Nova: Picture this: one in every four workers on the planet, right now, is in a job that generative AI could fundamentally reshape. Not replace, necessarily, but transform so deeply that the daily tasks they perform might look completely different within a few years. That is the headline finding from the International Labour Organization's landmark research on generative AI and the labor market.

Nova: : That number is staggering, Nova. One in four. That is roughly a billion people. But here is what grabs me: the ILO is not predicting mass unemployment. They are telling a much more nuanced story. So what is actually going on?

Nova: Exactly. The ILO, which is the United Nations agency focused on the world of work, published a major working paper in August 2023 and then followed it up with a refined global index in 2025. The research was led by economists Paweł Gmyrek, Janine Berg, and David Bescond along with collaborators from Poland's National Research Institute. And their central insight flips a lot of the scary headlines on their head.

Nova: : I have seen those headlines. Robots are coming for your job. AI will replace everyone. But you are saying the ILO sees it differently?

Nova: Yes, and the distinction matters enormously. The ILO found that the overwhelming impact of generative AI will be augmentation, not automation. In other words, AI will change how we do our jobs, take over certain tasks, but leave the human worker in place doing the things only humans can do. Only about 3.3 percent of global employment falls into what they call the highest exposure category, where automation could genuinely displace roles.

Nova: : So the real story is not about jobs vanishing. It is about jobs changing. But that raises a whole different set of questions. Who gets hit hardest? Which countries? And are we ready for this?

Nova: Those are exactly the questions the ILO research tackles, and the answers are revealing and uncomfortable. So today, we are diving deep into the ILO's findings on how generative AI is reshaping the global labor market, who stands to win and lose, and what policymakers need to do about it. Let us get into it.

Nova: Let us start with the methodology because it helps explain why the ILO's conclusions are different from a lot of the doom-and-gloom forecasts. The researchers did something clever: they used GPT-4 itself to analyze thousands of occupational tasks drawn from the International Standard Classification of Occupations, known as ISCO. They broke jobs down not by job title, but by individual tasks.

Nova: : So instead of asking, will AI replace accountants, they asked, how many of the specific tasks an accountant performs could AI handle?

Nova: Precisely. And when you look at it at the task level, the picture shifts dramatically. For most occupational groups, the share of highly exposed tasks is tiny, between 1 and 4 percent. Even medium-exposure tasks do not exceed 25 percent for the vast majority of occupations. The one glaring exception is clerical work.

Nova: : Clerical work. That is secretaries, data entry clerks, administrative assistants?

Nova: Exactly. For clerical occupations, 24 percent of tasks are considered highly exposed to generative AI, and an additional 58 percent have medium exposure. That is a combined 82 percent of clerical tasks that AI could significantly affect. No other occupational group comes close.

Nova: : That makes intuitive sense though, right? Those roles involve a lot of text processing, scheduling, data formatting, document preparation, things that large language models are genuinely good at.

Nova: Right. But here is the crucial distinction the ILO makes. Even for clerical workers, the most likely outcome is not that entire jobs disappear. It is that the nature of those jobs changes. A clerical worker might stop spending three hours a day on data entry and instead focus on more complex problem-solving, client interaction, or oversight of AI-generated outputs. That is augmentation.

Nova: : So the job title stays but the day-to-day experience is transformed. But does that also mean fewer clerical workers are needed overall?

Nova: That is the big unknown, and the ILO is honest about it. If one worker augmented by AI can do the work of two, demand for clerical labor could shrink even if individual jobs are not fully automated. The researchers identified what they call the big unknown, a gray zone between clear augmentation and clear automation, where outcomes will depend on company choices, market dynamics, and policy interventions.

Nova: : It feels like we are at a fork in the road. Companies could use AI to make workers more empowered and productive, or they could use it to squeeze out headcount while intensifying the workload on those who remain.

Nova: That tension runs through the entire report, and we will come back to it. But first, let us look at who bears the brunt of this transformation, because the ILO's findings on gender are some of the most striking in the whole study.

