fixing substandard work makes achieving speed and efficiency with AI even more difficult.
Workslop isn’t just a temporary growing pain of AI integration, however; it’s persistent, and eradicating it will take a concerted effort from managers, employees and solo operators.
How to Eliminate Workslop
- Treat the problem as a management issue: Focus on employee education rather than technical fixes.
- Establish guardrails: Define specific use cases where AI is permitted and require manual verification.
- Keep a human in the loop: Treating AI output as a draft that requires human interrogation and subject matter expertise.
Why Workslop? Why Now?
Workslop is exploding as companies push employees to demonstrate AI usage and the technology exacerbates existing knowledge gaps.
BCG found that 82 percent of business leaders feel competitive pressure to invest in AI. We’re still in the initial stages of the enterprise AI integration era, so many of these leaders are racing to add AI tools to their tech stacks because of a fear of missing out.
Even sole proprietors aren’t immune to workslop. In pursuit of a competitive edge, solopreneurs might leap to outsource routine client reports or emails to AI. Any lapse in quality still reflects on them, however, not the AI platforms they employ, so this strategy can often backfire.
In the enterprise setting, if workers are unprepared to make the most of AI or leadership mandates them to use it, a dynamic of uneven output quality will quickly emerge. As those more skilled in using AI begin to increase their productivity, management will notice and applaud their usage. Meanwhile, those unequipped will feel pressure to pass AI-created work off as their own just to keep up. This pattern can hide an employee’s lack of subject matter expertise, perpetuating their reliance on AI.
Without the proper mechanisms in place to identify workslop, team leaders can inadvertently lower the internal standard of acceptable work while leaving other members to clean up the mess. Everyone using AI must take steps daily to solve this problem.
What Managers Can Do to End Workslop
Team leaders must treat workslop as an employee education and management problem, not a technical one.
Pearl’s recent proprietary research reveals 48 percent of white-collar workers feel ill-equipped to learn AI skills. So, managers must begin with the assumption that employees are starting from scratch and inform them of the limitations of AI’s “mental model.” They must emphasize that AI doesn’t think for itself, but rather predicts the next best logical response. As a result, workers should instead treat every response as a draft to interrogate, not a truth to blindly follow. Managers can also host demonstrations in which they reveal a large language model’s (LLM’s) shortcomings with nuanced questions about their industry to give employees the confidence that they still know best.
Explicit guardrails will solidify these best practices. Team leaders must establish cases for which they’ll allow AI usage, instances when workers must check its output and who else must verify its accuracy. These rules will vary in rigor between creative marketing and cybersecurity teams, for example, but they must always build upon company-wide AI use policy. Instead of setting and forgetting these rules, managers must actively maintain accountability for all work by asking employees to clarify which elements of their content are AI-generated and how they validated key facts or decisions.
Ultimately, managing around workslop requires communicating an openness to AI, so leaders can correct improper usage rather than guess when workers are masking AI responses. In the longer term, this openness will transform a team’s AI culture from speed to improvement. Managers that recognize specific cases in which team members used AI to improve clarity, depth and performance outcomes will be rewarded with more effective work.
How Employees and Sole Proprietors Can Fix Workslop
The same Pearl report found that 50 percent of white-collar employees believe that AI could handle half their job responsibilities in the next five years, but this doesn’t mean it actually should handle this proportion. Instead, workers must continually interrogate AI’s output and find their own unique strategies for combining AI’s work with their own creativity.
AI will only ever provide half a final answer; what’s missing is the due diligence on the employee’s part to prove its veracity and enhance it with nuance from their own experience. This practice is even more essential for sole proprietors, who build client trust on the uniqueness of their own insight.
Effective AI use also relies on avoiding the technology’s weaknesses and capitalizing on its strengths. Generalist AI often falls short of human expert knowledge, jeopardizing critical decisions and essential research. So, workers must also continually assess whether they’re using the best AI tool for every task.
Even with the best intentions, however, workers can still let workslop creep into their content if they fail to interrogate the limits of their own expertise. Prompting the model to conduct deeper research, comparing its statements to verified sources and identifying where it strayed beyond the original brief is only half the battle. For AI’s supposed statements of fact, employees should seek the truth outside of the model itself.
By continually maintaining these work standards, employees and solopreneurs will discover that not every problem needs an AI consultation. They’ll instead find the best instances to use AI in their own workflow and avoid using it for those where their creative thinking on its own is more efficient.
An LLM can be an initial researcher, pulling verified links under strict parameters for further reading. Or it can be a brainstorm buddy, helping flesh out multiple versions of a worker’s original creative thought. However individuals collaborate with AI, they should aim for the final product to fuse its power with their own judgement.
Moving Beyond Workslop
Just because AI makes mistakes doesn’t mean we should demonize it or call every AI error “slop.” Some inaccuracies are a natural part of working with a probabilistic tool. It’s when workers and managers fail to identify these slip-ups or pass them off as their own work that quality begins to erode.
Workslop has infected employees at every level, preventing them from using AI responsibly and effectively. Blanket AI bans, however, are impractical here; data from Pearl shows 46 percent of white-collar workers expect to treat AI as a coworker within the next two years.
Rather than a scorched-earth approach to eradicate this virus, managers, workers and solopreneurs must focus on disciplined use, or risk eliminating real productivity gains along with the slop.
