Europe's Deskless Workforce 2026

Europe does not have one labour market and it certainly does not have one working week. The further you move from office employment, the more the standard assumptions break down: the hours are longer or shorter than the average rather than close to it, more of the work happens at the weekend, the safety exposure is materially higher, and the vacancies are harder to fill. Those differences are visible in official statistics, but they are usually reported one indicator at a time, which makes them easy to miss.

We put the indicators next to each other for a consistent set of sectors, so the picture can be read as a whole. The sectors we return to throughout are manufacturing, construction, wholesale and retail trade, transport and storage, accommodation and food services, and administrative and support service activities. They are the parts of the business economy where non-desk work concentrates, and they are the sectors where Eurostat publishes comparable figures across all of the indicators used here.

The last part of the report is the part that is not in the statistics. Enterprise digital adoption is broad at company level, but a large share of the workforce cannot reach the systems being adopted, because reaching them assumes a company email account, a computer, and someone available during office hours. We call that the HR Access Gap, and it is a framework for reading the data rather than a score we calculated.

1. The European working week is shrinking, almost everywhere. Actual weekly hours for the EU fell from 36.9 in 2015 to 35.9 in 2025, a decline of one hour, and 26 of 27 Member States moved in the same direction.
2. The average hides a seven-hour gap. In 2025 the spread between Member States was 7.7 hours, from 39.6 in Greece to 31.9 in the Netherlands. A single European scheduling standard would fit almost nobody.
3. Weekend work and safety exposure follow the job, not the country. Around a fifth of employed people in the EU work weekends, but that rises to roughly half in agricultural and in service and sales occupations, and accident incidence in construction is more than double the all-activities figure.
4. Company digital adoption and employee digital access are two different things. Cloud adoption is above half of EU enterprises while AI adoption is around a fifth, yet adoption at company level says nothing about whether a warehouse operative can request leave without walking to an office.

Between 2015 and 2025 the average actual weekly working time in the EU fell from 36.9 hours to 35.9 hours. A one-hour decline over a decade sounds modest until you notice how broad it is: 26 of the 27 Member States recorded a decrease. Whatever sits behind it, the direction is close to continent-wide, so plans built on a stable European working week are being built on a moving figure.

The more useful number for anyone planning across borders is the spread. In 2025 the distance between the longest and shortest average weeks among Member States was 7.7 hours, with Greece at 39.6 hours and the Netherlands at 31.9. Two plants with the same headcount, the same output targets and the same shift software will produce completely different rosters in those two countries, and neither of them is an exception to be corrected.

For HR teams operating in several countries, this argues against harmonising the roster and in favour of harmonising the process. What a manager needs to see, approve and record can be the same everywhere. The pattern of hours underneath it cannot.

Weekend work in these statistics means usually working on Saturdays or Sundays, or both, as part of the normal job. It is not overtime and it is not an emergency call-out. It is simply how the week is shaped for a large part of the workforce.

Across the EU, 21.0% of employed persons aged 20 to 64 usually work weekends. Broken down by occupation the picture changes sharply: 47.6% of skilled agricultural, forestry and fishery workers and 47.3% of service and sales workers do so. Close to half of two large occupational groups treat Saturday or Sunday as a normal working day.

A second series, based on employees rather than all employed persons aged 20 to 64, gives an EU27 figure of 18.1%, ranging from 31.3% in Greece down to 3.0% in Lithuania. The two numbers are not in conflict. One series counts employees, the other counts employed persons, so they should not be treated as the same denominator. If you quote one of them internally, quote which one.

The practical consequence for HR is availability. When a meaningful share of a workforce is on site at the weekend, any process that depends on an administrator being reachable Monday to Friday between nine and five is not really available to those employees. It is available to their managers.

In 2025, 6.3% of employed persons aged 20 to 64 in the EU usually worked 49 hours or more per week. The two groups furthest above that line are skilled agricultural, forestry and fishery workers at 25.4% and managers at 20.0%.

That pairing cuts across the usual framing. Very long weeks are concentrated at two ends of the labour market that have little else in common, and the middle of the distribution, including most shift-based industrial and service work, is not where the 49-hour weeks live. The distribution of long hours therefore does not match the distribution of weekend work described above.

Reading hours, safety and vacancies together for the same six sections produces a clearer result than any one of them alone. Only one section sits above its relevant comparator on all three: construction, with 38.6 average weekly hours against 35.9 for all activities, 2,899.23 accidents per 100,000 against an EU total of 1,392.49, and a job vacancy rate of 2.9% against the 2.0% EU business-economy comparator.

That combination describes a sector working longer, exposed more, and struggling harder to fill roles at the same time. It does not tell us that any one of those causes another, and we make no such claim here.

Administrative and support service activities sits differently. Average hours are 33.6, below the all-activities figure, while accident incidence is 2,042.55 and the job vacancy rate is the highest of the six sections examined at 3.2%. Shorter average weeks combined with above-comparator accident incidence and the highest vacancy rate is the observed combination here.

