Bositting: The Work Draining AI Gains

Botsitting: The Hidden Labor Tax on AI at Work

A new report from Glean and the Work AI Institute puts a name to something that's been hiding in plain sight: the invisible labor of making AI actually work. 87% of digital workers now use AI, 75% say it makes them more productive, and the average worker saves 11 hours a week through automation — yet only 13% say their organization is performing significantly better. The report's argument is that those gains are being swallowed by a new form of labor they call "bot sitting." The counterargument — and the more interesting frame — is that even without bot sitting, individual productivity gains don't automatically become organizational gains. But bot sitting is still a real and growing problem worth understanding in its own right.

Glean Work AI Index 2026: The Hidden Labor of AI at Work

What Is Bot Sitting?
37% of AI time — more than active use of AI itself.
The report defines bot sitting as the work required to make AI usable: feeding it missing context, checking outputs, debugging mistakes, rerunning prompts, and cleaning up confident-but-wrong answers. Workers now burn an average of 6.4 hours a week on it. Broken down: 2.3 hours feeding AI context (14% of total AI time), 2.2 hours supervising outputs, and 1.7 hours debugging — with the remainder on cleanup and tool-switching. Crucially, they break bot sitting into productive flavors (verifying high-stakes outputs, iterating prompts, adding domain context the AI couldn't know) and unproductive ones (reloading the same context into multiple tools, comparing outputs because the first wasn't good enough). The former is just what it looks like to manage autonomous AI well. The latter is where things go wrong.

The Exhaustion Multiplier and Bot Sh*tting
The burnout-to-sloppiness pipeline is the real threat.
For every 10% more time workers spend feeding AI context, they're 25% more likely to feel worn out by it. Frequent bot sitters — those spending 40% or more of their AI time on bot sitting — are 73% more likely to be actively job hunting. And the fatigue doesn't just cause churn; it causes what the report calls "bot sh*tting": the gradual cognitive offloading that happens when workers stop verifying outputs and start shipping whatever looks good enough. When AI-generated work fails, 40% of workers blame the AI. Only 29% admit fault. Heavy AI users are 3.4 times more likely than light users to blame the tool. Tool sprawl makes all of this worse — workers using multiple AI tools are 35% more likely to be frequent bot sitters, and right now, 60% of workers are rerunning the same prompt across multiple tools because the first output wasn't good enough.

The Agentic Amplification Problem
This study was fielded in December–January. Agents weren't really in the picture yet.
The 6,000 workers surveyed answered questions in December 2025 and January 2026 — before the real ascendancy of agentic work. One of the report's most provocative data points is that the smarter the tool, the sloppier the worker. Among ChatGPT, Claude, Gemini, and Microsoft Copilot, the tools whose users reported the biggest productivity gains (ChatGPT at 67%, Claude at 59%) were also the tools whose users admitted to bot sh*tting most frequently (71% and 92% admitting to it at least monthly, respectively). As AI moves from efficiency tasks to genuinely new capabilities workers couldn't previously perform, the verification problem becomes structurally harder — you can't fact-check an output in a domain you've never worked in.

What High AI Achievers Do Differently
Better individual AI users still do more of their core work themselves — for now.
The report identifies "high AI achievers" as workers who report AI improving both productivity and quality of work. Three things distinguish them: they're more selective about where they deploy AI (spending closer to a third of their AI time on core job tasks vs. half for low achievers); they orient their bot sitting toward the productive variety, using it as a feedback mechanism to improve their own AI skills; and they reinvest their AI time dividend into learning new skills rather than absorbing more work. The caveat: this framing may already be outdated. When the engineers who built Claude Code essentially no longer write code themselves, the entire concept of "AI time spent" starts to break down in an agentic world.

How High-Achieving Teams Work
Peer-to-peer adoption is 5x more powerful than top-down mandates.
High-achieving AI teams treat AI as a teammate but keep accountability on the humans. When AI underperforms, they re-prompt or try other tools rather than abandoning AI and doing it themselves. They see adoption spreading peer-to-peer rather than top-down — and the numbers on this are striking: when a leader uses AI, it makes the average employee 2.4 times more likely to adopt it; when a direct teammate uses it, 3.2 times more likely; when a cross-functional teammate adopts it, 5 times more likely. The reason cross-functional teammates carry so much weight is that they build for the messy version of work, not a tidy fantasy of it — their workflows survive contact with reality. High-achieving managers also delegate 32% more of their coordination work to AI, freeing up time for coaching, developing, and inspiring their people rather than routing requests and summarizing meetings.

What Transformative Organizations Do Differently
The 13% doing this right share three things: better metrics, living governance, and investing in people.
Transformative organizations — those where employees say AI has meaningfully improved organizational performance — share a few distinguishing habits. They measure things that matter (quality, productivity, output, time saved) rather than vanity metrics, and 71% of workers in transformative organizations can see their own AI usage data versus 40% in non-transformative ones. They treat governance as a living system: 93% of their workers say their org reviews AI policy, versus 55% in non-transformative organizations. And they invest in people: 84% of workers in transformative organizations say their company formally rewards AI skills, versus 48% in non-transformative ones. Trust follows from all of this — 93% of workers in transformative organizations trust their company's AI strategy, compared to 57% elsewhere.