AI Saves Workers 11 Hours a Week—Then Takes Back 6 in 'Botsitting'

Artificial intelligence is delivering measurable time savings to office workers, but those gains come bundled with a new category of labor that companies are only beginning to understand.

A survey of 6,000 digital workers across the United States, United Kingdom, and Australia found that AI tools save employees roughly 11 hours per week. The same workers, however, spend more than six hours weekly on what researchers call "botsitting"—verifying AI output, correcting mistakes, and rerunning prompts to get usable results.

The research, conducted between December and January by the Work AI Institute with contributors from Stanford University and UC Berkeley, reveals a productivity paradox: three-quarters of individual workers report efficiency gains, yet only 13 percent of organizations say they've seen significant business benefits from AI adoption.

Why it matters

The gap between individual productivity and organizational value signals a fundamental challenge for enterprise AI strategies. Companies investing heavily in AI tools may be trading one form of work for another rather than capturing genuine efficiency gains. Understanding the hidden labor costs of AI supervision becomes critical as businesses plan workforce changes and calculate return on investment.

The hidden cost of AI management

For every hour workers spend getting useful output from AI, they spend roughly another hour making that output usable, according to the survey. Of total AI interaction time, 37 percent goes to botsitting while 36 percent goes to actual productive work.

More than one-third of AI sessions fail outright, requiring workers to start over or substantially rework results. The survey analyzed anonymized workplace data from companies using the Glean Work AI platform.

"Most people don't realize the amount of time that they're spending working on the tools to get the time savings that they're professing," said Paul Leonardi, Duca Family professor of technology management at UC Santa Barbara and co-author of the study.

The report describes a "thick, mostly invisible layer of human labor holding the whole thing together."

New risks emerge

As workers delegate more tasks to AI, they're also offloading judgment and accountability. The survey found 41 percent of workers sometimes deliver AI-generated work they couldn't explain if questioned.

The research highlighted a case where a junior software engineer pasted thousands of lines of AI-generated code before going to bed. When something broke, a senior engineer already behind schedule had to debug the problem while the junior engineer struggled to explain the code.

"We're essentially expecting individual contributors to act as managers," Leonardi said. "They're just managing these AI tools, AI agents, and we're expecting that they'll be able to produce way more, but we're not taking into account all of the work that actually goes into managing."

The findings come as Silicon Valley companies push employees to maximize AI usage. Some efforts have backfired—Uber reportedly burned through its entire 2026 AI budget in four months without shipping a usable feature.

These details were first reported by the Los Angeles Times.