AI Productivity Claims Clash With $9M Annual Cleanup Costs

Corporate America is pouring billions into artificial intelligence with the promise of unprecedented productivity gains, yet emerging data suggests the returns may be largely imaginary. A troubling disconnect has surfaced between executive claims of AI-driven efficiency and what employees actually experience on the ground.

Researchers at Stanford and BetterUp have identified a phenomenon they call "workslop"—AI-generated content that appears polished but lacks substance. According to their findings, roughly 40% of workers regularly receive slide decks without insight, reports without analysis, and summaries that miss critical points. Rather than saving time, this output creates additional work: someone still needs to review, correct, and rebuild what the AI produced.

The cleanup costs are substantial. Each workslop incident requires an average of two hours to resolve. For a 10,000-employee organization, that translates to approximately $9 million annually in lost productivity—a figure that sits uncomfortably alongside the billions being invested to generate efficiency gains in the first place.

The measurement gap

KPMG's 2026 Global AI Pulse survey found that only 7% of business leaders report established ROI from AI investments, even as adoption has become nearly universal. Meanwhile, 77% of workers told the Upwork Research Institute that AI has increased their workload rather than reduced it, according to reporting by Forbes contributor Kara Dennison.

The perception problem runs deeper than self-reporting bias. Research from METR found that experienced developers using AI tools completed tasks 19% slower than those working without AI assistance—yet believed they were working 20% faster. That 40-percentage-point gap between perceived and actual productivity raises serious questions about the self-reported gains appearing in corporate surveys.

Why it matters

Companies are making irreversible workforce decisions based on productivity improvements they cannot verify. AI-cited layoffs account for 24% of all announced job cuts in 2026, according to Challenger, Gray & Christmas. Yet Forrester research shows 55% of employers now regret AI-related layoffs, with over 30% of U.S. hiring managers reinstating eliminated positions at higher cost than the original savings.

The investment pressure comes from the top. A Harris Poll survey found 80% of CEOs worldwide believe their jobs are at risk if AI strategy fails by year's end, while 72% face active board pressure to demonstrate returns. KPMG research revealed that 78% of business leaders cite demonstrating AI value to investors as a critical strategic driver—not actual results. Goldman Sachs projects $1 trillion in global AI investment for 2026, including $581 billion in the United States alone.

What works instead

Organizations seeing genuine returns share common approaches, Dennison reports. They identify specific use cases where AI demonstrably improves outcomes before broad deployment. They invest in structured training so workers understand how to apply tools effectively to their actual tasks. They establish performance baselines before implementation, then measure real output changes rather than activity metrics like tool usage rates.

Crucially, successful adopters distinguish between AI activity and AI outcomes. Adoption statistics and prompt volumes are not productivity metrics—they measure motion, not progress. What matters is whether work gets completed faster, with fewer errors, and at lower cost than before AI deployment.

An Emergn survey of 700 senior leaders found that only 30% considered ending an underperforming AI initiative a normal, acceptable decision. Nearly a quarter admitted reluctance to acknowledge project failures, allowing unsuccessful initiatives to consume an average of 2.4% of annual revenue.

The details were first reported by Kara Dennison in Forbes, drawing on research from Stanford, BetterUp, KPMG, Upwork Research Institute, METR, Forrester, Harris Poll, and Challenger, Gray & Christmas.