Study Confirms AI Is Increasing Employee Workloads

The Study's Findings

Researchers surveyed and tracked work patterns of over 4,000 knowledge workers at companies that had recently deployed AI productivity tools. They measured actual time spent on tasks, job satisfaction, stress levels, and perceived productivity before and after AI adoption.

On average, AI tools reduced time spent on targeted tasks by roughly 30 percent. However, total working hours for employees using the tools increased by an average of 12 percent. The discrepancy is explained by several categories of new work.

First, there is direct overhead of managing AI tools: formulating prompts, evaluating outputs, iterating on unsatisfactory results, and correcting errors. This AI management work did not exist before the tools were deployed.

Second, AI tools enabled managers to raise expectations about output volume and speed. When a team demonstrated that AI could help produce more reports, the expectation quickly became the new normal, without adjustment to headcount or acknowledgment that maintaining quality required additional human effort.

The Expectations Ratchet

This expectations ratchet emerged as the study's most concerning finding. Across organizations, AI tool introduction was followed by increased output targets, expanded scope of responsibilities, or reduced staffing. The productivity gains that AI provided were captured by the organization through higher expectations rather than returned to employees through reduced workloads.

The dynamic mirrors historical patterns with previous automation waves. Email, spreadsheets, and enterprise software all promised to reduce work burdens but instead expanded the volume and speed of work expected from each individual. AI appears to follow the same trajectory, but with an added twist: because AI tools are positioned as transformative, the expectations inflation is correspondingly larger.

Quality and Satisfaction Impacts

Workers using AI tools reported lower confidence in the quality of their output, even when objective measures suggested quality was maintained. Job satisfaction declined modestly but consistently, driven by the sense that expertise was being devalued. Professionals reported feeling reduced to AI managers, reviewing machine outputs rather than applying their own knowledge directly.

Stress levels increased across the board, with the highest increases among mid-career professionals who felt pressure to demonstrate AI proficiency while maintaining quality standards established over years of practice.

What Organizations Can Do

The researchers recommend that organizations commit to returning at least a portion of AI-generated time savings to employees rather than immediately filling freed-up time with additional work. Setting realistic expectations about what AI can and cannot do well is critical.

Organizations with more modest and specific deployment strategies, targeting particular bottlenecks rather than mandating broad adoption, achieved better outcomes for both productivity and employee satisfaction. Tracking not just the tasks AI automates but the new work it creates is essential for drawing accurate conclusions about actual impact.

The findings arrive as companies across every industry race to integrate AI tools, often driven by competitive pressure rather than careful analysis. The study suggests the rush to adopt AI may be creating problems that will take years to fully understand. For workers caught in the middle, the promise of AI as labor-saving technology remains largely unfulfilled.

This article is based on reporting by Gizmodo. Read the original article.