12 min readLast updated 24 September 2026

The new working model during the AI transition: Freelancing

For the past century the default was to stay with one employer and slowly rise through the ranks. Globalisation and the internet made changing employers routine. With the rapid growth of AI, freelancing is becoming mainstream, driven by two key forces.

Summary

For the past century, the default working model for most people was to stay with one employer and slowly rise through the company ranks.

Globalisation, and more recently online job boards like LinkedIn, made changing employers far easier. The long career at a single company, once ordinary, became increasingly more rare. Algorithms could now suggest relevant job opportunities, and the friction of applying was reduced to a few clicks, that often required less than 10 seconds (e.g. with LinkedIn’s “Easy Apply” feature).

However, in the past ~15 years, and especially with the advent of artificial intelligence, a new paradigm has emerged in the job market: freelancing is becoming mainstream, and replacing full-time employment. Two key forces are primarily driving this new trend, and as I explain later in the article, it is expected to become even more prominent over the next 10 to 20 years, as we navigate the AI transition phase and approach ASI (Artificial Superintelligence).

The two key forces

Demand: AI is reshaping the talent companies need, with the demand for AI-related skills surging across industries. Companies are trying to adopt the latest AI technologies, but are struggling to keep up with the pace of change, making the skills gap wider. This creates a growing demand for skilled freelancers who can quickly fill the gap and help companies navigate the AI transition.
Supply: The unprecedented technological advancements since the launch of ChatGPT in 2022 enable a new wave of early-adopters of AI to begin freelancing. This trend, combined with the fact that younger generations hold the strongest preference for working independently, suggests a growing supply of skilled freelancers in the coming years. Evidence of this trend is already visible in the data since the early 2010s, and seems to have accelerated significantly since 2025 (as LLMs reached a level of ‘intelligence’ that enabled them to be used for a wide range of tasks, and not just for research purposes).

How we got here

The 3 phases of the labour market

Since the 20th century, there have been roughly three phases in the labour market. This section walks through all three:

a lifetime at one employer
constantly switching employers every few years
the rise of freelancing

Phase 1: A lifetime at one employer, climbing a single corporate ladder

For most of the last century the default was to stay with one employer and climb the company ranks. The career was a ladder inside a single organisation, and the reward for loyalty was seniority: a better title, more responsibility, and a comfortable pension at the end of it.

Stability and predictability mattered far more then than now, at a time when the available job options were more limited. Information travelled slowly, hiring was local, and a steady position with a known employer was something to hold on to, rather than to trade up from.

The evidence: workers once chose the secure job over the better-paid one, while nowadays, job-security ranks behind balance and pay

According to surveys, when individuals were offered a trade between pay and security, security won by a wide margin. In contrast, nowadays, job security usually ranks behind work-life balance and pay in what people say they want.

Insight #1

ILR Review, 1947US national poll, reported October 1947

In 1947, when given two options, Americans chose the low-paid job they were sure of keeping over the well-paid one they might lose, by about two to one.

The question was a forced choice: “a job which pays quite a low income but which you were sure of keeping” against “a job which pays a good income but which you have a fifty-fifty chance of losing”. About half chose the sure job and about one quarter the risky one. The same choice put to American students in 1943 came out the same way, with 47% picking the sure low-paid job.

Sources: a contemporary national poll quoted in the inaugural issue of the Industrial & Labor Relations Review (October 1947); the student figures are Occupations (March 1943, p. 571). Polls of that era used quota rather than probability sampling, so treat the exact splits as directional.

Insight #2

Gallup Panel, 202213,085 US employees, fielded October 2021

Contrary to the 20th century perception, nowadays, when employees are asked what matters in their next job: stability and job security come fourth, behind pay, work-life balance and the ability to do what they do best.

In Gallup’s probability-based panel, the shares calling each factor “very important” when deciding whether to take a job with a new employer were: income and benefits 64%, greater work-life balance and personal wellbeing 61%, the ability to do what they do best 58%, and greater stability and job security 53%. Since 2015 the balance share has climbed eight points while the security share has not moved.

One honest caveat: job security has not become unimportant. A majority still names it, long-running academic surveys still place it among the top-rated job attributes in absolute terms, and it reliably surges back after downturns, reaching 73.6% “very important” in 2016 in the shadow of the Great Recession. What has changed is its rank among the things people say they are looking for.

Sources: Gallup, “The Top 6 Things Employees Want in Their Next Job” (February 2022; Gallup Panel, 13,085 US employees, fieldwork October 2021). The absolute-importance and recession points are Arne Kalleberg and Peter Marsden, “Work Values in the United States” (The ANNALS of the AAPSS, 2019), from the General Social Survey; the same research finds the overall ranking of job values in the US fairly stable since the 1970s, which is why this pop-up pairs the 1940s with today rather than claiming a steady year-by-year decline.

Executive insights

You stayed loyal, and the company carried you up the ladder for life:

“Like many other large companies in the United States, Europe, and Japan, GE has had an implicit psychological contract based on perceived lifetime employment. … This produced a paternal, feudal, fuzzy kind of loyalty. You put in your time, worked hard, and the company took care of you for life.”

