Executives raced to adopt generative AI. They promised boards higher output, leaner teams and a future less dependent on human workers. Investors cheered the announcements. Stock prices responded. Yet two years on, many of those same organizations confront something quieter and more corrosive.
Call it knowledge decay. Or workslop. The terms emerged from researchers watching what happens when low-quality AI output floods internal systems. Errors compound. Colleagues waste hours verifying facts that once passed without question. Trust in shared information erodes. And the very processes meant to accelerate work slow to a crawl.
Futurism first highlighted the pattern in stark terms this month. Companies that embraced AI wholeheartedly now watch outputs crumble. The piece drew heavily from Harvard Business Review analysis that laid out the mechanics in detail.
Harvard Business Review described the spiral in September 2025. “Errors compound and pile up,” its authors wrote. “Trust in information erodes. People spend more time verifying facts or risk costly and dangerous mistakes. Eventually, people start to lose trust in the processes that they rely on to do their jobs.” The article, titled “AI-Generated ‘Workslop’ Is Destroying Productivity,” rested on a survey of 1,150 U.S. employees. Forty percent reported receiving workslop in the previous month. On average, 15.4 percent of content crossing their desks qualified as such. The productivity tax reached $186 per employee each month. Scale that to a 10,000-person organization at 41 percent prevalence and annual losses exceeded $9 million.
Workers quoted in the study captured the frustration. One finance professional received AI-generated material so vague he faced a choice: rewrite it, force the sender to fix it, or accept mediocrity. “It is furthering the agenda of creating a mentally lazy, slow-thinking society,” the person said. A tech manager spent hours clarifying an unclear AI-drafted email. A retail director described repeated cycles of verification, meetings and personal rework. These stories repeat across departments.
But the damage runs deeper than wasted hours. Organizational knowledge itself atrophies. Employees lean on AI for first drafts and summaries. They forget how to perform certain analyses from scratch. Processes that once relied on human judgment incorporate generic outputs laced with subtle inaccuracies. Over months, institutional memory frays. Decisions rest on increasingly shaky foundations.
Failure statistics paint an even bleaker picture. More than 80 percent of AI projects fail to deliver intended business value, according to a RAND Corporation study from 2024. That rate runs roughly twice as high as for conventional IT initiatives. S&P Global Market Intelligence found 42 percent of companies abandoned most AI initiatives in 2025, up from 17 percent the year before. The average organization scrapped 46 percent of proofs of concept before production.
Fortune reported on MIT research in August 2025. The Project NANDA report concluded that 95 percent of generative AI pilots produced no measurable profit-and-loss impact. Only 5 percent drove rapid revenue acceleration. The study, built on 150 interviews, a survey of 350 employees and analysis of 300 public deployments, pointed to a “learning gap.” Tools and organizations failed to adapt AI to actual workflows. Model quality was rarely the culprit. Integration, governance and realistic expectations were.
“The 95% failure rate for enterprise AI solutions represents the clearest manifestation of the GenAI Divide,” the MIT report stated. Startups often succeeded by targeting specific pain points and partnering with vendors. Enterprises that built custom tools internally stumbled more frequently.
Real-world examples from 2025 drive the point home. Volkswagen’s Cariad software unit, tasked with advanced driver-assistance and AI systems, recorded $7.5 billion in operating losses over three years. Product delays stretched beyond a year for models including the Porsche Macan Electric and Audi Q6 E-Tron. A 20-million-line codebase accumulated bugs. The company cut 1,600 jobs. An insider told investigators he arrived without a clear job description and simply built what he knew. The lesson, according to analysts at Nine Two Three, was clear: big-bang modernization efforts rarely work. AI integration demands modular, iterative approaches.
Taco Bell’s AI drive-thru experiment turned into a public embarrassment. One customer reportedly ordered “18,000 cups of water,” crashing the system. The AI repeatedly pushed add-on drinks despite refusals, forcing employees to intervene. National media coverage followed. Brand damage lingered. Chief Digital Officer Dane Mathews acknowledged in August 2025, “Sometimes it lets me down, but sometimes it really surprises me.” The chain shifted to a hybrid model. The broader takeaway: customer-facing automation without strong guardrails creates friction rather than convenience.