Nova: Here is a number that should make everyone pause. According to the ILO's 2025 refined index, 4.7 percent of women's jobs globally fall into the highest GenAI exposure category, compared to just 2.4 percent for men. That is nearly double.

Nova: : Double. And in high-income countries, the gap is even wider. I saw that 9.6 percent of female employment in wealthy nations is in the highest-risk category versus 3.5 percent for men. That is almost triple.

Nova: Yes. And the reason is straightforward but deeply rooted in structural patterns. Women are heavily overrepresented in clerical and administrative support roles. Think secretaries, receptionists, payroll clerks, accounting assistants, customer service representatives. These are the very occupations where AI exposure is highest.

Nova: : And it is not just about the occupations themselves. The ILO also published a research brief that found around 29 percent of female-dominated occupations are exposed to generative AI, compared to just 16 percent of male-dominated ones. For high automation risk specifically, it is 16 percent versus 3 percent.

Nova: That is a staggering disparity. And it compounds another problem. Women are simultaneously locked out of the AI-driven job opportunities. Globally, women account for only about 30 percent of the AI workforce, and that number has barely budged since 2016, moving only four percentage points in six years.

Nova: : So women are disproportionately exposed to the risks of AI and disproportionately excluded from the opportunities. That is a double whammy.

Nova: Exactly. Janine Berg, one of the lead authors, put it bluntly: generative AI is not entering a neutral labor market. Discriminatory social norms, unequal care responsibilities, and economic policies that do not fully address gender needs all shape who enters which occupations. AI lands on top of existing inequalities and risks amplifying them.

Nova: : And there is another layer. The ILO research brief also points out that AI systems themselves can reproduce gender biases. If you train a hiring algorithm on historical data that reflects past discrimination, the algorithm learns to discriminate. Women can be disadvantaged in recruitment, pay decisions, even credit scoring.

Nova: Right. So we have occupational segregation, underrepresentation in tech, and biased AI systems all feeding into each other. The ILO is essentially waving a red flag here: without deliberate intervention, generative AI could widen the gender gap in the labor market significantly.

Nova: : What countries see the most extreme version of this?

Nova: In several economies, more than 40 percent of women's employment is exposed to GenAI, including Switzerland, the United Kingdom, the Philippines, and several small island developing states in the Caribbean and Pacific. In 88 percent of countries analyzed, women are more exposed than men. It is nearly universal.

Nova: : So if you are a policymaker looking at this, addressing the gender dimension cannot be an afterthought. It has to be baked into the response from day one.

Nova: Let us zoom out to the global level. One of the most sobering findings in the ILO research is how differently generative AI will affect countries based on their income level.

Nova: : I saw the numbers on this and they are jarring. In high-income countries, 34 percent of jobs are in occupations exposed to GenAI. In low-income countries, that figure drops to 11 percent.

Nova: Right. And the automation potential specifically is 5.5 percent of employment in high-income countries versus just 0.4 percent in low-income countries. The reason is structural: high-income economies have a much larger share of employment in clerical, professional, and administrative occupations that AI affects. Low-income countries have larger shares of agricultural and manual labor, which are far less exposed to current generative AI capabilities.

Nova: : So at first glance, you might think low-income countries are getting off easy. Less exposure means less disruption, right?

Nova: That would be a dangerously naive reading. The ILO researchers explicitly warn against it. Lower exposure does not mean lower risk. In fact, the digital divide could make things worse for developing economies in several ways.

Nova: : How so?

Nova: First, there is the infrastructure gap. If you do not have reliable internet, electricity, or digital literacy, you cannot benefit from AI-driven productivity gains at all. The ILO notes that billions of people still lack internet access. So while high-income countries use AI to boost productivity, low-income countries risk falling further behind. The productivity gap widens.

Nova: : So it is not that AI will not affect them. It is that they will be excluded from the benefits while still facing whatever disruptions do occur.

