AI saves 13 hours a week, then "bot sitting" quietly takes back half of them
Glean research finds workers spend 6.4 hours a week doing the hidden labor that keeps AI functional, nearly canceling out the productivity gains they report.
Glean researcher Rebecca Hinds has a term for the unpaid, untracked overhead of working with AI tools: “bot sitting.” It covers feeding context to systems that lack it, correcting errors, and manually bridging disconnected tools. The headline finding from her team’s research is that this hidden labor consumes 6.4 hours per week, “roughly half of all the time that AI supposedly saves,” against a reported average saving of 13 hours per week.
We're seeing about 36% of all AI sessions fail, meaning a worker goes to use the technology and it's not successful. Rebecca Hinds
The self-reported productivity numbers already contain some tension. Seventy-three percent of workers say AI makes them more productive, yet 36% of all AI sessions fail outright, requiring workers to start over or do significant rework. The gap between perceived and actual gains may also be shaped by what Hinds calls “bot bullshit”: 40 to 41% of employees, on her figures, ship AI work they could not explain if asked to defend it.
Among all the bot-sitting tasks, feeding AI context carries the highest exhaustion toll. Hinds frames it plainly: that effort is “in the best case something your AI should know,” making it feel redundant on top of draining. Only 13% of organizations, she says, have employees reporting significant productivity gains, which points to a sizable gap between the tools most workers use and the workflows the most effective organizations have built around them.