Jack Welch, Chairman and CEO of General Electric, describing the old contract he was then dismantling, in the interview “Speed, Simplicity, Self-Confidence”, Harvard Business Review (September–October 1989); the ellipsis trims a sentence on the rare exceptions.

“For the last sixty or so years, the job market for educated workers worked like an escalator. After graduating from college, you landed an entry-level job at the bottom of the escalator at an IBM or a GE or a Goldman Sachs. … So long as you played nice and well, you moved steadily up the escalator, and each step brought with it more power, income, and job security.”

Reid Hoffman, co-founder of LinkedIn, with Ben Casnocha, in the opening chapter of The Start-up of You (2012); the ellipsis trims two sentences on employer-provided training.

Phase 2: Switching employers becomes easier, and more frequent

Globalisation opened hiring across borders. More recently, since the ‘dot-com’ boom in the early 2000s, the online job boards like LinkedIn further fuelled this new trend of frequently switching employers.

Over time, job-boards’ algorithms got increasingly more sophisticated, and employees could now receive personalized job recommendations, and apply with just a few clicks. This greatly reduced the friction of job searching, and the frequency of switching jobs grew, holding all else constant. The data shows that at any given age under fifty, people now hold a job for less time than their parents did. Also, in America, the ten-year or twenty-year career at a single company, once ordinary, has become the exception.

The evidence: the long single-employer career becomes much rarer, and at any given age under fifty, the typical job tenure is shorter

For economists, it is very complex to accurately measure the average job-tenure over time as it can be influenced by various factors, such as the state of the economy and women’s labour force participation. Despite these complexities, the data from the two studies below show that: (1) Spending a decade (or more) with the same employer went from normal to unusual. (2) The typical job tenure is shorter than it used to be before the 21st century.

Insight #1

Farber, NBER, 2010US private-sector men, 1973 vs 2006

Spending a decade with one employer went from normal to unusual: half of American men in mid-career had done it in 1973, only about a third by 2006.

The decline in long-term employment: share of American men who had spent 10 or 20 years with a single private-sector employer, 1973 vs 2006

19732006

50%35%

35%20%

in a 10+ year jobin a 20+ year job

The data shown by the economist Henry Farber demonstrate that the long career at a single employer, once ordinary, has become unusual. It is worth noting that the decline is specific to men in the private sector, as women’s tenure rose over the same period and public-sector tenure held, which is why the all-worker average barely moved.

Source: Henry S. Farber, “Job Loss and the Decline in Job Security in the United States”, in Labor in the New Economy (NBER / University of Chicago Press, 2010). Ten years and over for men aged 35–64, twenty years and over for men aged 45–64.

Insight #2

Goulart and Oesch, EJIR, 2024Six Western European economies, 1993 vs 2021

The average job-tenure is shorter than it used to be: by about a year for people in their late twenties, and nearly two for those in their late forties.

Shorter jobs at every age under 50: mean job tenure in six large European economies, 1993 vs 2021

19932021

4.5 yrs3.3 yrs

8.2 yrs6.5 yrs

13.2 yrs11.3 yrs

age 25–29age 30–39age 40–49

Hold age constant and the job-tenure is shrinking at every stage of life below fifty, which is the same thing as saying people change employers more often than their parents did. Only the over-fifties held on for longer than the same-age worker in 1993.

Source: Kimberly Goulart and Daniel Oesch, “Job tenure in Western Europe, 1993–2021: Decline or stability?”, European Journal of Industrial Relations 30(3) (2024), Table 2, EU-LFS columns. Mean job tenure in years, pooled across Germany, France, Italy, Spain, Poland and the United Kingdom. The aggregate average barely moved (10.9 → 11.1 years): tenure falls at every band under 50 but rises for the over-fifties, and an ageing workforce plus women’s rising tenure hold the total up.

Phase 3: A growing share of skilled workers choose to work independently, as a freelancer.

In the past ~15 years, a new paradigm has emerged in the job market, which has accelerated even further due to the rapid growth of artificial intelligence: freelancing is becoming mainstream, and replacing full-time employment. Across Europe even though the total number of solo self-employed is shrinking, the grew ~27% since 2011. grew fastest of all, up about 43%, and now make up half of Europe’s solo self-employed, up from about a third in 2011.

“Skilled solo self-employed”: working for yourself, employing nobody, in a skilled occupation

Someone who works for themselves, employs nobody else, and whose main job is a skilled occupation: a manager, a professional, or a technician. Those are the two highest of the ILO’s four official skill levels, which makes this a cut by skill, not by industry.

It covers, for example, software developers, graphic designers, consultants, architects, lawyers, accountants, translators, journalists and photographers. It does not cover a solo plumber, a taxi driver, a shopkeeper, a hairdresser or a farmer, however independently they work: those sit in other occupation groups.

This skill-based cut is how IPSE, the UK’s association of the self-employed, defines a freelancer. Treat it as a close proxy rather than a headcount: it also counts solo doctors and one-person company directors who would never use the word.