Nova: Precisely. Second, there is the nature of work in developing economies. When labor protections are weak and digital access is limited, even small-scale automation can destabilize vulnerable sectors. A call center that automates 20 percent of its work in a country where alternative employment is scarce can cause disproportionate harm.

Nova: : And third, the augmentation potential, which is the more widespread effect, affects 10.4 percent of employment even in low-income countries. That is not trivial.

Nova: Exactly. The ILO also points out something called the classic growth path. As countries develop, their occupational structures shift toward more clerical and professional roles. So today's low-income countries will become more exposed to AI over time. You cannot just look at a snapshot and assume the risk stays static.

Nova: : So we are looking at a world where AI could entrench global inequality. High-income countries get the productivity boost and manage the transition with stronger institutions. Low-income countries get left out of the gains while still absorbing disruption in whatever formal sectors they have.

Nova: That is the concern. And it is why the ILO insists that addressing the digital divide is not a side issue. It is central to any equitable AI future. Expanding internet access, digital infrastructure, and digital skills in developing regions is not charity. It is a prerequisite for preventing AI from becoming yet another driver of global divergence.

Nova: So we have established that AI will augment more than automate, that women face disproportionate risk, and that the global playing field is deeply uneven. The question the ILO then tackles is: what do we actually do about all of this?

Nova: : And this is where it gets practical. The ILO is not just a research body. It is the UN agency that sets international labor standards. So their policy recommendations carry real weight.

Nova: Absolutely. The report outlines three major policy areas. The first is mitigating the negative effects of automation. This includes traditional measures like unemployment insurance, retraining programs, and job placement services, but adapted specifically for AI-driven transitions. The key is to act proactively, before displacement happens, not react after the fact.

Nova: : So not waiting until a factory closes or an office downsizes and then scrambling. But actually anticipating which sectors and regions will be hit and preparing the workforce in advance.

Nova: Exactly. The second, and in many ways more novel, area is ensuring job quality under augmentation. This is a huge theme in the report and one that does not get enough attention.

Nova: : Explain that. Most people probably think augmentation sounds great. AI handles the boring stuff, humans do the interesting stuff. What is the catch?

Nova: The catch is that augmentation can cut both ways. Yes, AI could eliminate drudgery and free workers for more creative, fulfilling tasks. But it could also be used to intensify work, increase monitoring, reduce autonomy, and erode wages. Think about a customer service agent who is now monitored by AI on every call, with every pause and every word scored and ranked. That is augmentation too, and it is not empowering.

Nova: : So the same technology that could make work better could also make it more stressful, more surveilled, and less fairly compensated.

Nova: Precisely. The ILO stresses that job quality effects may ultimately be of greater consequence than job quantity effects. Losing autonomy, facing algorithmic management, having work intensity ratchet up, these are the everyday realities that shape whether AI makes working life better or worse.

Nova: : And the ILO's answer to this is social dialogue?

Nova: Yes. Social dialogue, which means bringing together governments, employers, and workers' representatives to shape how AI is introduced. The report is very clear: workers must have a voice in how generative AI is deployed in their workplaces. Without it, the risks of unequal outcomes, declining job quality, and widening gender gaps multiply.

Nova: : This feels like a core tension. Companies want to deploy AI quickly to gain competitive advantage. Workers want to ensure it does not crush their quality of life. And governments are caught in the middle, trying to regulate a technology that is evolving faster than legislation can keep up.

Nova: That is exactly the challenge. The third policy area is addressing the digital divide we already discussed. Expanding infrastructure, ensuring equitable access, and investing in digital skills globally. The ILO also emphasizes that preparing workforces requires not just technical training but modernized curricula and alignment between education systems and evolving labor market needs.

Nova: : So the policy toolkit is actually quite coherent: protect people during transitions, ensure technology improves rather than degrades job quality, close the digital gap between and within countries, and do all of it through inclusive processes where workers have a real seat at the table.

Nova: That is it. The ILO is essentially arguing that the future of AI in the workplace is not technologically determined. It is a set of choices. And those choices need to be made collectively, transparently, and with gender equity and global fairness at the center.

