The more commodified your job, the more likely AI can do it – lessons from online freelancing

Not long ago, if you needed a speech polished, a document translated or a logo designed, you would probably have hired a freelancer online. Millions of people did exactly that. They went to platforms such as Fiverr and Upwork and paid someone (maybe on the other side of the world) to do the job.

In 2023, online gig workers^[1] were estimated to number between 154 million and 435 million^[2] globally. As such, they could represent as much as 12.5%^[3] of the global labour force.

Today, however, many people do something else. They open ChatGPT. Generative AI now acts as a copy editor, translator, illustrator and research assistant in one. It can summarise a report in seconds, write social media posts, create a presentation or produce a simple logo at virtually no cost.

What, then, has happened to the freelancers who used to do this work?

AI has long been discussed as a threat to jobs and livelihoods. But what’s the reality? In this new series^[4], we explore the impact it is already having on different occupations – and how people really feel about their AI assistants. Some freelancers are struggling. But perhaps surprisingly, others are doing better than ever. Demand and wages have fallen^[5] for some kinds of online freelance work. Translation, basic copywriting and simple graphic design have been hardest hit. According to one study^[6], demand for freelance writers was found to have fallen by up to 30% after the release of generative AI tools. Other research^[7] suggests that freelancers who are highly exposed to AI saw earnings fall by as much as 14%. Yet there is also evidence that many freelancers are thriving. Freelancer platform Upwork reports that higher-value contracts – those worth more than US$1,000 (£745) – increased across various disciplines^[8] after the arrival of generative AI. Freelancers using AI-related skills earn around 40% more^[9] than comparable freelancers who do not. How can both of these things be true? The answer becomes clearer when you stop thinking about “freelancers” as one group and instead look at the tasks and skills they perform. Some kinds of freelance work are highly commodified. They consist of narrowly defined, repetitive tasks that can be clearly described and easily compared. This could be things like translating a document, summarising a report, drafting a press release or designing a basic logo. These tasks are exactly what generative AI is good at. They rely on patterns, templates and predictable instructions. The more closely a freelancer’s work resembles the tasks that AI can perform, the more likely it is to come under pressure. But other freelancers do not sell a single narrow skill. They sell a more complex bundle of expertise^[10]. A legal translator does not merely convert words from one language to another. They understand legal terminology, cultural nuance and the risks of getting a phrase wrong. Similarly, a branding consultant combines design with market research and consumer psychology. A software developer may use AI to generate code, but still needs to understand the client’s business problem to decide which solution actually works. These workers can use AI to automate the repetitive parts of their jobs while concentrating on the aspects that clients still value most: expertise, judgment and trust. This matters far beyond online freelancing platforms. Online labour markets often act as an early warning system^[11] for the wider economy. This work is more transactional and less protected by the institutions that shape conventional employment (things like long-term contracts, internal promotion ladders and unions). Because tasks are posted, bought and completed on the open market, technological change shows up there more quickly than in ordinary workplaces. What happens on Fiverr or Upwork today may happen in offices tomorrow. This is already becoming visible in law firms, consultancy companies and marketing agencies. Many junior employees spend much of their time summarising documents, preparing presentations, drafting reports or conducting basic research. These are precisely the kinds of tasks that AI can perform. Recent evidence^[12] from the US labour market suggests that younger and less-experienced workers are already bearing the brunt of AI-related disruption. Senior workers, by contrast, tend to do more complex work, combining technical knowledge with experience and human interaction.

The response should not be to compete with AI at the things AI already does well. Instead, workers need help building deeper forms of expertise and combining skills in ways that are harder to automate.

This is in the interest of workers, but also of the platforms themselves. Fiverr, Upwork and others promise clients efficient and high-quality work. If routine tasks are increasingly automated away, they will depend more heavily on workers who can offer something more than a standardised service.

That means platforms should actively provide skill-building courses, training resources and guidance on how to use AI productively. They could also offer micro-credentials that certify newly acquired expertise. These credentials have been found to help workers enter online labour markets and increase their earnings^[14].

The challenge, then, is not to stop people from using AI. It is to ensure that workers are not trapped in forms of work that are so narrow, standardised and commodified that they can easily be automated away. The future of online (and onsite) work may depend less on whether we use AI than on whether our jobs can be reduced to something an AI can easily imitate.

References

1. ^{^} gig workers (theconversation.com)
2. ^{^} 154 million and 435 million (doi.org)
3. ^{^} as much as 12.5% (open-research-europe.ec.europa.eu)
4. ^{^} this new series (theconversation.com)
5. ^{^} have fallen (www.museumsassociation.org)
6. ^{^} one study (pubsonline.informs.org)
7. ^{^} Other research (pubsonline.informs.org)
8. ^{^} across various disciplines (www.upwork.com)
9. ^{^} around 40% more (www.upwork.com)
10. ^{^} complex bundle of expertise (www.sciencedirect.com)
11. ^{^} early warning system (journals.sagepub.com)
12. ^{^} Recent evidence (digitaleconomy.stanford.edu)
13. ^{^} CrizzyStudio/Shutterstock (www.shutterstock.com)
14. ^{^} increase their earnings (jhr.uwpress.org)