Analysis: The automation of human jobs in the 2030s

The story of how machines change work has long been told in two very different ways. On one side, economists often look at the past to guess the future, suggesting that new technologies will slowly reshape jobs, making some tasks easier but rarely eliminating entire professions overnight. On the other side, the creators of the most advanced artificial intelligence systems predict a much faster, more dramatic shift, where machines capable of thinking and acting like humans could take over most cognitive work within years rather than decades. This gap between cautious economic models and bold technological forecasts leaves a difficult question unanswered: when will we actually see these changes, and how big will they be? To answer this, researchers need a way to measure not just what a machine can do in a lab, but whether it can do everything a specific human job requires, from solving complex problems to interacting with people.

A team of researchers from the Organisation for Economic Co-operation and Development has tackled this problem by building a new set of tools to track the progress of artificial intelligence against the full range of human abilities. Instead of trying to count individual tasks, which is nearly impossible to measure in real time, they focused on whole occupations. They created a scale that rates AI capabilities across nine different areas, such as language, social interaction, problem-solving, and physical manipulation, ranging from basic, solved functions to full human equivalence. By comparing these AI ratings against the demands of nearly nine hundred different jobs, they identified which occupations could be fully automated once the technology catches up. Their analysis suggests that while the changes are not yet visible in the broad economy, we are standing on the edge of a rapid transformation where the share of jobs that can be completely replaced by machines could grow from a tiny fraction today to nearly all employment within a decade.

The researchers began by acknowledging that the standard way of studying work often misses the most important signal. Traditional models look at how technology changes the mix of tasks within a job, but because we cannot easily measure how often a specific task happens, these models struggle to detect the arrival of artificial intelligence. Instead, the team looked at two places where data actually exists: the displacement of entire occupations and the resulting changes in the overall economy. They reasoned that even if AI is already changing tasks behind the scenes, the first clear sign of this shift will be when a machine becomes capable enough to replace a whole job title entirely. Only after that happens on a large scale will we see the broader economic effects, such as a surge in productivity or a drop in prices.

To make this measurement, the team developed a set of indicators that describe AI progress in terms anyone can follow. They defined nine domains of human ability, including language, social interaction, creativity, and physical manipulation, and created a five-level scale for each. Level one represents capabilities that have been solved for a long time, while level five represents performance that matches every aspect of human ability. In their initial assessment, current AI systems sat mostly between level two and level three across these domains. To get a clearer picture of where things are heading, they asked two of the most advanced AI systems available to re-evaluate their own progress based on recent research. This preliminary update suggested that in just eighteen months, AI had made substantial leaps, particularly in problem-solving and language, moving closer to the level where it could handle the full demands of many professional roles.

The researchers then mapped these capability levels against the requirements of nearly nine hundred occupations. They found that today, only about one percent of the workforce is in jobs that current AI can already fully automate. These are mostly routine office and administrative roles. However, the picture changes dramatically when they looked at the capabilities that are just becoming available. With the first wave of new capabilities, the share of automatable jobs jumps to about eight percent, covering parts of sales, finance, and computer work. The real shift, however, comes with the next two waves of projected progress. If AI continues to advance at the pace seen recently, the second wave could make about one-third of all jobs fully automatable, and a third wave could eventually cover nearly all employment. The only jobs that would remain largely untouched are those that rely heavily on deep human connection, such as community social services and certain aspects of healthcare.

The data already shows the beginning of this trend. The researchers examined employment figures for the jobs that are already automatable and found that employment in these areas had already begun to decline. For the jobs that become automatable with the first wave of new capabilities, the decline is just starting to appear, while jobs further down the timeline remain stable or are still growing. This pattern suggests that the signal of change is arriving exactly where the researchers predicted: first in the specific occupations that machines can now take over, and only later in the overall economy. They estimate that as the share of automatable work grows, the economic effects will become historically large, potentially leading to massive increases in productivity and significant drops in prices, a combination that has no modern precedent.

This trajectory presents a profound challenge for policy and society. The researchers argue that waiting for the economy to adjust on its own is risky, because the speed of this transition could outpace the ability of workers to retrain for new roles. If machines can perform almost any job, the traditional idea that new tasks will automatically create new jobs for humans may no longer hold true. Instead, governments may need to prepare for an economy where high productivity and falling prices coincide with widespread job displacement. The study concludes that the time to act is now, using the specific list of occupations identified as automatable to guide retraining efforts and social safety nets, rather than relying on models that assume change will be slow and modest. By focusing on the jobs that are actually within reach of current and near-future technology, society can begin to prepare for a future where the definition of human work is fundamentally altered.