A “professional on their own account”

“Professionals” is an official occupation group (ISCO major group 2): jobs that normally require a university-level qualification, such as engineers, doctors, lawyers, accountants, software developers, designers and journalists.

“On their own account” means self-employed with no employees: they sell their own work, rather than running a company with staff.

Only two occupation groups grew, and professionals grew fastest: from 3.46 to 4.95 million, up 43%

Solo self-employed in the EU by skilled occupation group, millions of people, 2011 vs 2025

20112025

0.76M0.73M

3.46M4.95M

2.30M2.62M

Managers −4.5%Professionals +43.0%Technicians +14.1%

Own-account professionals went from 3,461,200 to 4,950,600 people, up 43.0%, which is more than four fifths of the skilled group’s entire growth. Technicians and associate professionals grew 14.1%, managers declined 4.5%, and every other occupation group shrank: clerical support workers by 23%, service and sales workers by 14%, farmers by 43%, craft workers by 5%, machine operators by 10% and elementary occupations by 27%. Professionals did not just grow fastest, they are the only large group growing at all.

Source: Eurostat EU Labour Force Survey, dataset lfsa_esgais: self-employed persons without employees (own-account workers), EU27, age 20–64, by ISCO-08 major group. “Professionals” is major group 2 (software developers, engineers, doctors, lawyers, accountants, designers, journalists and the like); managers are group 1, technicians and associate professionals group 3, and together they are the “skilled” line of the chart below. Growth is 2011 to 2025; the same 2011 base and 2021 break apply as in the main chart.

Solo self-employment in the EU: the total is shrinking while the skilled part grows (indexed, 2011 = 100)

Skilled solo self-employed · managers, professionals and technicians working for themselves, with no employees
All solo self-employed · everyone working for themselves with no employees

Source: Eurostat EU Labour Force Survey, dataset lfsa_esgais, EU27, age 20–64, indexed to 2011 = 100.

More about how this series is measured

Skilled solo self-employed counts people who work on their own (and employ nobody), and whose main job sits in ISCO-08 major groups 1–3: managers, professionals, and technicians and associate professionals. Those are the two highest of the ILO’s four skill levels, which is what makes this a skill cut rather than an industry one. IPSE defines a freelancer with the same skill-based cut, applied to the UK’s occupational classification rather than the international ISCO that Eurostat uses.

Treat it as a proxy rather than a headcount: it counts solo doctors and one-person company directors who would never use the word, and it cannot see anyone who freelances alongside a main job, because only the main job is recorded. The series starts in 2011 because the ISCO-88 to ISCO-08 reclassification that year cuts the recorded number of self-employed managers from 2.5 million to 0.8 million overnight, which makes any earlier comparison meaningless. The 2021 LFS redesign adds a smaller break.

What is driving the growth of freelancing

Two “forces” are driving the growth of freelancing: one on each side of the job market

The previous section described the three phases that have shaped the labour market since the 20th century. The question that matters for the next decade is why the shift toward independent work should continue, and possibly accelerate further.

I see two “forces” driving the growth in freelancing, and they sit on opposite sides of the job market:

On the demand side, AI is reshaping the talent companies need, with the demand for AI-related skills surging and changing constantly. This constant need for new skills is creating a growing demand for skilled freelancers who can quickly fill the gap, and help companies navigate the AI transition.
On the supply side, there is a growing preference for independent work, fuelled by demographic shifts and changing career aspirations (e.g. work-life balance, desire for working remotely, independence). The data shows that this trend is especially prominent since 2025, when LLMs reached a level of ‘intelligence’ that enabled them to be used for a wide range of tasks.

1 · Demand force

Artificial Intelligence is reshaping the talent companies need, constantly

Since the release of ChatGPT in November 2022, companies have been trying to adopt AI. This has proven to be a very difficult task, because the capabilities of AI keep evolving. Every few months a new state-of-the-art model sets new records across a range of benchmarks.

That creates two main challenges for every company:

