Harvard Business Review cautions that AI 'workslop' is deteriorating companies internally.

      **Summary**

      Harvard Business Review highlights that companies heavily investing in AI are facing “knowledge decay,” where low-quality outputs accumulate, erode trust, and lead to an annual loss of approximately $9 million due to rework. Organizations that aggressively adopted generative AI are now grappling with deteriorating work quality, which contradicts the technology's intended improvements. Two recent Harvard Business Review articles explain this cycle, where AI-generated poor-quality work undermines the information companies utilize for decision-making, a phenomenon termed “knowledge decay.”

      In a June 2026 article, professors Matthias Holweg and Thomas Davenport argue that the repercussions extend beyond individual errors. When employees utilize AI to produce seemingly polished but flawed work, their colleagues waste time on verification and corrections, leading to a decline in the overall knowledge base within the organization.

      Low-quality AI output has been termed “workslop” by BetterUp Labs and Stanford’s Social Media Lab in a September 2025 article, denoting AI-generated content that appears good but lacks the necessary substance. A survey of 1,150 full-time U.S. workers revealed that 41 percent encountered workslop in the prior month, requiring an average of one hour and 56 minutes to resolve each issue.

      The financial impact is considerable. Based on survey responses, researchers estimate that workslop costs around $186 per worker monthly. For a company with 10,000 employees, this results in over $9 million in annual productivity losses, not including the negative effects on morale and trust.

      The social repercussions may be even more significant. In the BetterUp-Stanford survey, 53 percent of workers who received workslop reported feeling annoyed, 42 percent perceived the sender as less trustworthy, and roughly half considered the colleague less creative or reliable. One-third indicated they would be less inclined to work with that individual again.

      The overall productivity landscape is equally troubling. A July 2025 MIT Media Lab report found that 95 percent of organizations reported no measurable return on their generative AI investments, despite expenditures in the billions. Goldman Sachs reached a similar conclusion in March 2026, noting no significant correlation between AI adoption and productivity gains at a broader economic level, even as 70 percent of S&P 500 management teams discussed AI during earnings calls.

      Knowledge decay differs from the commonly acknowledged issue of AI hallucinations, which involve factual inaccuracies in AI outputs. Instead, knowledge decay pertains to the degradation of an organization’s information quality over time due to the accumulation of errors and subpar AI-generated work. This results in a lack of trust in internal documents, unreliable processes, and diminished institutional memory as employees rely on AI over personal expertise.

      Holweg and Davenport caution that the hiring process has been especially harmed. AI-generated resumes inundate recruiters, AI-generated job postings mislead candidates, and AI-driven screening tools may filter out qualified individuals. Consequently, trust in the hiring process has reached "all-time lows for both job seekers and recruiters," according to HBR.

      A measurable worker backlash is already evident. A 2026 survey of 2,400 employees in the U.S., UK, and Europe indicated that 29 percent actively undermined their employer’s AI strategy by ignoring guidelines, avoiding training, or intentionally altering performance data. Among Gen Z workers, this figure climbed to 44 percent, primarily due to concerns about job security.

      This pushback occurs alongside a broader trend of AI-related layoffs, often without clear evidence that the roles eliminated were genuinely replaced by AI systems. In 2026, the tech sector saw over 95,000 job cuts across 247 incidents, nearly half of which were attributed to AI, despite doubts about the maturity of those companies' AI implementations.

      Ironically, addressing the workslop issue necessitates the kind of labor that AI was supposed to alleviate. Business leaders must now allocate resources toward verification processes, quality standards, and human oversight to ensure that AI-generated content meets acceptable criteria, which consumes employees’ time. HBR suggests establishing an additional layer of human review around AI output, undermining the efficiency rationale for AI adoption.

      Both HBR articles distinguish between blanket AI mandates and targeted implementations. The June article notes that proprietary models trained on company-specific data can deliver real value, while public large language models used in ill-suited tasks create "generic prose that frequently contains errors." Companies that froze hiring based on anticipated AI productivity gains are discovering that these gains might be unfounded if work quality deteriorates quicker than the workforce shrinks.

      The knowledge decay framework reshapes the conversation around AI productivity. The focus has shifted from merely whether AI speeds up individual tasks to whether widespread AI use ultimately enhances or diminishes an organization’s decision-making capabilities. HBR concludes that, for firms that adopted AI without effective quality controls, the outcome is detrimental.

      While Holweg and Davenport’s credentials add credibility to this argument, it is important to note that the knowledge decay concept has not yet been validated through controlled empirical studies