Making a task easier does not necessarily reduce the work around it. | Image credit: Tanía Diaz / Ordinal
Main takeaways
- Faster tasks do not necessarily mean less work.
- Technology can change expectations as well as productivity.
- Research shows AI can both save time and intensify work.
- New tasks can emerge around checking, correcting and coordinating AI outputs.
- Organisations influence whether efficiency becomes capacity or additional workload.
The vacuum cleaner made housework easier without necessarily reducing the work involved. AI raises a similar question for businesses: what happens when technology makes it possible to do more?
The vacuum cleaner made cleaning easier. It reduced the physical effort involved in removing dust and dirt, particularly from carpets and upholstery.
Yet easier cleaning did not necessarily produce more leisure time.
The Vacuum Cleaner Problem
In More Work for Mother, historian Ruth Schwartz Cowan examined a paradox of household technology. Appliances reduced the effort involved in individual domestic tasks, but the organisation of household labour changed alongside them. Standards of cleanliness evolved, some activities became more frequent, and responsibilities shifted.
A technology designed to reduce labour could therefore contribute to a household in which people continued working just as hard.
The vacuum cleaner offers a useful way to think about AI in business.
A tool might reduce the time required to write a report, analyse information or prepare a customer response. But whether the organisation ends up doing less work depends on what happens around that improvement.
Technology Does Not Decide How We Work
The philosophical distinction here is between technological determinism and the social shaping of technology.
Technological determinism treats technology as a principal driver of change, often implying that its capabilities lead towards predictable social or organisational outcomes.
Research on the social shaping of technology offers a different perspective. Technologies are developed, adopted and used within existing institutions, practices and relationships. Their effects depend partly on the decisions people make about them.
The vacuum cleaner did not decide how frequently a house should be cleaned. Its capabilities interacted with changing expectations about cleanliness and domestic responsibility.
AI creates a similar choice.
If a task that previously took three hours can now be completed in thirty minutes, a business might use the recovered time to reduce pressure on employees, improve the quality of its work or increase output.
It might also introduce more reporting, more reviews and higher production targets.
The technology enables these possibilities. It does not independently decide which one the organisation should pursue.
When AI Makes the Working Day Longer
Evidence from contemporary workplaces suggests that AI can increase the amount of work people undertake, even when individual tasks become easier.
In an eight-month study at a US technology company, researchers Xingqi Maggie Ye and Aruna Ranganathan examined how AI affected employees’ working patterns.
They observed workers taking on tasks they previously would not have attempted, moving between multiple AI-assisted activities and allowing work to extend into breaks and other parts of the day.
Some employees expanded their responsibilities voluntarily because AI made additional work seem manageable. But this created a further problem: as employees demonstrated that they could produce more, greater output could become the expectation.
The researchers described a pattern of work intensification rather than straightforward labour reduction.
Their findings come from one organisation and should not be generalised to every business. Nevertheless, they identify a mechanism that leaders should recognise: when the effort required to begin a task falls, people may take on more tasks.
Consider a marketing team using AI to develop campaign materials.
A draft that once took three hours can now be produced in thirty minutes. The team begins generating more variations, testing additional messages and preparing campaigns more frequently.
Someone must still check the outputs for accuracy, relevance and consistency. Managers may also expect shorter turnaround times.
The original task becomes faster, but the surrounding workload expands.
The same connection has been drawn outside academic research. Writing on X, the account @Kpaxs reached Cowan’s book by a different route:
AI cut the cost of starting a piece of work to almost zero. A first draft takes thirty seconds but finishing costs about what it always did. Someone still has to check the output, decide whether it’s right, deal with the edge cases, get other people to agree and live with the consequences.
When starting is cheap and finishing is expensive, you get more unfinished work.
AI Can Save Time Without Expanding Work
The opposite outcome is also possible.
In a randomised field experiment involving 7,137 knowledge workers across 66 firms, researchers studied how access to generative AI changed working patterns.
Employees using AI tools integrated into their everyday applications spent approximately two fewer hours per week on email and reduced work outside regular hours.
The researchers did not detect broader changes in the quantity or composition of tasks from individual access to the technology.
This finding is important because it demonstrates that AI can deliver measurable time savings without necessarily expanding employees’ responsibilities.
The two studies examined different environments and used different methods. They cannot establish that one management approach caused the difference.
But together they challenge the assumption that a technology’s capabilities are enough to predict its organisational consequences.
AI can reduce the effort involved in work. It can also make additional work easier to initiate. What follows depends on how people respond and how the organisation manages those changes.
What Happens to the Time Saved?
For CEOs and COOs, the question is therefore broader than how much faster an AI tool performs an individual task.
They need to understand whether the technology improves the operation as a whole.
A customer service team may use AI to draft responses faster, only to spend additional time correcting inaccurate answers. A finance team may generate reports more frequently without improving the decisions those reports support.
Some additional oversight is necessary. Accuracy, accountability and risk management cannot simply disappear because a task has been automated.
But businesses should distinguish between work that creates value and work that exists primarily because a new tool has introduced complexity.
They should also examine whether increased output is useful. Producing twice as many reports offers little benefit if nobody needs the additional information.
Before introducing AI, organisations can establish how much work a process requires, what the technology is expected to remove and what new responsibilities its use might introduce. After implementation, they can examine the entire workflow rather than relying exclusively on time saved at task level.
That assessment should include the people reviewing outputs, managing exceptions and responding to changes in expectations.
Who Decides What Productivity Means?
The vacuum cleaner was a valuable invention. Its limitations as a labour-saving device were not simply failures of engineering. They reflected the relationship between technological capability and the expectations surrounding its use.
AI presents businesses with a similar question.
If employees can complete their work faster, leaders still have to decide whether the benefit should appear as greater output, better quality, additional capacity or less pressure on the workforce.
Those decisions determine whether an efficiency improvement becomes an organisational improvement.
The vacuum cleaner made cleaning easier. It did not decide how clean a house needed to be.
AI can make work faster. Businesses still have to decide how much work is worth doing.
References
- Cowan, R. S. (1983). More Work for Mother: The Ironies of Household Technology from the Open Hearth to the Microwave. Basic Books. https://www.hachettebookgroup.com/titles/ruth-schwartz-cowan/more-work-for-mother/9780465047321/
- Dillon, E. W., Jaffe, S., Immorlica, N., & Stanton, C. T. (2025). Shifting work patterns with generative AI. NBER Working Paper No. 33795. https://doi.org/10.3386/w33795
- MacKenzie, D., & Wajcman, J. (Eds.). (1999). The Social Shaping of Technology (2nd ed.). Open University Press. https://mitpress.mit.edu/9780262631600/the-social-shaping-of-technology/
- Ye, X. M., & Ranganathan, A. (2026). AI doesn’t reduce work, it intensifies it. Harvard Business Review, 9 February. https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it