Staying up to date with the latest technological capabilities. Companies need to continuously & quickly update their plans to ensure they do not fall behind. Every couple of weeks, there are significant improvements in the latest AI technology that can change the way companies operate, and the talent they need to hire.
The evidence: state-of-the-art AI models keep getting better, constantly exceeding expectations of what AI can do
The cleanest way to see the pace is through benchmarks: fixed sets of hard problems that researchers use to score every new model, so results are comparable over time. Below, each cluster is one ability, and each bar shows the best score any AI system released by the end of that year had reached on that ability’s test. Every cluster tells the same story: what AI could barely start two or three years ago is close to solved today.
Best score by any available AI system at the end of each year, in % of test problems solved
42%77%93%95%
Science¹
–31%77%84%
Coding²
2%28%54%93%
Puzzles³
––73%84%
Computer use⁴
–8%38%46%
Multidisciplinary reasoning⁵
Sources: Epoch AI, “AI Benchmarking Hub” (science and coding, from Epoch’s own standardised runs rather than lab self-reports) and the official ARC Prize (puzzles), OSWorld-Verified (computer use) and Humanity’s Last Exam (multidisciplinary reasoning) leaderboards, retrieved 17 and 21 July 2026.
More about the method, each test, and the models behind the bars
Each bar is a running best: the highest score among all systems released by the end of that year, with older systems scored after the fact where a test was created later. A dash means no comparable measurement exists for that year. The computer-use board uses its 100-step budget.
¹ GPQA Diamond: 198 graduate-level science questions. Four-choice, so blind guessing scores 25%; maxed out since March 2026.
² SWE-bench Verified: 500 real bug reports from open-source projects; a fix counts when the project’s own tests pass. Created August 2024.
³ ARC-AGI-2: visual logic puzzles, easy for people, built to resist AI. The 2023–24 bars are bespoke competition systems, not models you could buy.
⁴ OSWorld-Verified: real tasks on a real computer desktop, completed within 100 actions; people managed about 72% in the original study. Leading entries are agent systems built on a model.
⁵ Humanity’s Last Exam: 2,500 expert questions across more than a hundred subjects; closed-book, no tools or web search.
⁶ Best scores as of 21 July 2026, placed by each system’s release date. The July 2026 bars belong to four different systems: OpenAI’s GPT-5.4 Pro (science), Anthropic’s Claude Opus 4.7 (coding), OpenAI’s GPT-5.6 (puzzles), the Pointer agent running on Opus 4.7 (computer use) and Google’s Gemini 3.1 Pro (the exam); no lab stays in front for long. When a cluster stops growing, as the science one has since March 2026, it is the test running out of headroom rather than the models running out of progress, and benchmark makers respond by writing harder tests. The newest releases, Claude Fable 5 and Kimi K3, hold no bar yet: the chart uses only independent, same-harness scores, which mostly had not been run for them, and the labs’ own reported numbers are not comparable. Epoch’s aggregate capability index, which does cover them, puts Fable 5 first.
CEO Perspectives
“Just like Moore’s Law, we saw the doubling in performance every 18 months. With AI, we’ve now started to see that doubling every six months or so.”
Satya Nadella, CEO of Microsoft, opening the Microsoft Ignite 2024 keynote (November 2024).
“As we roll out more Generative AI and agents, it should change the way our work is done. We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs.”
Andy Jassy, CEO of Amazon, in a memo to all Amazon employees (June 2025).
Making sure their employees have the skills to use, or learn how to use, the latest AI technologies.
Even when the leadership knows exactly what the best models can do, they face the enormous challenge of implementing these technologies across the organization. They need to make sure they have the right talent to turn their plans into reality. That is a much slower process, and it is where the skills gap comes in: the people who know how to use AI effectively are scarce, and the people who do not know how to use it are not learning fast enough to keep up with the pace of change.
The data shows: employers expect AI to reshape their business, and admit they currently lack the expertise to adopt AI technologies.
1. The forecast: employers expect AI to reshape a large share of the skills their people use.
2. The admission: the expertise needed for that transition is exactly what firms say they are missing.
3. The measurement: In the EU, only one in five firms actually uses AI as of 2025.
Put together: a large adjustment is coming, the skills for it are scarce, and nearly all of it is still ahead.
Insight #1
WEF Future of Jobs 20251,000+ employers, 55 economies, surveyed 2024
86% of employers name AI the single biggest force reshaping their business, and expect about 39% of workers’ core skills to change by 2030.
86% of the employers surveyed named AI and information processing as a technology they expect to transform their business by 2030, more than named any other technology.
Source: World Economic Forum, Future of Jobs Report 2025, Figures 1.2 and 3.1. A survey of over 1,000 employers representing more than 14 million workers across 55 economies, fielded May to September 2024. It samples large firms by design, so it is an employer expectation rather than a probability-based measurement.
Insight #2
Eurostat, 2025Official EU enterprise survey, ~157,000 firms
70% of firms that looked into AI and walked away name one reason above all the others: they lack the expertise.
Why EU firms that considered AI ended up not using it (share naming each reason, 2025)
Lack of relevant expertise70%
Unclear legal consequences54%
Data protection & privacy concerns53%
Availability or quality of data44%
Incompatible with existing systems42%
Costs seem too high38%
Ethical considerations25%
AI not useful for the enterprise18%
The excuse is not that AI is useless; that answer comes last, at 18%. The missing piece is people who know how to make it work, and that holds at every company size: expertise is named by 71% of small firms, 69% of medium ones and 65% of large ones. The World Economic Forum’s employer survey says the same thing globally: skill gaps rank as the single biggest barrier to transformation (63%, first in 52 of 55 economies), lack of skills is the top barrier to AI adoption specifically (50%), and 77% of employers plan reskilling programmes so their people can work alongside AI.
Sources
Eurostat, dataset isoc_eb_ai (2025 wave; this question is asked of enterprises with 10 or more staff that have considered AI without adopting it, about one EU enterprise in nine; multiple answers allowed, so shares do not sum to 100); World Economic Forum, Future of Jobs Report 2025, Figures 4.1, 4.13 and 4.14 (the 50% figure comes from the WEF Executive Opinion Survey of 11,000+ executives).
Insight #3
Eurostat, 2025Official EU enterprise survey
We are still at the beginning of the AI transition: as of 2025 only 20% of EU firms were using at least one AI technology.
Still the early phase: share of EU firms (10 or more staff) using at least one AI technology, 2023 vs 2025
20232025
8%20%
EU firms (10+ staff) using AI
Adoption is climbing steadily, roughly doubling in two years: 8.1% in 2023, 13.5% in 2024, 20.0% in 2025, with large firms in front at 55%. But the level is the point. A firm counts if any single team uses any single AI tool, so this measures how widely AI has arrived, not how deeply it has changed the work, and even by that generous yardstick four in five EU firms had not started by 2025. The late adopters are only now beginning to move: we are in the opening phase of the transition, not its late one.
Source: Eurostat, dataset isoc_eb_ai, enterprises with 10 or more persons employed, EU27. An official probability survey run by the national statistical offices, so unlike the WEF figures this is a measurement rather than a forecast.
The importance of the skills gap is undeniable, as research shows that employees using AI are significantly more productive. In randomized experiments with working professionals, workers using AI delivered 15% to 40% more () than the control group on the tasks measured, and employers are already acting on it: two in three leaders say they would not hire someone without AI skills (Microsoft & LinkedIn, 2024).
This is what makes freelancing so attractive during the transition: hiring an AI power-user as a freelancer for a couple of weeks, or months, helps the company stay competitive, while at the same time reducing the risk of hiring the wrong person, at a time when the talent needs keep changing.
The evidence: workers using AI deliver 15% to 40% more on measured tasks, the less experienced gain the most, and employers now hire for the skill
The cleanest evidence comes from experiments: give a randomly chosen group of workers an AI tool, keep comparable colleagues as the control group, and measure real output. Four well-known ones, spanning writing, coding, consulting and customer support:
How much more the AI group delivered than the control group
Professional writing tasks¹40%
Developers, day-to-day work at three firms²26%
Knowledge work³25%
Customer-support agents⁴15%
Each bar compares workers given an AI tool with a randomized or matched control group doing the same work, in the study’s own headline measure; the measures differ and are given below, and every gain is specific to the tasks studied.
More about each experiment
¹ Noy and Zhang, Science (2023): 453 professionals, incentivised writing tasks; 40% less time and 18% higher quality.
² Cui et al., Management Science (2026): 4,867 developers at Microsoft, Accenture and a Fortune-100 firm during normal work; 26% more completed tasks, with a wide error band.
³ Dell’Acqua et al., Organization Science (2026): 758 BCG consultants; 25.1% faster, with significantly improved quality on in-scope tasks.
⁴ Brynjolfsson, Li and Raymond, Quarterly Journal of Economics (2025): 5,172 customer-support agents; 15% more issues resolved per hour on average.
Executive insights
Two of the world’s biggest software companies put their own numbers on it:
“How much has our engineering velocity increased as a company due to AI? … Our estimates are that number is now at 10%.”
Sundar Pichai, CEO of Google, describing a number Google measures rigorously, on the Lex Fridman podcast (June 2025); the ellipsis trims conversational filler.
“We’re seeing 30% productivity increase on engineering, and we’re going to really continue to ride that up.”
Marc Benioff, CEO of Salesforce, on the company’s Q4 2025 earnings call (February 2025), where he also said Salesforce would not hire any new engineers that year.