Nova: I want to spend a moment on what the ILO researchers themselves flag as the big unknown. It is one of the most intellectually honest parts of the entire report.

Nova: : I love this concept. They basically identify a gray zone of occupations where they genuinely cannot predict whether augmentation or automation will dominate.

Nova: Right. Some occupations have task profiles that sit right on the boundary. Depending on how companies choose to deploy AI, how quickly the technology improves, and how labor markets and regulations respond, these jobs could go either way.

Nova: : Can you give a concrete example?

Nova: Think about a paralegal. Many paralegal tasks involve document review, legal research, drafting standard filings, all things that generative AI is getting quite good at. A law firm could use AI to make each paralegal dramatically more productive, handling more cases with the same headcount. Or the firm could decide that with AI doing 60 percent of the task load, it only needs half as many paralegals. The technology does not make that choice. Humans do.

Nova: : So the big unknown is really about human decisions, not technological capabilities. And that is why policy matters so much.

Nova: Exactly. Another thing keeping researchers up at night is the pace of change. The 2025 update found that AI capabilities in voice, image, and video generation have already expanded the exposure of media and web-related occupations since the 2023 report. This technology is not standing still.

Nova: : And the infrastructure constraints you mentioned earlier. The ILO is clear that their exposure estimates are upper-bound estimates. They do not account for the fact that in many countries, reliable internet, electricity, or the necessary hardware simply does not exist at scale. So the gap between theoretical exposure and actual impact may be large, but in ways that are themselves inequitable.

Nova: That is such an important point. The estimates assume ideal conditions for AI adoption. In reality, adoption will be uneven, shaped by infrastructure, regulation, corporate strategy, and cultural factors. The ILO's primary value, as they themselves state, is not the precise numbers but the insights the overall distribution provides about the nature of possible changes.

Nova: : It is a map, not a prophecy.

Nova: Beautifully put. And maps help you navigate, but they do not tell you where to go. That is up to all of us.

Nova: So let us bring this together. The International Labour Organization has given us one of the most rigorous, globally scoped analyses of what generative AI means for the world of work. And their message is nuanced, urgent, and ultimately hopeful.

Nova: : The core insight is that transformation, not replacement, is the real story. Most jobs will be augmented, not automated. Only about 3.3 percent of global employment is in the highest exposure category where outright displacement is plausible. But augmentation itself comes with profound challenges around job quality, work intensity, autonomy, and fair compensation.

Nova: The gender dimension is impossible to ignore. Women are concentrated in the clerical and administrative roles most exposed to AI, with female-dominated occupations nearly twice as likely to be affected. At the same time, women remain underrepresented in the AI and STEM fields where new opportunities are emerging. Without deliberate intervention, AI could widen existing gender gaps.

Nova: : And the global picture is starkly unequal. High-income countries face 34 percent occupational exposure. Low-income countries face 11 percent. But lower exposure is not a blessing. It reflects a digital divide that could leave developing economies excluded from AI-driven productivity gains while still absorbing disruption in their formal sectors.

Nova: The policy path forward, according to the ILO, rests on three pillars. First, proactive social protection and retraining to cushion automation's effects. Second, a deliberate focus on job quality, ensuring that augmentation empowers workers rather than intensifying their work and eroding their autonomy. Third, closing the digital divide through infrastructure investment and skills development globally.

Nova: : And running through all of it is the principle of social dialogue. Workers must have a voice. Employers, governments, and labor representatives need to sit at the same table. The choices about how AI enters the workplace cannot be made in boardrooms alone.

Nova: The ILO's research reminds us that technology does not determine its own impact. We do. The decisions made today, about regulation, about training, about who gets consulted and who gets left out, will shape whether generative AI becomes a force for broader prosperity or a driver of deeper inequality.

Nova: : That is the challenge. That is the opportunity. And according to the ILO, the window for getting it right is open, but it will not stay open forever.

Nova: This is Aibrary. Congratulations on your growth!