As the use of artificial intelligence spreads across companies worldwide, it is relieving workers of tedious old chores but creating new ones.

A new survey of individuals using AI found it made them more productive, saving each roughly 11 hours per week. But at the same time, the workers on average have to spend more than six hours “botsitting,” checking the AI output, fixing mistakes and rerunning the prompt.

“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.

Leonardi is one of the co-authors of the new study published by the Work AI Institute, whose contributors include academics from Stanford University and UC Berkeley. The institute is sponsored by AI company Glean. Leonardi said its research output maps broad trends in understanding AI’s impact on work.

The research surveyed 6,000 digital workers across the United States, the United Kingdom, and Australia between December and January.

The report found that we are in a phase of significant personal productivity gains, but few companies are translating these gains into revenue and business growth.

While 75% of individuals reported a boost in productivity, only 13% of the organizations say they have seen significant business gains as a result of AI adoption, the survey found.

The survey analyzed anonymized, aggregated workplace data from companies using the Glean Work AI platform, a private search tool used to manage their internal information.

Over the past six months, Silicon Valley companies have been urging their employees to max out their AI use . But the benefits of merely maximizing AI usage have been unclear, with instances such as Uber burning through their entire 2026 AI budget in four months, without shipping a usable feature.

The reason the boost in productivity sometimes leads to waste, Leonardi said, is the time people spend correcting the bot’s work and gathering the right files, documentation, and tacit knowledge required for it to produce high-quality output.

“It’s pretty striking the amount of time and effort people are spending,” Leonardi said.

Most employees now spend over six hours a week of their workday babysitting their work chatbots, the survey said.

There is a “thick, mostly invisible layer of human labor holding the whole thing together,” the report said.

The survey found that for every hour a worker spends getting useful output from AI, they spend roughly another hour making it usable.

Of the total time workers spend interacting with AI each week, 37% goes to botsitting, 36% to actually using the tool to produce work.

Part of the reason so much time disappears into botsitting is how often the tools fall short: Workers report that more than a third of AI sessions fail outright, requiring a full restart or substantial rework.

Paradoxically, as more workers hand over bigger parts of their jobs to AI, they are offloading personal judgment and responsibilities to the bots. The survey found 41% of workers say they sometimes deliver AI-generated work they couldn’t explain if asked.

The report highlighted an example of a junior software engineer, Robin, who pasted thousands of lines of AI-generated code before going to bed. But something in there was broken, which a senior engineer already behind on a deadline had to untangle, while Robin struggled to explain.

“I think what’s happening with a lot of these Gen AI tools right now is 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.”

This problem isn’t likely to go away.