2 · Supply force

Growing preference for self-employment, especially since the unprecedented advancements in AI

Understanding why people want to work for themselves has been measured for decades, and the answer barely moves: personal independence, self-fulfilment, freedom to choose where and when to work are at the top of the list.

The evidence: autonomy is the motive, and the young hold the wish most

The sources here are the EU’s own preference survey, retired in 2012, and the 50,000-person German probability survey that is its closest living successor. Together they support the first half of the story: the wish is old, autonomy-driven, and strongest among the young.

Insight #1

Flash Eurobarometer 354, 2012EU-wide official survey, ~27,000 adults

Ask people why they would prefer self-employment: 62% say personal independence, followed by flexibility to choose when & where to work.

“Why would you prefer to be self-employed rather than an employee?” (EU27, share naming each reason, 2012)

Personal independence & self-fulfilment62%

Freedom to choose place & time of work30%

Better income prospects16%

The hierarchy was the same in 2009 (68%, 35% and 20%), so this is not a crisis-year quirk: the attraction of working for yourself is autonomy, not money.

Source: Flash Eurobarometer 354, “Entrepreneurship in the EU and beyond” (2012), question Q6, asked of respondents who would prefer self-employment (EU27 base 10,082); multiple answers allowed, so shares do not sum to 100. Figures are the EU27 row of the Q6 trend table (report pp. 137–138; summary p. 134). The related employee-or-self-employed preference question (Q1) ran from 2000 to 2012 and was then discontinued.

Insight #2

KfW-Gründungsmonitor, 2026Germany, ~50,000-person probability survey

36% of under-thirties would choose self-employment, against 20% of the over-fifties.

