Babysitting the Machine: Glean's Rebecca Hinds on the Hidden Human Labor of AI at Work

Episode Summary

Executive Summary: Rebecca Hines discusses Glean’s Work AI Index 2026, showing that AI is widely adopted by knowledge workers but often fails to improve organizations because of bot sitting (manual cleanup and context-giving) and bot shitting (shipping AI output people can’t defend). The conversation explores why context, mission, incentives, and psychological safety are central to effective enterprise AI.

Main Topics: The Work AI Index findings (Priority: 5/5): The report’s headline result is a large gap between individual and organizational impact: most workers use AI and feel more productive, but few say their company is materially better because of it. Bot sitting as hidden labor (Priority: 5/5): Hines defines bot sitting as the untracked human work of feeding AI context, debugging outputs, and cleaning up mistakes; it consumes a substantial share of the time AI saves. Bot shitting and coordination neglect (Priority: 5/5): Bot shitting refers to passing along AI-generated work that the user cannot explain or defend, often driven by exhaustion, good-enough behavior, and broken cross-tool coordination. Context, enterprise graphs, and AI systems design (Priority: 4/5): A major solution theme is better context-aware systems that connect documents, people, workflows, recency, and authority so AI can be useful without excessive human babysitting. Meaning, alienation, and work design (Priority: 4/5): The discussion examines why workers threatened by AI may adopt it most aggressively, including the risk of automating away the parts of work that give people meaning and pride. Leadership, culture, and incentives (Priority: 5/5): Hines argues that AI transformation requires transparency, psychological safety, mission alignment, and reward systems that value collaboration and effective use rather than raw tool usage. Future org design and the enterprise graph (Priority: 4/5): The conversation looks ahead to smaller teams, more dynamic staffing, and AI-assisted role allocation based on missions, tasks, skills, and employee ambitions.

Key Arguments: Individual AI productivity gains are not translating into organizational gains because time saved is being consumed by bot sitting, hidden rework, and coordination overhead. The biggest bottleneck is not just model capability but missing organizational context: AI tools often lack authority, recency, and cross-system integration. Bot sitting is a leading indicator of both dissatisfaction and future disengagement because it is exhausting, unrewarded, and often signals poor organizational AI strategy. Bot shitting is partly a symptom of exhaustion and alienation: people reach for “good enough” and then ship output they cannot fully explain. Organizations that succeed with AI tend to combine top-down leadership with bottom-up champions, and they reward collaboration and experimentation rather than only individual throughput. Meaning matters: if AI automates the parts of work employees value most, it can undermine ownership, pride, and performance even if efficiency rises. AI should be treated as a teammate, not a peer or a human scapegoat; leaders must preserve human accountability while using AI to augment work. The future of enterprise AI is less about one universal model and more about a context-rich orchestration layer that routes tasks to the right model/tool for the job. Mission clarity becomes more important as hierarchy flattens, because employees need a shared compass when managers and org charts provide less guidance. Effective AI deployment may shrink teams and rebundle roles, but successful organizations will intentionally redesign work instead of copy-pasting AI-native practices onto legacy companies.

Data Points: AI usage among surveyed workers: 87% - Share of digital/knowledge workers who now use AI Workers reporting higher productivity: 73% - Share saying AI makes them more productive Average weekly time saved: 13 hours/week - Average self-reported time saved from AI use Share saying organization performs significantly better: 13% - Only a small minority see major organizational improvement from AI Bot sitting time: 6.4 hours/week - Average time spent feeding context, debugging, and cleaning up AI outputs AI sessions that fail: 36% - Share of AI sessions that do not successfully move work forward and require rework or restart Workers admitting to bot shitting: 69% - Share who say they ship AI-generated work they cannot explain or defend Employees shipping explain-unsafe AI work: 40-41% - Another reported measure of AI work that workers could not explain if asked Knowledge workers surveyed: 6,000 - Total survey sample across the US, UK, and Australia US respondents: 3,000 - Portion of survey sample from the United States UK respondents: 1,500 - Portion of survey sample from the United Kingdom Australia respondents: 1,500 - Portion of survey sample from Australia Automation time savings estimate: 11 hours/week - Time savings attributed to work output fully automated with AI in one framing Organizations using AI in performance-related decisions: Large portion / majority trend - Employees report AI is already informing performance management, hiring, and firing in many workplaces

Pivotal Quotes: "This is a human change, just as it is a technology change." — Rebecca Hines: On why AI adoption is mostly a change-management and psychology challenge, not just a technical one "Bot sitting is the hidden human labor that is required to make the technology usable." — Rebecca Hines: Definition of the report’s key term for manual context-giving, debugging, and cleanup "The digital employee experience is increasingly the employee experience." — Rebecca Hines: Explaining why technology friction now directly shapes engagement, retention, and organizational outcomes

Implications: Enterprise AI wins when organizations invest in context, mission, and incentives—not just tools. Leaders who ignore hidden labor and psychological costs risk more slop, more turnover, and weaker ROI; those who redesign work around meaningful human-AI collaboration may gain durable advantage.