Freelancers are learning constantly, but often without a training budget, learning department or employer paying for the time. As skills change faster, staying current is becoming part of the cost of working independently.
Learning a new skill every few years used to sound ambitious. For many knowledge workers, six months is starting to look more realistic.
A 2025 workforce study from Upwork surveyed 3,000 skilled US knowledge workers and found that 87% of skilled freelancers had learned a new skill during the previous six months. Among full-time employees, the figure was also high at 82%. The study of 3,000 skilled knowledge workers
The five-point difference is not the most interesting part of the research.
What separates the groups is how the learning happens.
Sixty-four percent of freelancers relied on self-directed learning, while 41% used formal certification programs and 28% learned through apprenticeships or mentorships. Full-time employees were more likely to rely on formal training provided by their organization and employer-sponsored development programs.
That difference points to a larger change in professional life. Workers still need to keep learning, but the responsibility for deciding what to learn, finding the training and making time for it is increasingly moving toward the individual.
For freelancers, that responsibility already comes with the job.
The 87% figure is really about how fast skills are moving
It would be easy to frame the finding as evidence that freelancers are more motivated than employees.
The study does not establish that.
Most full-time employees in the same research had also learned something new recently. Both groups reported being comfortable with continuous learning, and 32% of skilled workers said they were picking up new skills “all the time.”
The broader labor market helps explain why.
The World Economic Forum’s 2025 Future of Jobs Report found that employers expect 39% of workers’ core skills to change by 2030. AI and big data are among the fastest-growing technical skills, but employers also expect greater demand for creative thinking, resilience, flexibility, leadership and other human capabilities. The global employer research on changing skills through 2030
LinkedIn reaches the issue from a different dataset. Its analysis suggests that by 2030, 70% of the skills used in most jobs will have changed compared with 2015, with AI accelerating the shift. Its Work Change research on how job skills are evolving
The percentages are not directly comparable because the studies use different methods and time periods. Their direction, however, is consistent. A job title can stay the same while the bundle of skills required to do that job changes significantly.
That makes continuous learning less of a career advantage and more of a maintenance requirement.
AI is making the skills cycle move faster
The pressure is particularly visible in jobs exposed to AI.
PwC’s analysis of close to a billion job advertisements found that the skills employers requested were changing 66% faster in occupations most exposed to AI than in less exposed occupations in its 2025 analysis. Its 2026 research, expanded to more than a billion postings, found that skills in the most AI-exposed jobs were changing at more than twice the rate seen in the least exposed jobs. The global analysis of how AI is accelerating skill change
That does not mean every worker needs to become an AI specialist.
In fact, some of the skills becoming more important are distinctly human. The 2026 analysis found that new tasks appearing in AI-exposed roles were 2.5 times more likely to depend on capabilities such as judgment, empathy and creativity.
That is an important distinction for freelancers.
Learning AI is not simply about becoming better at prompting a model. As automation handles more routine production, professionals may need to strengthen the skills required to review, interpret and improve what the technology produces.
Flexable’s analysis of the 70% rise in jobs fixing AI-generated work shows what that transition can look like in practice. Clients are increasingly hiring developers, designers and other specialists after AI has already produced the first version.
The skill being sold in those cases is not simply the ability to use AI. It is the ability to know when AI has produced something that is not good enough.
Freelancers do not have a training department
Full-time employees often underestimate one advantage of being inside an organization: some of the cost of staying useful is carried by the employer.
Companies buy training platforms, reimburse qualifications, run workshops, provide managers and mentors, give employees access to internal experts and sometimes allow learning to happen during paid working hours.
Across OECD countries, 65% of adults participating in non-formal job-related learning received at least some financial support from their employer, according to the organization’s latest analysis of adult learning. The OECD research on who pays for workplace learning
A freelancer usually operates under a different arrangement.
Clients generally pay for the work they need today. They rarely pay an independent professional to spend ten hours learning a tool that may be necessary for the next client.
That makes skill development a business investment.
A freelance developer learning a new framework may spend evenings working through documentation before being confident enough to sell the service. A designer may pay for software and training before discovering whether clients actually want the capability. A marketer experimenting with new AI tools may spend hours testing workflows that never become commercially useful.