Share who would choose self-employment whatever their situation, by age (Germany, 2025)

Age 18–2936%

Age 30–3932%

Age 40–4923%

Age 50–6720%

The gradient is the point: the wish exists at every age, but the younger the group, the stronger it is, with the under-thirties nearly double the over-fifties. The overall average has recovered from its 23% pandemic trough to 26%, still short of 2017’s 28%.

Source: KfW-Gründungsmonitor 2026, an annual probability survey of about 50,000 people and the closest live successor to the discontinued EU preference question. “Gründergeist”: the share who would choose self-employment whatever their situation.

Side noteThis wish is even more prominent in Greece than in almost any other European country, possibly because of the economic situation (low salaries combined with long work-hours).

Every time the EU asked, Greeks preferred self-employment more than almost anyone else in Europe

“Would you prefer to be an employee or self-employed?”: share choosing self-employment, Greece vs the EU average, 2000 to 2012

EU averageGreece

51.4%70.2%

44.6%52%

45.1%60.3%

37%50%

2000200420092012

Greece sat above the EU average in every one of the eight waves the question was asked, at or near the top of all member states, with the gap ranging from about 3 points in the 2002 and 2003 waves to about 20 points in 2000 and 2001. The question was retired after 2012, but its one revival points the same way: a 2022 survey of 15-to-30-year-olds found 55% of young Greeks would prefer self-employment, against an EU average of 39%. Two more Greek details fit the picture: 66% of Greeks who preferred self-employment named personal independence as the reason, and part of the Greek wish is pressed by circumstance, with 28% of Greece’s self-employed saying they are there for lack of alternatives, one of the highest shares in Europe.

Sources

the Flash Eurobarometer entrepreneurship series, 2000–2012, read from the official trend tables published with Flash Eurobarometer 354. Four of the eight waves are charted (2000, 2004, 2009, 2012); Greece was above the EU average in all eight. National samples are roughly 500 to 1,000 people per wave, so single-wave swings carry noise, and the EU aggregate widens over the series (EU15 in 2000, EU25 in 2004, EU27 from 2009). Levels everywhere were lower in 2012 than in 2000; what persists is the gap, not the level. The 2022 youth figures are Flash Eurobarometer 513 (fieldwork October 2022, ages 15–30: Greece 54.9%, EU27 38.7%); the 66% is Flash Eurobarometer 354’s motive question; the 28% lack-of-alternatives share is Eurofound, “Exploring self-employment in the European Union” (2017, from the 2015 European Working Conditions Survey).

The desire to be self-employed has existed for years. However, what is new is how strongly people have started acting on it, and how safe doing so has begun to feel. In one of the longest-running surveys of independent workers, the share who feel more secure working independently than in a traditional job has doubled, from 32% to 67% (MBO Partners, 2025). This insight is reinforced by the actual number of people starting their own businesses: Germany measures this through a 50,000-person annual survey, and counted 690,000 new founders in 2025, up 18% in a single year, the sharpest annual rise the survey has measured since it began in 2002, with the under-thirties supplying four in ten of them.

Working as an independent has felt safer every year for a decade, and Germany’s founder numbers just made their sharpest jump since 2002

Two series, one for the mindset and one for the behaviour: a fourteen-year survey of how safe independent work feels, and Germany’s 50,000-person count of who actually starts. They point the same way: the felt risk of independence keeps falling, and the entries have risen every year since 2022.

Insight #1

MBO Partners, 2011–2025Annual US survey, same instrument for 14 years

The confidence keeps growing: the share of independents who feel more secure than in a traditional job has doubled since 2011, and employees increasingly agree.

Working for yourself keeps looking safer: US independents who feel more secure than in a traditional job, 2011 to 2025

The same survey tracks the social normalisation: half of independents (51%, up from 14% in 2012) and 43% of traditional workers (up from 13% in 2017) now say more of their friends and family work independently. One honest caveat: MBO is a US firm serving this market, so lean on the fourteen-year direction of a stable questionnaire rather than any single level.

Source: MBO Partners, State of Independence in America 2025, trend series (annual since 2011; ~6,500 respondents, weighted to US demographics; industry-funded, which is why only the long-run direction is leaned on; the chart connects the years published in the report’s trend charts).

Insight #2

KfW-Gründungsmonitor, 2026Germany, ~50,000-person probability survey

690,000 Germans founded something in 2025, up 18% in a single year, the sharpest annual rise since the survey began in 2002.

People who started something of their own in Germany, per year, in thousands (full-time and side-business founders combined), 2018 to 2025

From 2018 to 2024 the count moved sideways, between roughly 540,000 and 610,000 new founders a year. Then 2025 broke the pattern: 690,000 founders, the sharpest annual increase since the survey began in 2002 and the highest count since 2015, with each of the last three years higher than the one before. The composition changed too: a record 70% founded side-businesses alongside a job, and the under-thirties supplied four founders in ten. The turn is not only German: the country’s early-stage entrepreneurship rate hit an all-time high in the 26 years it has been measured, and across Europe the share of adults intending to start within three years rose about two points between 2022 and 2025.

