Survey: Knowledge Workers Spend More Time Managing AI Than Producing Output
A survey of 6,000 digital workers by Glean's Work AI Institute found that three-quarters of knowledge workers believe AI boosts personal productivity, but only 13% say technology has significantly improved company performance. Automation saves an average of about 11 hours per week, but employees spend slightly more than half of that interaction time managing tools rather than producing output. The report notes that successful enterprise AI adoption relies on a 'human infrastructure' built around AI, including context implementation, use-case training, shadow AI signal identification, and governance mechanisms.
Briefing Overview
- Although three-quarters of knowledge workers say AI has boosted their productivity, only 13% report that the technology has significantly improved their company's performance—a finding from a survey by Glean's Work AI Institute. Glean surveyed 6,000 digital workers and combined insights from AI leaders on its platform to compile the firstWork AI Indexreport.
- The survey found that automation saves employees an average of about 11 hours per week, but employees report spending much of that saved time on managing AI. Of the total time spent interacting with AI, employees say they spend slightly more time managing the tools than producing work. Only 27% of the time is spent learning how to use the tools and building agents.
- The report found that the success of enterprise AI applications depends on the "human infrastructure" built around them. To improve use cases, leaders must anchor AI in business scenarios, train employees on use cases, and treatshadow AIas a signal that company-approved tools are insufficient, and embed governance into daily decision-making.
Deep Insights
Although AI is removing tasks previously done by humans from employees' to-do lists, employees are now spending time on low-visibility tasks to make AI outputs usable—such as providing context for agents, checking their work, flagging errors, and cleaning up answers.
The report found that employees spend nearly six and a half hours per week on these maintenance tasks. And monotonous work can easily turn into errors. If employees stop carefully reviewing outputs or verifying whether AI suggestions are reasonable—as 69% of respondents reported—errors slip through the cracks.
"Too many companies treat AI adoption as a vanity metric—more seats, more prompts, more usage," said Rebecca Hinds, head of Glean's Work AI Institute, in the report.
But she noted that more AI usage does not equal productivity or digital transformation. The report found that for every hour employees spend getting useful output from AI, they need another hour to make it usable. Glean found that more than a third of AI sessions fail completely, forcing employees to start over or do significant rework.
Hinds said that if employees spend too much time on AI management, companies are not eliminating work—they are just creating new types of work and adding more overhead for employees and managers.
Hinds believes that companies that integrate AI into actual workflows will be more successful than those that use AI just for the sake of using it. Successful companies build morehuman infrastructurearound their AI applications, training employees on how and when to use AI, and establishing clear guardrails.
The report also noted that these companies learn to reinvest the time saved by AI into higher-quality, human-centered work and stronger AI skills, rather than using AI as much as possible.
Enterprise-level success is more likely to be built on foundations at the individual, team, and organizational levels.
"Anchored in the right context, measured by real outcomes, and governed in a way that helps employees act faster without lowering the quality bar," Hinds said.