There is no guarantee that the investment pays back.
This is one reason the 64% self-directed learning figure deserves more attention than the headline 87%.
It shows how much professional development is happening outside a formal training system.
The real cost of learning is often time
Training itself does not need to be expensive to carry a meaningful cost.
For someone billing clients by the hour or project, an afternoon spent learning is also an afternoon that cannot be billed.
The OECD’s 2026 work on barriers to adult learning found that among adults prevented from taking desired training, lack of time because of work was the most common barrier, affecting 29%. Family responsibilities were cited by 18%, while 13% pointed to the cost of training. The research on what prevents adults from participating in training
The International Labour Organization makes a similar point in its 2026 work on lifelong learning. Cost, time constraints, lack of information and rigid training formats continue to prevent workers from accessing learning, with people in less secure forms of employment facing particular disadvantages. The 2026 global review of lifelong learning and skills
Those studies do not isolate highly skilled freelancers, so they should not be used to estimate how many freelancers struggle to afford training.
They do show why continuous learning cannot be treated as free.
Someone pays for it in money, time or foregone work.
In a traditional job, that cost can be shared between worker and employer. In independent work, much more of it can land directly on the worker.
Freelancers are using projects as part of the learning process
There is another reason independent workers may learn differently.
The work itself changes more frequently.
A full-time employee can spend years inside the same technology stack, company systems and set of responsibilities. Freelancers often move between clients that use different tools, operate in different industries and expect different deliverables.
That variety can create a practical learning loop.
A new client requires a capability. The freelancer learns enough to perform it. The capability becomes part of the freelancer’s service offering. Another client then pays for a more advanced version of the same skill.
The original study provides some evidence for this behavior. Eighty-seven percent of skilled freelancers said they preferred work that helped them improve existing skills or learn new ones over work that simply allowed them to use skills they already had.
That suggests learning is influencing which projects freelancers choose, not simply what they do between projects.
For anyone unfamiliar with how independent careers are structured, Flexable’s guide to how freelancing works explains why workers often move across clients and projects rather than developing inside a single organization.
The arrangement can expose workers to more variety, but it also places more pressure on them to identify which skills will remain valuable.
Nobody inside the company is creating an individual development plan for them.
AI training shows the difference particularly clearly
The study found that 51% of skilled freelancers were teaching themselves how to use AI tools, compared with 39% of full-time employees.
Freelancers were also using multiple learning routes. They experimented directly with AI, pursued formal training and certifications, and reported higher levels of proficiency. Fifty-four percent described their professional AI skills as advanced or expert, while 62% said they were using AI tools several times a week or more.
Self-reported proficiency should be treated cautiously. The study did not independently test whether one group was objectively better at using AI.
What it does show is a different pattern of learning.
Freelancers cannot necessarily wait for someone to decide that AI training belongs in next quarter’s learning budget. If clients start asking for AI-related work, market demand itself becomes the training signal.
That speed can be useful.
It can also produce bad decisions. Workers can chase fashionable tools, buy certifications with little market value or spend time learning technologies that disappear before they become commercially important.
Continuous learning is not automatically good learning.
The harder skill is knowing what deserves to be learned.
The market is giving workers conflicting signals
This is one of the least discussed problems in the push for constant upskilling.
There are now more ways to learn than ever, but also more things competing to be learned.
AI literacy. Large language models. Agents. Automation. Cybersecurity. Data analysis. New programming frameworks. New design tools. New marketing channels. New certifications.
At the same time, employers continue to value broad human skills.
LinkedIn’s 2025 analysis of fast-growing skills found AI literacy appearing near the top across multiple countries, but communication, strategic thinking, adaptability, relationship building and creativity were also repeatedly prominent. The data on skills growing fastest across countries
That creates a practical allocation problem for workers.
Time spent learning one thing is time not spent learning another.
A freelancer who spends 100 hours mastering a highly specific tool needs that tool to remain useful long enough for the investment to make sense. Learning a broader skill such as negotiation, client management or strategic thinking may be harder to advertise on a profile, but could remain valuable across many technologies.