Sources

KfW-Gründungsmonitor, the ~50,000-person survey behind the preference figures above; a founder is anyone aged 18–64 who started a commercial or freelance self-employment, full-time or as a side business, within the 12 months before the interview. The full series goes back to 2002, and its level then was about twice today’s: roughly 1.5 million founders a year in the early 2000s, just after the dot-com boom, followed by two decades of decline as Germany’s labour market strengthened and founding out of necessity fell away; the chart starts in 2018 to keep the recent turn legible. Mobile-phone sampling was added in 2010 and online interviews in 2024–25; revisions between editions stay within 1%. GEM Germany 2025/26 for the early-stage entrepreneurship record; the European intention figure is the Global Entrepreneurship Monitor panel, an online panel that tends to inflate levels, so it carries direction, not size.

This trend is not unique to survey evidence, but is also reflected in official statistics. The United States set an all-time record for new business applications in 2025 and 2026 is running ahead of it. Similarly, EU business registrations stand at the highest level of their series, and the world’s largest entrepreneurship survey reports, in its own words, that “more people agree that it is becoming easier to start a business”.

The evidence: new-business numbers set records on both sides of the Atlantic, and starting keeps getting easier

Two official statistics and the world’s largest entrepreneurship survey, all pointing the same way: a record number of people are starting their own businesses, and starting keeps getting easier. It is worth keeping in mind that becoming a freelancer, offering your expertise as a service to clients, is similar in many aspects to becoming an entrepreneur.

Insight #1

US Census Bureau, 2004–2026Business Formation Statistics, official count

2025 was the biggest year for new business applications the United States has ever recorded, and 2026 has started faster still.

New business applications filed in the United States, per year, in millions

Every year since 2021 has run above five million applications, against 3.5 million in 2019, and the first half of 2026 is the highest first half ever, 12% above the same months of 2025. Two honest notes: an application is not yet a business, and the subset most likely to become employer firms peaked in 2023, so the record is being driven by the one-person, no-employee kind, which is exactly the segment this essay is about.

Source: US Census Bureau, Business Formation Statistics, total business applications, annual sums of the monthly series, retrieved 23 July 2026. The series begins in mid-2004 (2005 is its first full year); every year from 2005 to 2019 sits between 2.4 and 3.5 million, so no pre-pandemic year comes close to the 5-million-plus levels since 2021.

Insight #2

Eurostat & GEM, 2024–2026Official EU statistics; 51-economy survey

EU business registrations are at the highest level of their series, and across countries more people agree that starting a business is getting easier.

New business registrations in the EU, quarterly, index (2021 = 100): at the highest level since Q1 2019

Eurostat’s registrations index reached its highest level in late 2025 and held it in early 2026, “the highest level since Q1 2019” in Eurostat’s own words. The Global Entrepreneurship Monitor’s 2024/25 report puts the ease trend plainly: “more people agree that it is becoming easier to start a business”, with the number of surveyed economies where over two adults in five call starting easy rising from 29 of 50 in 2019 to 35 of 51 in 2024; an industry survey adds that 60% of 2025 founders used AI to help launch, up from 21% in 2023. The honest edges: the same GEM series finds fear of failure rising, with the share of economies where a majority of would-be founders are deterred nearly doubling since 2019; EU bankruptcy declarations sit at their own post-2019 high; and Greece is the exception that echoes the side note above, ranking 47th of 51 on finding it easy to start and 7th on fear of failure. The door is open; it does not feel open everywhere yet.

Sources

Eurostat, quarterly business registrations (index, 2021 = 100, seasonally adjusted; an EU-wide absolute count does not exist because national practices differ); GEM 2024/25 Global Report and GEM 2025/26 Global Report (population surveys across ~50 economies; Greece figures are its 2024 wave, the last it participated in); the AI-launch figures are Gusto’s 2026 New Business Formation Report, a payroll provider’s survey of ~1,000 US founders, a vendor source carried for colour rather than proof.

Executive insights

Executives agree that it is the best time in history to become an entrepreneur:

“This is the golden age of entrepreneurship. It is the best time in the history of the world to start a business. … For the first time in the history of entrepreneurship the cost of failure is almost zero.”

Harley Finkelstein, President of Shopify, on the Aspire podcast (December 2025). In the same conversation he warned against glamourising it: “It’s really hard.”

“Let’s at least say it’s the best time yet.”

Sam Altman, CEO of OpenAI, asked whether this is the best time ever to start a technology company, interviewed by Y Combinator (November 2024). He has also described a betting pool among tech CEOs on the first one-person billion-dollar company.

So the supply side of the shift is not a new fashion. It is an old, deep wish for autonomy, held most strongly by the younger age groups, that more people are finally acting on (possibly due to AI), just as the demand side described above starts looking for precisely this kind of flexible, up-to-date skill.

Conclusion

What the evidence points to: freelancing becoming a mainstream way of working through the AI transition phase.

To recap, I set out to explain one shift and why it should last: freelancing is becoming mainstream, and is expected to keep growing through the AI transition. The labour market reached this point in three phases: a lifetime at one employer, then decades of switching jobs more and more often, and now a growing share of skilled people working for themselves.

