You Don’t Know It Yet, But ‘Botsitting’ Is Stealing 6 Hours of Your Life Each Week

Artificial intelligence has promised to revolutionize work processes, lightening the load on employees and freeing up precious time. However, new research from the Work AI Institute at Glean, conducted with the help of prestigious universities like Notre Dame, Stanford, and UC Berkeley, has revealed a very different reality. So-called “white-collar” workers spend an average of 6.4 hours a week on “botsitting”—that is, supervising AI systems.

The term “botsitting,” coined by the authors of the report (source: Business Insider), describes the often-overlooked and hidden work necessary to make AI genuinely useful. This activity includes providing context for models, verifying generated outputs, resolving errors, and cleaning data. The research involved 6,000 full-time workers in the United States, the United Kingdom, and Australia between December 2025 and January 2026.

What is Botsitting: Impact on Morale and Risk of Turnover

The results highlight a marked gap between individual productivity gains and overall company performance, a phenomenon dubbed the “productivity paradox.” Although 87% of employees surveyed reported using AI at work and 75% feel individually more productive, only 13% perceived a significant improvement in company performance thanks to the implementation of these technologies. Rebecca Hinds, head of the Work AI Institute at Glean, described botsitting as a job that is “often tedious,” “exhausting,” unrecognized, underappreciated, and not even tracked or incentivized within organizations.

This additional burden is having repercussions on employee morale. The report found that workers dedicating a disproportionate amount of their AI time to botsitting are 73% more likely to actively seek new employment. The survey highlighted that frustration extends beyond the extra work: many employees spend time transferring information between disconnected AI systems, correcting errors, and providing context that the tools should already inherently possess. They are effectively acting as a bridge between technologies that do not effectively communicate with each other.

Hinds also pointed out that, in some cases, workers are asked to automate the parts of their jobs that they find most fulfilling. She cited the example of customer service employees who value building human relationships but are increasingly called upon to supervise AI agents. This, she warned, can erode the sense of joy and meaning in work, posing a serious risk to staff retention.

To break this cycle, the solution does not simply lie in implementing more AI, according to the study. Organizations benefiting the most are those that invest the most in “work around AI”: helping employees access the right context, teaching them how to effectively use technology, and establishing clear standards to define the quality of AI-assisted work. These companies do not increase the amount of time spent directly with AI but improve the quality by setting context, defining what constitutes a good outcome, sharpening human judgment, and deciding what should never be delegated to a model. The alternative, the authors of the report warned, is to continue to pay the price in terms of “botsitting” and the constant turnover of talent, tired of having to constantly “clean up” after the bots.