Study finds workers spend 6.4 hours botsitting AI weekly

Glean's Work AI Institute published its Work AI Index 2026 based on a survey of 6,000 full-time digital workers in the US, UK, and Australia (December 2025 - January 2026). The report was co-authored by researchers from the Work AI Institute, Emory University, Stanford University, UC Berkeley, UC Santa Barbara, UNC Charlotte, University College London, and University of Notre Dame. Key findings reported by Computerworld include: 87% of digital workers use AI; AI is automating more than a quarter of digital work and saves roughly 11 hours per week per worker on average; yet only 13% say the use of AI has significantly improved their company's performance.

The report coins the term 'botsitting' for the unrecognized work required to make AI usable: feeding large language models missing enterprise context, checking and correcting outputs, debugging mistakes, re-running prompts, and switching between disconnected AI tools. Workers average 6.4 hours per week on this, a figure that erodes the headline productivity gains. Rebecca Hinds, head of Glean's Work AI Institute, told Computerworld: 'It's definitely in many ways a vicious cycle that feeds itself.' She noted that LLMs trained on the public internet - but not enterprise-specific data - regularly require workers to re-supply company context, creating repetitive overhead across multiple tools.

A companion concept introduced in the report, 'botshitting', refers to shipping AI-generated work that hasn't been verified - because workers are overwhelmed or time-constrained. The report found 69% of users admit to doing this; 41% say they sometimes deliver work they could not explain if asked; and 28% blame AI for mistakes they themselves made. Hinds told Computerworld: 'Botshitting is offloading your critical human thinking, judgment, and understanding.' Workers using multiple AI agents are more prone to this, she said, because agentic systems can spiral without proper controls, leading users to abandon verification.

The productivity paradox the report identifies is: workers report large individual time savings, yet organizational metrics barely move. Only 13% say company performance has improved significantly. The report attributes high-performing organizations' success to redesigning work processes, formally rewarding AI skills, providing governance with continuous policy updates, and measuring quality and efficiency via existing KPIs rather than token usage.

vendor and platform features aimed at reducing manual oversight (automated validation, provenance tooling, connected enterprise knowledge layers); organizational metrics that capture end-to-end workflow impact rather than per-worker time savings; and whether the botsitting burden is transitional (as tools improve) or structural (as agentic complexity grows). The academic co-authorship and large sample size give these findings more weight than typical vendor surveys, though Glean - as an enterprise AI search provider - has a commercial interest in findings that highlight the cost of disconnected tools.