The question for the next decade is why this should continue, and possibly accelerate. To answer it, I pointed to two forces that sit on opposite sides of the job market:

Demand. AI keeps reshaping the skills companies need, and it does so constantly, as every few months a new model resets what the technology can do. The skills gap is large and needs to be filled: 86% of employers call AI the single biggest force reshaping their business and expect about 39% of core skills to change by 2030. Despite that, only one EU firm in five was using any AI by 2025, and when asked why not, 70% of them said it was because they lacked in-house expertise. Since workers who use AI deliver 15% to 40% more on the tasks measured, bringing in a flexible, AI-skilled freelancer for a few weeks or months lets a company adapt and lower its risk while the skills it needs keep shifting.
Supply. The wish to work independently is old and stable, driven by the desire for autonomy rather than pay, and it runs strongest among the younger age groups. What is new is that people are now acting on it, and that doing so now feels safe: the share of independents who feel more secure on their own than in a traditional job has doubled, from 32% to 67%. The behaviour follows the mood, with 690,000 new founders in Germany in 2025, up 18% in a single year, and an all-time record for new business applications in the United States.

Taken together, the two forces fit each other. The demand side is looking for precisely the flexible, up-to-date skill that the supply side is increasingly willing to offer.

In fact, we are still early in this transition phase. Most of the demand described here has not arrived yet, as the data shows that only one in five EU firms were using any AI as of 2025. Similarly, on the supply side confidence and founder counts are still rising rather than peaking, and AI adoption is growing across domains.

However, what remains unclear is whether the structural shift we are seeing in the labour market will be permanent once the AI transition phase is over, and we reach a new steady-state, or whether the growth of freelancing is a temporary phenomenon that will eventually revert to the old status quo. The answer to this question will depend on how the labour market evolves in the coming years, and whether the forces driving the shift towards freelancing continue to strengthen or weaken over time.

Written by

Panagiotis Belias

Founder of Hellenic Talent · ex-BCG · London School of Economics

Get started

Find the right talent.
Or become one.

For clients

Schedule a free call

Tell us about your project needs and we will share pre-vetted elite freelancers that meet your needs, typically within 2 days.

Get started

For freelancers

Join our network

Apply to join our network of top-tier freelancers and experts, and get matched with projects and clients.

Apply now

Recent articles

A charcoal coat draped over a leather desk chair beside a briefcase, a brass lamp and a stack of bound reports, in an office overlooking a whitewashed Aegean harbour town at dusk

What is a fractional CFO? A guide for Greece and Cyprus

A fractional CFO gives a company real finance leadership a few days a month. The model is growing fast in the US and across Europe. Here is the data, country by country, where Greece stands, and how to use it well.

9 min readUpdated November 2026

Four plain, worn paper folders fanned across a dark walnut table beside a navy leather portfolio closed with a red wax seal

How is Hellenic Talent different from other players in the space?

Open marketplaces, global talent networks, search and staffing firms, and job platforms all solve part of the problem. Here is the full comparison, and what we do differently.

6 min readUpdated September 2026

Two leather armchairs facing each other in a quiet, lamp-lit private interview room

How does Hellenic Talent screen candidates before admitting them to its elite network?

Every elite member on Hellenic Talent is interviewed before joining. Here is exactly how the vetting works, why it matters, and what it filters out.

6 min readUpdated September 2026

The 3 phases of the labour market

Phase 1: A lifetime at one employer, climbing a single corporate ladder

The evidence: workers once chose the secure job over the better-paid one, while nowadays, job-security ranks behind balance and pay

Phase 2: Switching employers becomes easier, and more frequent

The evidence: the long single-employer career becomes much rarer, and at any given age under fifty, the typical job tenure is shorter

Phase 3: A growing share of skilled workers choose to work independently, as a freelancer.

“Skilled solo self-employed”: working for yourself, employing nobody, in a skilled occupation

A “professional on their own account”

Only two occupation groups grew, and professionals grew fastest: from 3.46 to 4.95 million, up 43%

Two “forces” are driving the growth of freelancing: one on each side of the job market

Artificial Intelligence is reshaping the talent companies need, constantly

The evidence: state-of-the-art AI models keep getting better, constantly exceeding expectations of what AI can do

The data shows: employers expect AI to reshape their business, and admit they currently lack the expertise to adopt AI technologies.

The evidence: workers using AI deliver 15% to 40% more on measured tasks, the less experienced gain the most, and employers now hire for the skill

Growing preference for self-employment, especially since the unprecedented advancements in AI

The evidence: autonomy is the motive, and the young hold the wish most

Working as an independent has felt safer every year for a decade, and Germany’s founder numbers just made their sharpest jump since 2002

The evidence: new-business numbers set records on both sides of the Atlantic, and starting keeps getting easier

What the evidence points to: freelancing becoming a mainstream way of working through the AI transition phase.

Find the right talent.Or become one.

Schedule a free call

Join our network

Recent articles

What is a fractional CFO? A guide for Greece and Cyprus

How is Hellenic Talent different from other players in the space?

How does Hellenic Talent screen candidates before admitting them to its elite network?

Find the right talent.
Or become one.