The Hours Nobody Is Counting
Ask a team whether AI saves them time and most will say yes. Ask where that time went and the answers become vaguer.
Glean's Work AI Index 2026, a survey of 6,000 full-time digital workers in the United States, the United Kingdom and Australia, puts numbers on the gap. Workers estimate that AI automation saves them around 11 hours a week, and 75% say it makes them more productive. Yet only 13% say their organisation is performing significantly better as a result.
The report gives a name to part of the reason: "botsitting", which it describes as the "largely unrecognized, unbudgeted, and untracked labor of making AI usable". On average, workers spend 6.4 hours a week on it: feeding AI context (2.3 hours), supervising its output (2.2 hours) and debugging its mistakes (1.7 hours). The report estimates that for every hour spent getting a useful output from AI, people spend roughly another hour making it usable. In the past month, 77% of workers had corrected or redone AI-assisted work.
Official data points the same way. The US Census Bureau's Household Trends and Outlook Pulse Survey (published August 2026) found that among workers who used AI at work in the previous week, 56% saved two hours or less, 10% saved no time and 4% said AI added time.
AI is saving time. A good share of it is being spent again on the same task.
The Organisational Shift: The Problem Is the Workflow, Not the Tool
Not all botsitting is wasted. Checking a client proposal or a financial figure before it goes out is exactly what a responsible person should do. The report separates this productive checking from the unproductive kind: re-typing the same background into several tools, trying a second tool because the first answer was poor, and cleaning up AI-generated work after it has already been passed to a colleague.
The unproductive kind is a workflow problem. It appears when every person uses AI their own way, with no shared starting point, no agreed standard of "good enough" and no rule about who checks what.
The report's own comparison is telling. At organisations reporting a transformative impact from AI, 90% of workers say their employer treats AI as a chance to redesign work, compared with 54% elsewhere. Those organisations also spend less of their AI time inside AI tools: 27%, against 49% at organisations reporting no or negative impact.
In other words, the teams getting the most from AI are not the ones spending the longest with it. They are the ones that changed how the work flows around it.
The Strategic Blueprint: Four Ways to Cut Botsitting
1. Write the context once and share it. Feeding AI context is the largest single block of botsitting time. Most of it is repeated: who the customer is, what tone the business uses, what the product does, what format the report takes. Write a short context note for each recurring task, keep it somewhere everyone can reach and paste it in at the start. A one-page note can save every person on the team from typing the same background each day.
2. Agree what "good enough" looks like before AI starts. A large share of rework happens because nobody defined the target. For each task you use AI for, write down three or four checks the finished piece must pass: the facts are verified, the numbers match the source, the tone fits the reader, the length is right. People stop re-prompting endlessly when they know when to stop.
3. Make the person who uses AI responsible for the output. The most costly botsitting happens downstream, when a colleague receives AI-drafted work that has not been checked. A simple rule fixes most of it: whoever uses AI owns what it produces, and checks it against the agreed standard before passing it on. This keeps the accountability with a person, which is where the report says it belongs.
4. Measure the finished result, not the first draft. If the team only counts how fast drafts appear, AI will look excellent and the rework will stay hidden. Pick one recurring task and track the whole cycle, from start to the moment the work is accepted. Ask how many rounds of correction it needed. That figure tells you whether AI is helping on that task, and whether the context note and checks are working.
The Human Factor: Judgment Is the Part You Cannot Hand Over
My view is that AI enhances human strategy, critical thinking and judgment. It does not replace them. Botsitting is what happens when that division of labour is left unclear.
When people treat AI output as finished work, colleagues spend their time fixing it. When people treat AI as a starting point that they direct, check and shape, the same tools save time that stays saved. The difference is not technical skill. It is a clear sense of what the person is there to decide.
In my work with teams, the most effective habit I see is simple: before using AI on a task, people ask themselves what part of the task needs their judgment, and keep that part for themselves. The report finds something similar. Its highest achievers use AI less on the core of their craft, not more.
Conclusion: Where Did Your Saved Hours Go?
If your team saves several hours a week with AI and spends a similar amount checking, correcting and re-explaining, the business has gained very little. The good news is that most of that lost time comes from habits a manager can change in a week: a shared context note, an agreed standard, a clear owner and an honest measure.
So here is a question worth putting to your team this week: on the task where you use AI most, how much of the time saved is being spent again on fixing the output?
If this was useful, share it with a colleague who manages a team's day-to-day work.
Further reading
For a practical look at how to map a recurring task and decide where AI fits into it, see AI Workflow Design & Process Digitisation. To see where your organisation stands on AI use today, the free AI Readiness Assessment is a useful starting point.
Build this skill in your team. Our AI Workflow Design & Process Digitisation course covers it hands-on, and it is HRDC (HRD Corp) claimable for Malaysian employers. See how to claim it with HRDC.
Sources
Glean Work AI Institute, Work AI Index 2026 (survey of 6,000 full-time digital workers in the US, UK and Australia, conducted December 2025 to January 2026; published June 2026): https://www.glean.com/work-ai-institute/reports/work-ai-index
US Census Bureau, "About a Third of Workers Who Used AI in the Last Week Said They Completed Tasks One to Two Hours Faster", Household Trends and Outlook Pulse Survey, 11 August 2026: https://www.census.gov/library/stories/2026/08/ai-use-at-work.html