Google’s AI Overviews feature produced hallucinations that spread quickly. It suggested adding non-toxic glue to pizza sauce to help cheese stick, drawing from an 11-year-old Reddit joke. Other responses invented meanings for nonsense phrases or recommended eating rocks for digestion. Trust in search results suffered. Persistent negative press followed. For a company whose product is information, accuracy remains the core asset. Verification cannot be an afterthought.
Hiring has grown complicated too. Harvard Business Review noted that AI-augmented recruiting sinks trust for both candidates and recruiters to all-time lows. Automated screening tools generate errors. Candidates submit AI-written materials that blur signals of genuine capability. Recruiters spend more time verifying than deciding. The entire pipeline frays.
Worker morale tells another story. Employees forced to adopt AI tools they distrust have pushed back. Some sabotage systems. Others simply refuse. Surveys show white-collar workers bypassing company AI even when mandated. A Futurism report on enterprise worker sentiment captured widespread disillusionment. Low morale compounds the productivity losses already measured in workslop taxes.
Data itself poses a stubborn barrier. Enterprises feed models on years of accumulated files, many redundant, obsolete or trivial. A startup called Clario raised $6 million in June 2026 to tackle exactly this “data rot” across tools like Google Drive, SharePoint and Confluence. The company routes cleanup decisions into Slack and Teams, training maintenance systems over time. Its emergence signals that AI budgets have moved from pilots to production demands. Yet the prerequisite work remains unglamorous and expensive.
Backlash extends beyond office walls. Axios reported in May 2026 that AI resentment could slow adoption and create investor risk. Protests target data centers. Executives face boos at events. Public sentiment surveys show growing skepticism. Computer Weekly described a “grassroots backlash” era in February 2026, with writers, artists and communities questioning deployment choices.
So what separates the minority that succeeds? Pertama Partners analyzed the data in early 2026. Clear success metrics defined before project approval matter. Honest assessment of data foundations comes first. Sustained executive sponsorship proves essential. Organizations that treat AI as business transformation rather than technology experiment fare better. Michael Hauge of Pertama Partners put it directly: failures stem from ambiguity about success, ownership and data readiness. “The projects that fail are rarely beaten by the technology. They are beaten by ambiguity.”
Harvard Business Review authors Kate Niederhoffer, Gabriella Rosen Kellerman and their colleagues offered practical guidance. Leaders must model discernment rather than blanket mandates. Distinguish “pilots” who use AI to enhance creativity and achieve goals from “passengers” who deploy it to avoid work. Set explicit guardrails. Recommit to collaboration that keeps humans accountable for outcomes. Uphold standards of excellence even when AI participates.
The early hype has faded. Billions poured into generative AI. Global spending forecasts reached hundreds of billions for 2025 alone. Yet measurable returns remain elusive for most. Companies now confront the hangover. Some quietly scale back. Others hire humans specifically to fix AI mistakes, a practice documented years ago and still common. A few invest in proprietary models trained on clean, internal data. Those efforts show more promise than public large language models churning generic prose.
The pattern feels familiar. Technology waves often crest with over-adoption followed by correction. This time the correction carries unique risks. When the product is knowledge itself, decay spreads silently. Decisions rest on polluted information. Skills atrophy. Culture shifts toward acceptance of mediocrity dressed in polished language.
Executives who treat AI as a universal productivity lever now face the bill. Those who view it as one tool among many, best deployed with clear boundaries and human oversight, stand a better chance. The data is in. The anecdotes accumulate. The rot is real. And the organizations that recognize it earliest may yet avoid becoming the cautionary tales of the next cycle.
But recognition alone falls short. Real change demands hard choices about which tasks truly benefit from AI and which suffer from it. It requires investment in data hygiene that delivers no immediate headline. It asks leaders to resist the temptation to declare victory after a successful pilot. Most of all, it insists that humans remain responsible for the quality of what their organizations produce and decide.
WebProNews is an iEntry Publication