The rapid change in AI makes that tradeoff particularly difficult.
The best learning strategy may therefore be a combination of short-lived technical capabilities and slower-changing professional judgment.
That is consistent with another pattern appearing in the freelance market. Flexable’s report on why clients trust human and AI work far more than AI alone found that the human contribution increasingly centers on judgment, oversight and accountability rather than simply producing every part of the work manually.
Those capabilities are harder to make obsolete with a software update.
Companies are struggling with the same problem at a larger scale
Continuous learning is not only a worker problem.
The World Economic Forum found that 63% of employers identified skills gaps as a major barrier to business transformation, making it the most frequently cited obstacle in its 2025 employer survey.
Companies are responding. Half of workers had completed some form of training, reskilling or upskilling as part of long-term learning strategies, up from 41% in the previous edition of the research.
But organizational learning has a timing problem.
A company needs to identify a skill gap, decide whether it matters, approve a training response and deliver it across the workforce. Independent professionals can often react much faster because the decision involves one person.
That does not make freelance learning inherently better.
Organizational training can be more rigorous, better funded and better connected to long-term career development. A freelancer watching tutorials alone may learn quickly but unevenly.
The difference is responsiveness.
A freelancer who loses a potential project because a client wanted a capability they did not have receives immediate information from the market. The incentive to learn is direct.
Continuous learning can become another form of unpaid work
There is a positive way to read the 87% statistic.
Freelancers are taking control of their professional development, learning across projects and adapting quickly as the market changes.
There is also a less comfortable interpretation.
If a worker must repeatedly retrain simply to maintain the same earning power, learning stops being an optional investment and becomes another cost attached to having a career.
For an employee, some of that burden may be absorbed by the employer.
For a freelancer, it can sit beside the other costs that independent workers already carry themselves: finding clients, managing contracts, chasing payments, handling administration and absorbing gaps between projects.
Those pressures matter because independence does not automatically mean stability. Flexable’s research into why inconsistent work remains a major freelance challenge shows how much uncertainty can already sit around the actual paid work.
Adding constant reskilling to that equation changes how we should think about freelance earnings.
An hourly or project rate does not capture every hour required to remain capable of charging that rate.
There is a selection effect worth taking seriously
The research should not be interpreted as proving that freelancing makes people more adaptable.
People who choose independent work may already be more comfortable with uncertainty, self-direction and learning without formal support.
The study itself is observational and relies on self-reported behavior. It cannot tell us whether freelancing caused respondents to become more proactive learners or whether proactive learners were more likely to become freelancers in the first place.
There is another important limitation.
The respondents were skilled US-based knowledge workers, not a representative sample of every freelancer in the United States or around the world. Participants were working above the administrative level across skilled organizational functions and met an earnings threshold based on Bureau of Labor Statistics data.
A delivery driver, construction subcontractor, independent care worker and freelance machine-learning engineer can all technically be independent workers, but their access to training and exposure to changing skill demands are very different.
The 87% figure belongs to the skilled knowledge-worker population studied.
That precision makes the finding more useful, not less.
What the 87% figure actually tells us
The survey was conducted between December 2024 and February 2025 among 3,000 skilled US knowledge workers. The reported margin of error was 1.8% at the 95% confidence level. Skilled freelancers were identified through self-report and had either been freelancing at the time of the survey or had done so during the previous 12 months.
The headline finding is that 87% of those skilled freelancers said they had learned a new skill during the previous six months.
Full-time employees were not far behind at 82%.
That comparison changes the meaning of the statistic. The defining difference is not simply willingness to learn. It is the infrastructure around the learning.
Employees can draw on employers, managers and formal development programs. Freelancers are more likely to assemble their own system through self-teaching, certifications, mentorship, experimentation and the work itself.
That model offers speed and autonomy, but it also moves more responsibility and more cost onto the individual.
As the global freelance economy continues to expand, this question deserves more attention. A labor market built around constantly changing skills also needs to answer who pays for workers to acquire them.
For a growing number of independent professionals, the answer is already clear.
They do.