Harvard Business Review cautions that AI 'workslop' is undermining organizations from within.

      **Summary**

      Harvard Business Review highlights that companies heavily invested in AI are experiencing “knowledge decay,” where low-quality outputs accumulate, eroding trust and incurring costs of $9 million annually for rework. Businesses that aggressively adopted generative AI are now facing a challenge that the technology was meant to alleviate: a decline in the quality of their work. Two recent HBR articles describe a cycle where low-quality AI-generated outputs diminish the quality of information vital for decision-making, a process termed “knowledge decay.”

      The June 2026 article by professors Matthias Holweg and Thomas Davenport indicates that the fallout extends beyond simple mistakes. When employees leverage AI to create seemingly polished work that is erroneous or lacking in substance, others spend time verifying and correcting these flaws. As these mistakes accumulate across teams, the overall knowledge of the organization deteriorates.

      The term “workslop” was introduced in a September 2025 HBR piece by BetterUp Labs and Stanford’s Social Media Lab to define AI-generated content that appears competent but lacks the necessary depth to complete a task. In a survey of 1,150 full-time US workers, it was found that 41% encountered workslop in the previous month, with each instance requiring nearly two hours on average to address.

      The financial implications are substantial. Based on self-reported salaries and time estimates from respondents, researchers determined that workslop costs around $186 per worker monthly. Consequently, for a company with 10,000 employees, this results in over $9 million lost each year in productivity, not factoring in the repercussions on morale and trust.

      These social costs may be even more significant than the financial ones. In the BetterUp-Stanford survey, 53% of those who received workslop expressed annoyance, while 42% viewed the sender as less trustworthy, and about half perceived the colleague as less creative, capable, or dependable than before. A third reported they were less inclined to collaborate with that person again.

      The overall productivity landscape is similarly grim. A July 2025 MIT Media Lab report revealed that 95% of organizations did not see measurable returns on their generative AI investments despite significant expenditures. Goldman Sachs reached a comparable conclusion in March 2026, finding no substantial link between AI adoption and productivity increases at the macroeconomic level, even with 70% of S&P 500 management teams discussing AI during earnings calls.

      Knowledge decay is a distinct issue from known problems like AI hallucinations, which are factual errors in AI outputs. Knowledge decay refers to the decline in an organization’s quality when errors and the trend of low-quality AI-generated work accumulate over time.

      As trust in internal documents diminishes, processes based on unreliable data yield unreliable results. Institutional knowledge suffers as employees rely on AI instead of cultivating their own expertise. Holweg and Davenport highlight that the recruitment process has particularly suffered; AI-created resumes inundate recruiters, misleading job seekers, while AI-driven screening tools filter out qualified candidates. Consequently, trust in the hiring process has reached “all-time lows for both job seekers and recruiters.”

      There is already measurable worker backlash. A 2026 survey of 2,400 employees across the US, UK, and Europe found that 29% admitted to sabotaging their employer’s AI strategy by disregarding guidelines, avoiding training, or skewing performance data, with 44% of Gen Z workers engaging in such resistance, largely due to fears of job loss.

      This pushback coincides with a broader trend of layoffs justified by AI, despite lacking clear evidence that AI displaced the roles eliminated. The tech sector saw over 95,000 job cuts in 2026 across 247 events, nearly half attributed to AI, although analysts have raised concerns about whether many of those companies had established AI systems sophisticated enough to handle the work.

      Ironically, addressing the workslop issue necessitates the kind of labor that AI was intended to reduce. Business leaders now need to invest in verification processes, quality standards, and human oversight to ensure that AI-generated content meets established criteria, which demands real employee time. HBR suggests creating a new layer of human review around AI outputs, countering the efficiency argument that initially justified AI adoption.

      Both HBR articles differentiate between indiscriminate AI use and strategic applications. The June article notes that proprietary models trained on company-specific data can add real value, while public LLMs may produce “generic prose containing errors” when applied to unsuitable tasks. Companies that halted hiring due to anticipated AI productivity gains are now realizing those gains may be illusory if the quality of output declines more rapidly than their workforce shrinks.

      The knowledge decay framework reframes the discourse on AI productivity. The question evolves from whether AI speeds up individual tasks to whether widespread AI use improves or hinders organizational decision-making. HBR’s conclusion for companies that integrated AI without maintaining quality controls is that it worsens decision-making.

      While Holweg and Davenport's credentials add credibility, it's