The CEO of a mid-sized Australian financial advice firm remembers the moment the four-day week became a moral question for her business. "We felt that when we sat down with clients every day, encouraging them to live their good life, it was a little rich if we didn't do the same," she told researchers at Swinburne University [1]. Two years later, her firm is one of fourteen in a small but telling Australian cohort still running the model, and the anecdote captures something the spreadsheets cannot: the four-day week in 2026 is being adopted not as a productivity experiment, but as an answer to workers who have stopped pretending they are fine.
The first half of 2026 produced the first credible dataset for the four-day week at scale. A peer-reviewed Australian study of the 100:80:100 model appeared in Nature in May [1]. Germany's largest-ever trial released its two-year follow-up [3][4]. In April, OpenAI published a policy paper proposing that governments actively encourage companies to pilot four-day weeks at full pay [6]. Each on its own would be a footnote. Together, they form the basis for saying the model works at scale.
What the Australian study actually found
The Hopkins, Bardoel and Djurkovic paper, published in May in Humanities and Social Sciences Communications (a Nature Portfolio journal), is the first academic study of the 100:80:100 model: 100% of pay, 80% of hours, 100% of output [1]. The researchers interviewed 15 Australian firms over two years using semi-structured qualitative methods, building a richer picture than the surveys that have dominated the field until now.
The headline finding is uncomfortable for those who have argued the four-day week only works with a productivity miracle. The authors found that "a 25% increase in hourly productivity was not required to maintain previous performance levels" [1]. In other words, the maths works without heroic assumptions. Workers simply reorganised their week: cutting low-value meetings, batching client work, trimming the long tail of email traffic. Productivity did not need to soar. It needed to be used differently.
A companion piece by the lead author, John Hopkins, republished by The Independent in May, adds the operational detail [2]. Fourteen of the fifteen firms were still on the 100:80:100 model at interview; one had abandoned it, citing timing problems during a separate organisational change rather than the policy itself. Six of the fifteen reported productivity had increased. The rest said it was "about the same". None reported a drop. Six firms named burnout reduction as their primary motivation, and the paper's authors situate that finding against a national mental health picture that is harder to ignore each year: half of Australians surveyed by Beyond Blue in mid-2025 said they had experienced burnout in the past twelve months, with young adults and primary caregivers the most exposed [7].
The study has clear limits. Fifteen firms is a small sample. The firms are self-selected, and qualitative interviews capture narrative, not statistical effect. The Nature publisher's note flagged the manuscript as an "unedited version" being circulated for early access [1]. Treat the Australian evidence as suggestive, not conclusive. It is, however, the first peer-reviewed window into what the model looks like in workplaces that have run it long enough to forget the novelty.
Germany's largest trial: two years on
Germany's pilot is harder to dismiss. Coordinated by 4 Day Week Global, the consultancy Intraprenör and the University of Münster's Chair for Transformation of Work, the trial involved 45 organisations from diverse sectors, running for six months from early 2024 [3]. The first-results report appeared in October 2024, and the academic follow-up, published in February 2026, tracks what happened over the next two years [4].
The follow-up found that 70% of pilot organisations continue with some form of reduced working time two years in, with 22% having further adapted their model [3]. (Earlier media coverage cited a figure closer to 73% based on the October 2024 first-results report; the slight difference reflects a different measurement point, not a contradiction [3].) Nearly 40% of pilots were run in selected teams rather than organisation-wide, an important caveat for anyone tempted to generalise from headline numbers.
The Münster researchers report sustained gains in innovation, retention, employer attractiveness and work-life balance, with stable key financial indicators across the cohort [4]. Where the model creaked, it creaked in predictable places: workload peaks during absences, and the broader economic volatility that has unsettled European business since 2024. The pattern across both the Australian and German studies is consistent. The model does not collapse under scrutiny, but it requires the kind of managerial discipline that a slack economy tends to erode.
The surprising new drivers: AI, fuel and burnout
The four-day week has been knocking around since at least the 2010s. Why is the 2026 evidence landing differently? Three forces, two of them new, have converged.
