Gig economy: highs and lows

In recent years I’ve also worked as a contributor to AI training, the industry term for helping build and refine artificial intelligence models. I only do it for one selected company, with rates that are decently humane, that pays out bonuses when they’re genuinely needed and treats its workforce like people, not numbers to plug into a spreadsheet. That’s an exception, and I know it well: just look at what happens across the rest of the industry to see how rare it really is.

The gig economy promises autonomy, flexible hours, the chance to pick projects that match your skills. For a minority of professionals that promise holds up. For most people, it means juggling dozens of underpaid microtasks just hoping to make it to the end of the month, or at least to the end of a pizza night with friends.

What the gig economy looks like today, in numbers and platforms

The term gig economy covers work organized through digital platforms, usually built around single assignments, the “gigs” themselves, rather than an ongoing employment relationship. According to the World Bank report “Working Without Borders,” online gig work today involves between 154 and 435 million people worldwide, depending on whether marginal activities are counted alongside more continuous ones. In the United States, Upwork’s Freelance Forward report estimates that 64 million Americans freelanced in 2023, 38% of the total workforce.

In Italy the phenomenon is smaller but growing: early Inapp surveys counted over 210,000 platform workers, 42% of them without any contract at all; more recent research from independent centers puts the figure closer to 570,000 people, spanning riders, drivers and digital contributors. The market is already worth hundreds of billions of dollars globally and grows by double digits every year, driven mainly by markets like India, Australia and China.

Generational differences: who chooses gig work

Not every generation approaches the gig economy the same way. According to a 2024 Upwork study, 52% of Gen Z professionals have freelanced, compared with 44% of Millennials, 30% of Gen X and 26% of Baby Boomers. For those born after 1997, project-based work is no longer a fallback option but often a deliberate choice: 70% cite schedule flexibility as their main motivation, 64% want an environment free from discrimination based on age, gender or background, 62% want to work on something they consider genuinely meaningful.

There’s an interesting tech angle too: 61% of Gen Z freelancers already use generative AI in their day-to-day work, compared with 41% of peers in traditional employment. This is a generation that doesn’t experience AI as a threat but handles it as a tool, and that’s partly why it gravitates toward jobs like model training. Millennials remain the largest group in absolute numbers, often out of economic necessity rather than choice; Baby Boomers, when they enter the sector, mostly do it to supplement their pension.

When gigs turn into scraps: the microtask problem

The clearest risk in the gig economy is extreme project fragmentation. People sign up for a platform expecting to earn a bit extra, maybe just enough for a treat, and end up chasing micro-assignments worth a few cents each, with no continuity and no room to grow.

A study by researchers from Northeastern University, Microsoft Research and Mexico’s National Autonomous University, published at the CHI conference and tracking more than 40,000 tasks completed by 100 Amazon Mechanical Turk workers, calculated a median hourly wage of just $3.76, dropping to $2.83 once invisible work like searching for tasks is factored in. That’s roughly 39.5% of the US federal minimum wage.

The picture isn’t any better on the protections front. Fairwork, a project run by the Oxford Internet Institute that rates digital labor platforms on criteria like fair pay, contract terms and collective representation, reviewed fifteen platforms operating in the UK and found that only three guaranteed workers the minimum wage. Thirteen out of fifteen, moreover, offered no form of union representation at all. Even the International Labour Organization, which in 2025 opened negotiations for an international convention on platform work, acknowledges that most of these jobs still lack the basic protections that come with standard employment.

The dark side of AI training: the Kenya case

The starkest paradox shows up precisely in AI training, the field I work in. A TIME investigation documented how OpenAI, through contractor Sama, had Kenyan workers label extremely violent content to train ChatGPT’s safety filters: descriptions of child abuse, sexual violence, torture. The workers were paid between $1.32 and $2 an hour, while Sama billed OpenAI roughly $12.50 an hour per person, up to nine times as much. One of them told TIME: “That was torture. You will read a number of statements like that all through the week. By the time it gets to Friday, you are disturbed from thinking through that picture.” The project shut down eight months ahead of schedule, but the psychological toll on those involved remains well documented.

The Data Workers’ Inquiry, coordinated by researcher Milagros Miceli and the DAIR Institute, has spent years collecting firsthand accounts from annotators in Kenya, Venezuela, Syria, Argentina and elsewhere, giving voice to a category of workers who remain largely invisible to the public that uses, every day, the tools they help train.

My experience, and that of others who got it right

Not every AI training experience looks like the one above, thankfully. The company I work with pays rates that let people live decently, hands out bonuses when workload spikes or a project demands extra effort, and treats the people who work for it the way you’d treat a colleague, not an anonymous vendor. In an industry used to treating people like rows in a spreadsheet, that kind of relationship changes everything.

In the interest of honesty: my role is more managerial than hands-on, mostly supervision and review rather than annotating content one piece after another for hours on end. My better pay likely has something to do with that position, and I can’t claim that people doing the more manual, repetitive work have it as good as I do.

Some recent testimonials from people in the same field confirm that, when the terms are fair, AI training can become a solid source of income. One woman from Florida said she earns between $35 and $50 an hour on a specialized training platform, now her main source of income. A former chemistry teacher reported earning around $15,000 in a year working 15 to 20 hours a week on advanced science projects. Those numbers are worlds away from what was documented in Kenya, and the gap almost always comes down to the same variable: how seriously the company commissioning the work takes its responsibilities.

Finding your way through, between risks and real opportunities

The gig economy isn’t a single, uniform thing, it’s a patchwork of very different realities, and the generation coming into it today knows that better than the ones before: people research, compare platforms and swap advice in online communities before taking on an assignment. Still, there’s an underlying problem that touches the whole sector, not just AI training: most jobs are designed with lower cost-of-living countries in mind, and people living in Western economies struggle to find genuinely sustainable terms, unless, as in my case, they run into one of the exceptions.

Anyone weighing whether to get into this world would do well to check a platform’s reputation, how transparent its rates are and whether it offers any real support, before signing up expecting to pad their paycheck and instead ending up chasing a few euros per task. I wrote about this before when reflecting on food delivery, a blessing and a curse, another field where the gig economy shows its most contradictory side, as well as in the piece on AI and the labor market. The generational differences at work covered in another post on the blog also help explain why Gen Z and Millennials approach gig work with such different expectations. And for anyone starting out in freelancing more broadly, it’s worth reading the tips on freelancing and networking.