Accommodation and food services is the quiet case. Average hours land at 35.9, exactly the all-activities comparator, which is why the sector rarely appears in working-time commentary. Its accident incidence of 1,467.88 and vacancy rate of 2.4% are both above their comparators, so an average-looking week conceals above-average pressure on the two things HR is judged on: keeping people safe and keeping shifts covered.

Administrative & support services: staffing/temp agencies · security · cleaning/facilities · call centres.

The job vacancy rate is the share of posts that are vacant out of all posts, occupied and vacant together. It is a measure of unfilled demand at a point in time. It is not a turnover rate, not a resignation rate, and not a measure of how hard people are to retain. A sector can have a low vacancy rate and still churn heavily if it refills roles quickly.

In 2025-Q4 the EU comparator across the business economy stood at 2.0%. Of the six sections examined, administrative and support services reported 3.2%, construction 2.9%, accommodation and food services 2.4%, transport and storage 1.9%, wholesale and retail trade 1.8% and manufacturing 1.5%.

Read as an ordering, the six sections examined rank from administrative and support services at the top to manufacturing at the bottom on this measure. That ordering describes unfilled demand in 2025-Q4 and nothing further.

In 2023 the EU recorded 1,392.49 non-fatal accidents at work per 100,000 persons employed. Around that figure the sectoral variation is wide: construction 2,899.23, transport and storage 2,365.98, administrative and support services 2,042.55, manufacturing 1,686.95, agriculture 1,610.12, human health 1,506.59, accommodation and food services 1,467.88 and retail 1,215.10.

Of the sectors examined here, construction carries the highest incidence, at more than double the all-activities figure. Health and accommodation and food, sectors that are not usually described as hazardous, both sit above the EU total as well.

The accident data measures incidence only. It says nothing about training or record keeping. As a separate business implication, and not something these figures demonstrate, organisations in higher-incidence sectors generally need training records, instruction acknowledgements and certification validity to be current, provable and reachable by people who are rarely in front of a computer, in the language they work in.

Using Structural Business Statistics for 2024, preliminary, the six sections examined here employ 110,403,152 people, which is 67.2% of the 164,201,284 employed in the SBS business economy total.

One clarification matters more than the number itself. This is a count of people employed in these sections, not a count of deskless workers. Every one of these sections contains office roles: finance, planning, engineering, sales, HR. The figure tells you how much of the business economy sits in sectors where non-desk work is concentrated. It does not tell you what share of European employees are deskless, and we do not present it that way.

Among enterprises with 10 or more employees, 19.95% in the EU report using AI, ranging from 42.03% in Denmark to 5.21% in Romania. Cloud adoption is far more established, at 52.74% for the EU, led by Finland at 79.21%, followed by Italy at 75.60%, Malta at 75.07%, Ireland at 73.04% and Sweden at 72.00%.

These are company-level indicators. They record what the enterprise has adopted, which is not the same as what an individual employee can use during their shift. A company can be counted as a cloud adopter while its cleaning teams still sign a paper sheet at reception.

Eurofound's work on the effects of technology at work points in a consistent direction: technology is more likely to add tasks to jobs than to remove them. For workforces that are already administratively underserved, that is the important part. New systems tend to arrive as additional steps, and someone has to absorb them.

For many non-desk roles, a corporate email address and a company laptop are simply not part of the job. That is not a technology failure or a training gap. It is the shape of the work.

A process can be fully digital from the company's point of view and still be inaccessible to the employee if it depends on a corporate email address they were never given, a computer that is not part of their job, a language they are not comfortable using, or an administrator who is only available during office hours. Each of those conditions is invisible on an adoption dashboard, and each of them turns a self-service process back into a queue at the HR office.

The principle we work from is straightforward: the employee should not need office infrastructure to access HR if office infrastructure is not part of their job. The HR Access Gap is the distance between the two. It is a framework for diagnosis, not a numeric composite score, and nothing in this report calculates it.

• Which of our HR processes assume a company email address, and how many of our employees do not have one?
• If a third of our people are on site at the weekend, who answers their questions on Saturday?
• Can we prove, today, that every worker on our highest-exposure sites has current, valid training in a language they understand?
• Are our vacancies concentrated in the roles where the working pattern is hardest, or where the administration around the job is hardest?
• When we adopted our last system, did it remove steps for site or shift supervisors or add them?
• Do our country rosters differ because the work differs, or because we exported one country's assumptions?

The Data & AI Pack contains the indicator tables behind this report in a structured, machine-readable form, with reference periods and status flags preserved. You can compare your own country and sector against the figures here, or hand the pack to an AI assistant and interrogate it directly.

Sample prompts

• "Using this dataset, compare Romania, Poland and Germany on actual weekly hours, weekend work and AI adoption, and note the reference period for each figure."
• "Rank the six sections by job vacancy rate and by accident incidence, and tell me where the two rankings disagree."
• "My workforce is 70% construction and 30% administrative and support services in Greece. Summarise the working-time and safety context I should plan around."
• "List every figure in this pack that carries a Eurostat status flag and explain what each flag means for interpretation."