AI 'workslop' backlash
Workers and reporters say AI is producing polished but flawed outputs that increase review and correction work rather than speeding delivery, a phenomenon labelled "workslop." (theguardian.com). Social commentary and analysis add that DORA metrics can fall in large organisations and that firms have burned AI coding budgets when review and operational costs outpaced local productivity gains ( ).
Workers say artificial intelligence is creating a new kind of office drag: polished-looking output that is wrong often enough to leave colleagues doing the cleanup. (theguardian.com) The term for it is “workslop,” shorthand for low-effort, artificial-intelligence-generated work that looks finished but pushes checking, rewriting and fact-finding onto the next person. Harvard Business Review said in January that this kind of output “offloads cognitive work onto the recipient.” (hbr.org) Harvard Business Review reported in September 2025 that 41% of workers had encountered this kind of artificial-intelligence output, and that each instance created nearly two hours of rework. The article said many companies were seeing little measurable return on their generative artificial intelligence spending. (hbr.org) That complaint is colliding with a second set of numbers from software teams. Google Cloud’s 2024 DevOps Research and Assessment report said a 25% increase in artificial intelligence adoption was associated with a 1.5% drop in delivery throughput and a 7.2% drop in delivery stability, even as many developers said they felt more productive. (cloud.google.com) DevOps Research and Assessment metrics are the standard scorecard many engineering groups use to track how often they ship code, how long changes take, and how often releases break things. Google Cloud said more than 75% of respondents used artificial intelligence for at least one daily professional task, but system-level results did not automatically improve. (dora.dev; cloud.google.com) New York Federal Reserve researchers added a wider labor-market snapshot on April 14. Using the November 2025 Survey of Consumer Expectations, they found 39% of employed respondents had used artificial intelligence at work in the prior 12 months, and about 66% of those users said the tools raised their personal productivity. (libertystreeteconomics.newyorkfed.org) The same New York Federal Reserve post showed access and training were uneven. College graduates reported workplace artificial intelligence use at 58.7%, compared with 22.9% for workers without a college degree, and usage rose from 15.9% for workers earning under $50,000 to 66.3% for those earning over $200,000. (libertystreeteconomics.newyorkfed.org) That gap helps explain why the backlash is not a simple anti-artificial-intelligence story. Some workers say the tools help them finish drafts faster, while managers and coworkers say the time savings disappear when outputs arrive without context, sources or reliable facts. (libertystreeteconomics.newyorkfed.org; theguardian.com) The cost question is now showing up in corporate budgets as well as inboxes. The Information reported on April 15 that Uber’s chief technology officer, Praveen Neppalli Naga, said the company’s heavy use of coding assistants such as Claude Code had already exhausted its full-year artificial intelligence budget just months into 2026. (theinformation.com) The thread running through all of it is narrower than the hype cycle suggests. Artificial intelligence can speed up a person’s first draft, but companies still have to pay for review, correction, instability and usage bills when the draft is not good enough to ship. (cloud.google.com; theinformation.com; hbr.org)