The first is AI. In April 2026, OpenAI released a paper called Industrial Policy for the Intelligence Age, proposing that firms be incentivised to pilot four-day weeks with no loss of pay [6]. The argument is that if AI systems can do the work of months in a fraction of the time, the gains should flow to workers as time, not just to firms as output. The document also proposes more jobs in childcare, education and healthcare, higher retirement contributions, subsidised childcare, and a public wealth fund. The paper is explicit that these are "initial ideas" chiefly aimed at the United States and intended to prompt policy discussion rather than bind firm behaviour [6]. OpenAI has a commercial interest in the framing: the same paper assumes workers are paid more while firms pay for AI subscriptions. Cambridge's Gina Neff made the obvious point in the same BBC coverage: "The difference now is that OpenAI wants other companies to pay workers more while also paying them for subscriptions to their services" [6]. The proposal is best read as policy thought-experiment and corporate positioning at the same time. The fact that it has been made at all is the news.
The second driver is oil. Iran's closure of the Strait of Hormuz in early 2026 disrupted a large share of global crude and product flows through the chokepoint, with the IEA reporting flows falling from roughly 20 million barrels per day to a trickle and Gulf producers cutting total oil production by at least 10 million barrels per day [9]. The IEA's March 2026 Oil Market Report described the disruption as the largest supply disruption in the history of the global oil market, with stagflation and recession concerns rising in its wake [9]. Commuting is a meaningful line item for working households, and the four-day week is, among other things, a 20% reduction in fuel use for the workers who can do their jobs remotely. Australian firms cited the fuel disruption and burnout as twin macro drivers of renewed interest in the model [2]. Climate and cost pressures, long backgrounded in the four-day-week conversation, are now foregrounded.
The third is burnout, the oldest of the three. The Beyond Blue data is unambiguous: half of working Australians, and half of those doing unpaid care work, have hit burnout in the past year [7]. Young adults aged 18-29 are the most affected cohort. The UK pilot of 2022, still the world's largest completed trial, found that 71% of workers reported reduced burnout during the pilot, and 92% of the 61 participating companies chose to continue the model afterwards [8]. Burnout is no longer a soft argument for the four-day week. It is the hard economic case.
Who is actually doing it, and what is going wrong
The temptation, faced with this much positive evidence, is to declare the model solved. Resist that. The German data shows the model works in roughly seven of ten organisations, which is a strong result and a real failure rate. The Australian data shows one firm in fifteen walked away, and that was with a small, motivated cohort that had already opted in [1][2]. The UK pilot's 92% continuation rate is the strongest number in the literature, and it is the number most often cited in isolation [8]. The Münster researchers, in their longer arc, found that economic volatility and workload peaks are the two main remaining failure modes [4].
The model also has blind spots. Most of the evidence base is from knowledge work, professional services, and the public sector. Manufacturing, hospitality, retail, and healthcare, the sectors that employ the most workers globally, are barely represented in the trial data. Where they do appear, the pilots tend to be team-by-team rather than organisation-wide, which limits what they can tell us about structural change [3]. A four-day week in a call centre looks different from a four-day week in a marketing agency, and the evidence to support the former is thin.
There is also the question of who gets to participate. The 100:80:100 model is, in effect, a wage protection for workers whose jobs can absorb a 20% reduction in time. For low-wage, hourly, or shift-based workers, the model often translates into longer shifts, denser rosters, or compressed schedules that do not deliver the same well-being gains. The Australian study maps its findings to UN Sustainable Development Goals 3, 5 and 8, and the gender dimension is real: unpaid care work falls disproportionately on women, and a four-day week that does not address that asymmetry will entrench it [1]. The Hopkins paper is candid about this; the broader policy conversation has been less so.
A serious policy option, not a fringe idea
In 2022, the UK pilot established that the four-day week was not a slogan. In 2025, the Fan, Schor, Kelly and Gu paper in Nature Human Behaviour added peer-reviewed evidence of well-being improvements across multiple countries under organisation-wide trials [5]. In the first half of 2026, the Australian and German studies, plus the OpenAI proposal, have produced something resembling a policy case.
The cautious take is that the four-day week works well enough, often enough, in the right kinds of workplaces, that governments and large employers should treat it as a live option rather than a curiosity. The harder case the model must now answer is this: it has matured into a genuinely useful tool, but a tool with a known failure rate, a narrow evidence base outside professional services, and a structural risk that it benefits already-privileged workers most.
What is no longer defensible is the position that there is no good evidence for the four-day week. There is. The argument now is about who it works for, under what conditions, and at what cost. That is a different, and more useful, debate.