Overwork feels free to the organization that demands it. The extra hours are already paid for, the salaried worker cannot bill more, and the output looks like a bonus extracted at no cost. That impression is an accounting illusion. The cost of running people past a sustainable pace does not vanish. It moves onto ledgers the company is not looking at: lost productivity, medical leave, and the expense of replacing the people who eventually walk out. Someone is always paying for overwork. It is rarely the person who ordered it.
The productivity ledger
Start with the largest and least visible cost. Gallup (2025), surveying 263,810 people, found global employee engagement at just 20 percent in 2025, the lowest reading since 2020, and estimated the productivity lost to that disengagement at roughly 10 trillion dollars worldwide, about 9 percent of global GDP. Overwork feeds disengagement directly: people pushed past what they can sustain stop investing discretionary effort, going through the motions instead. The paradox of overwork is that demanding more hours often buys less real output, because the marginal hours come from a depleted worker producing depleted work. The company sees a full timesheet and misses the hollow productivity behind it.
The medical ledger
Where overwork is chronic, the cost eventually shows up in a form no spreadsheet can hide: people become too ill to work. ANAMT (2026), using official INSS data for leaves longer than fifteen days, recorded mental-health absences in Brazil rising from 219,850 in 2023 to 393,670 through November 2025, with the benefits paid crossing 954 million reais in 2024. Burnout cases tripled over that span. This is the cost of pressure rendered in currency: a near-billion-real bill for the fraction of harm that reached the threshold of formal, certified leave. The full cost is larger, because most of the damage from overwork never becomes a claim. It stays as diminished performance, quiet suffering, and the slow erosion of people still at their desks.
That these figures come from administrative records rather than surveys matters. They are not reports of how people feel. They are counts of people a doctor signed off work. The pressure translated into medical fact.
The intervention that lowers the bill
The useful news is that at least one intervention has been shown to reduce the harm while improving output, not despite it. Fan, Schor, Kelly and Gu (2025), in a six-month trial across 141 organizations in six countries covering 2,896 workers, found that cutting working time without cutting pay lowered burnout and improved mental and physical health, a pattern absent from control firms. Crucially, as Nature (2025) noted in its coverage, workers felt just as productive on four days as on five. The study is quasi-experimental, resting on companies that volunteered rather than a randomized draw, so it is a strong signal rather than proof. But it undercuts the core assumption behind overwork, that more hours mean more output. If people can produce comparable work in less time and end up healthier, the extra hours were never buying what they seemed to.
Why the mechanism holds
The reason a saner pace can lower cost without lowering output points back to how depletion works. Fan, Schor, Kelly and Gu (2025) traced their wellbeing gains to better work ability, less fatigue, and fewer sleep problems. A rested worker produces more per hour than an exhausted one, so trimming the exhausted hours costs less real output than the raw math suggests. And Morikawa, Martela and Hakanen (2025), across samples of 425 and 2,000 workers in Finland, found autonomy associated with lower burnout partly through more manageable job demands, though their design is correlational and cannot prove causation. The through-line is that human output is not linear in hours. Past a point, more time yields less value and more harm, and the harm is what turns into cost.
This non-linearity is why overwork can look profitable in the short run and prove costly over any longer horizon. In a single crunch, extra hours from a still-fresh team genuinely add output, which is why the tactic persists. The damage accrues later and elsewhere: in the burnout claim filed months on, in the resignation that follows, in the slow decline of a team that never fully recovered. Because the benefit is immediate and visible while the cost is delayed and diffuse, the accounting almost always flatters overwork at the moment the decision is made. Correcting that bias means pricing the lagged costs into the present choice rather than discovering them after the fact.
Put the cost where it belongs
The practical discipline is to stop treating overwork as free and start accounting for it. Model the productivity drag Gallup (2025) quantifies, the medical-leave expense ANAMT (2026) makes concrete, and the turnover that trails burnout, and set that total against the cost of a more sustainable cadence. Fan, Schor, Kelly and Gu (2025) suggest the redesign can lower the bill and improve health at the same time. Overwork does not save money. It relocates the cost onto ledgers the person demanding the hours does not have to read, and then calls the result efficiency. The bill comes due regardless. The only decision is whether a company chooses to see it before its people pay it in full.
Sources
- Gallup (2025). State of the Global Workplace 2025. Engagement at 20 percent and roughly 10 trillion dollars in lost productivity quantify the hidden productivity cost of disengagement and overwork. https://www.gallup.com/workplace/659610/state-global-workplace.aspx
- ANAMT (2026). INSS data on mental-health leaves. A near-billion-real benefits bill in 2024 and tripled burnout cases render the medical cost of pressure in hard currency. https://www.anamt.org.br/portal/2026/01/27/levantamento-anamt-com-dados-oficiais-do-inss-revela-crescimento-dos-afastamentos-decorrentes-de-problemas-de-saude-mental-entre-2023-e-2025/
- Fan, Schor, Kelly and Gu (2025). Work time reduction via a four-day workweek. Cutting hours lowered burnout and improved health while workers felt just as productive, undercutting the assumption that overwork buys output. https://www.nature.com/articles/s41562-025-02259-6
- Nature (2025). News coverage of the four-day-week trial. Reported workers felt just as productive on four days, the accessible framing of no output loss. https://www.nature.com/articles/d41586-025-02295-2
- Morikawa, Martela and Hakanen (2025). Self-management and employee well-being. Autonomy was linked to lower burnout via more manageable demands, explaining why output is not linear in hours. https://journals.sagepub.com/doi/10.1177/23409444241307734