The illusion of time freed up by AI, the importance of governance

The prevailing narrative surrounding artificial intelligence is deceptively simple: by automating routine tasks and speeding up knowledge work, AI is supposed to reduce the human workload and free up time. Yet the reality unfolding within organisations suggests the opposite. Workers are not experiencing less work, but more: more tasks, more output, more expectations. The paradox is not technological; it is systemic.

A growing body of research from institutions such as the Harvard Business Review, the National Bureau of Economic Research and the London School of Economics confirms that artificial intelligence significantly boosts productivity at the level of individual tasks (HBR, 2023; NBER, 2024; LSE, 2023)[1,2,3]. Writing, coding, analysis: everything is completed in a fraction of the time previously required. However, embedded within this literature is a largely unquestioned assumption: that productivity gains can translate into reduced workloads. This assumption is flawed, because it treats productivity as a purely technical variable rather than as a resource that must be managed.

The key issue is not whether AI saves time, but who controls what happens to that time once it has been created.

The time saved does not go back to the workers; it becomes capacity

Once this distinction has been made, the apparent contradiction becomes clearer. The time saved through AI does not automatically revert to workers: it is transformed into capacity, and that capacity must be allocated. Organisations rarely interpret efficiency gains as an opportunity to reduce workload. On the contrary, they see them as a sign of underutilised resources. Consequently, what appears to be ‘freed-up time’ at the level of individual tasks is quickly reabsorbed into the system through new assignments, tighter deadlines and expanding expectations.

At the individual level, AI reduces the time needed to complete specific tasks, but this reduction lowers the perceived cost of the effort and broadens the scope of what is considered feasible. Tasks that were previously postponed, delegated or deemed unnecessary become standard practice. Workers, responding to evolving incentives and expectations, take on additional responsibilities because the marginal cost of doing so appears lower. What begins as a productivity gain at the micro level thus triggers a broader process of expanding the workload.

How organisations turn efficiency into pressure

This dynamic becomes more pronounced at the organisational level. Companies do not treat efficiency gains as spare capacity but as capacity to be reallocated. Performance metrics are adjusted upwards, turning yesterday’s exceptional output into today’s baseline. Managers reallocate resources, raise targets and shorten delivery cycles. At the same time, AI introduces new forms of work that offset some of the time saved, including verification, correction, integration and supervision.

These activities are not peripheral: they are fundamental, and ensure that human involvement remains central even as automation increases (McKinsey, 2023; WEF, 2023)[4,5]. Execution becomes faster, but the overall effort does not decrease.

Performance metrics are being raised; yesterday’s exceptional output becomes today’s baseline.

Beyond individual organisations, market forces reinforce this pattern. Competition ensures that efficiency gains cannot go to waste. Companies that translate productivity gains into faster deliveries or higher output gain an advantage, forcing others to follow suit. Increased supply stimulates demand, and customer expectations adjust accordingly (WEF, 2023; BCG, 2023)[5,6]. What was once considered fast becomes the norm, and what was once sufficient becomes inadequate. Speed becomes a competitive requirement rather than a discretionary choice.

The empirical evidence: a consistent pattern on a global scale

The empirical evidence supports this systemic interpretation. Research by the National Bureau of Economic Research (NBER, 2024)[2] suggests that the adoption of AI reshapes the labour supply and the allocation of tasks rather than reducing total working hours. Analyses published by the Harvard Business Review (HBR, 2023)[1] show that generative AI accelerates knowledge work whilst simultaneously expanding the scope of tasks and increasing the need for supervision.

Findings from the London School of Economics (LSE, 2023)[3] highlight that productivity gains derived from AI are primarily captured as increased output rather than reduced working hours. Complementary analyses by McKinsey (McKinsey, 2023)[4] and the World Economic Forum (WEF, 2023)[5] reinforce this pattern, showing that technological productivity gains tend to drive economic expansion rather than reductions in working hours. Taken together, these findings do not contradict one another: they reveal a consistent dynamic in which productivity gains are reabsorbed rather than redistributed.

Productivity gains do not distribute themselves. Without explicit mechanisms to channel them, they are systematically absorbed into increased output, greater coordination requirements and ever-expanding expectations.

The Jevons paradox applied to AI: efficiency increases, not decreases

This pattern is not new. As William Stanley Jevons observed in 1865 in the context of coal consumption, increases in efficiency often lead to greater consumption rather than conservation — a dynamic now known as the Jevons paradox (Jevons, 1865)[7]. In complex systems, such dynamics are reinforced through feedback loops, as described by John D. Sterman (Sterman, 2000)[8]. Applied to modern organisations, these principles explain why local efficiency gains generated by AI can lead to system-wide expansion rather than contraction. Efficiency does not reduce activity: it reshapes and amplifies it.

The deeper implication is that artificial intelligence does not determine outcomes on its own. Technology changes what is possible, but it is institutions that determine how those possibilities are realised. Historically, reductions in working hours have not emerged automatically from technological progress: they have required deliberate governance mechanisms, labour regulation and collective bargaining, which convert productivity gains into protected time. In the absence of such mechanisms, the default outcome is predictable: organisations convert efficiency into output, markets convert it into competitive pressure, and workers absorb the resulting increase in workload.

The implications for leaders: measuring capacity, not just throughput

For leaders, this shift in perspective has significant implications. The question is not whether AI boosts productivity – it clearly does. The question is how the capacity created by that productivity is managed. AI generates capacity, and if that capacity is not explicitly managed, it will be consumed by the system.

Organisations that continue to measure performance solely in terms of throughput will find that AI intensifies the workload (HBR, 2023; McKinsey, 2023)[1,4]. Those that incorporate metrics relating to quality, sustainability and human capacity may begin to see a different outcome.

AI does not save time by default: it creates capacity. And it is the system, not the technology, that determines how that capacity is used.

The governance gap: a dimension that has been overlooked in the literature

The failure of AI to free up time for workers is therefore not a limitation of the technology itself, but rather a reflection of a governance gap. Whilst the literature has made substantial progress in explaining how AI improves productivity, it has paid far less attention to how those gains are allocated.

This analysis fills that gap by demonstrating that productivity gains are not self-distributing. Without explicit mechanisms to channel them, they are systematically reabsorbed into increased output, greater coordination requirements and expanding expectations. (Photo by Mohamed Nohassi on Unsplash)

References

[1] Harvard Business Review. (2023). How generative AI changes knowledge work. https://hbr.org/2023/07/how-generative-ai-changes-knowledge-work

[2] National Bureau of Economic Research. (2024). Generative AI at work (NBER Working Paper No. 31161). https://www.nber.org/papers/w31161

[3] London School of Economics. (2023). Generative AI and productivity: Evidence from early adoption. https://www.lse.ac.uk/research/research-for-the-world/economics/generative-ai-productivity

[4] McKinsey & Company. (2023). The economic potential of generative AI: The next productivity frontier. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier

[5] World Economic Forum. (2023). The future of jobs report 2023. https://www.weforum.org/reports/the-future-of-jobs-report-2023

[6] Boston Consulting Group. (2023). Navigating the generative AI revolution in business. https://www.bcg.com/publications/2023/generative-ai-business-impact

[7] Jevons, W. S. (1865). The coal question: An inquiry concerning the progress of the nation and the probable exhaustion of our coal-mines. Macmillan.

[8] Sterman, J. D. (2000). Business dynamics: Systems thinking and modeling for a complex world. McGraw-Hill.

ALL RIGHTS RESERVED ©