Botsitting Drains 6.4 Hours Weekly as Firms Ignore Integration Costs

Glean's Work AI Institute report, drawing on 6,000 surveyed workers across the US, UK and Australia from December 2025 to January 2026, documents an average 6.4 hours spent weekly on context-feeding, output verification and error correction, a figure that aligns with Stanford and UC Berkeley co-authors' measurements of untracked AI overhead. This matches patterns in McKinsey's 2025 Global Survey on AI, which recorded similar productivity leakage when organizations deployed models without workflow redesign. The 73% elevated job-search likelihood among high-botsitting employees echoes findings from a 2024 Berkeley Labor Center study on task fragmentation in knowledge work. White-collar users report 87% adoption yet only 13% see firm-level gains, reproducing the productivity paradox first quantified in Brynjolfsson et al.'s 2023 NBER paper on AI deployment lags. Companies that allocate more time to context-setting and standards rather than raw usage show faster performance lifts, per the Glean data, while those treating AI as plug-and-play continue exporting supervision labor onto individuals. The resulting frustration concentrates in roles where relational tasks are automated first, amplifying exit signals already visible in LinkedIn's 2025 workforce trends report on AI-related burnout.