Predictions about how AI will reshape the labor market are multiplying, but they still rest on fragile ground. A recent line of research from GovAI and the Brookings Institution takes a slightly more structured approach. Instead of only asking which jobs are exposed to AI, it also looks at how easily workers in those roles could move into new ones. The result is less dramatic than many headlines suggest, but more useful.
The core finding is counterintuitive. Many workers whose jobs overlap heavily with AI capabilities may also be the ones best positioned to adapt. Skills that translate well across roles, higher levels of education, and access to dense job markets all increase the chances of recovery. In other words, exposure does not automatically mean displacement, and displacement does not necessarily mean long-term unemployment.
That said, the picture is uneven. Some occupations sit in a more precarious position, especially those combining high exposure with limited mobility. Clerical and administrative roles stand out here. These jobs often involve structured, repeatable tasks that AI can handle efficiently, while also offering fewer pathways into adjacent, higher-paying work. The result is a smaller but more vulnerable group of workers who may struggle to reposition themselves.
There is also a clear demographic dimension. A large share of the most exposed and least adaptable roles are held by women, reflecting long-standing patterns in the labor market. This suggests that the impact of AI will not be evenly distributed, and that existing inequalities could deepen if transitions are not managed carefully.
Despite the growing volume of research, the broader signal remains uncertain. Studies frequently contradict each other. Some suggest that AI is already reducing opportunities for younger workers in fields like software development and customer service. Others find the opposite, with those same workers performing better than peers in less AI-exposed sectors. At the same time, institutions such as regional Federal Reserve banks express skepticism about near-term job losses, while executives continue to warn of large-scale disruption.
This inconsistency is not surprising. The historical record shows that economists have struggled to forecast the labor effects of major technologies. Earlier predictions about automation often missed both the speed and the direction of change. ATMs did not eliminate bank tellers. Early AI did not erase radiology. New tools tend to remove certain tasks, reshape others, and create entirely new categories of work that are difficult to anticipate in advance.
What does hold across past transitions is the pattern of uneven adjustment. Some workers are able to move quickly into new roles. Others experience longer periods of disruption, lower wages, or exit the workforce entirely. The story of telephone switchboard operators illustrates both sides. Many workers struggled after losing their jobs, but new opportunities emerged within a few years in adjacent fields. The system rebalanced, but not without cost.
The current wave of AI appears to follow a similar trajectory. There is little evidence so far that it is reducing overall employment at a large scale. What is changing is the type of work that is most exposed. Unlike earlier waves of automation that targeted manual labor, this shift is focused more on white collar roles. Tasks involving writing, analysis, and coordination are now within reach of machines.
The more useful question is not which jobs will disappear, but which workers will have options when their roles change. Adaptability depends on more than technical skill. It includes access to education, geographic mobility, financial stability, and the ability to navigate transitions. Workers who lack these buffers face a more difficult path, even if their jobs are not the most exposed on paper.
For policymakers and businesses, this creates a practical challenge. It is easier to focus on broad narratives about job loss or job creation than to address the uneven distribution of risk. The workers most likely to struggle are often the least visible in public debates. Without targeted support, they are also the least likely to benefit from whatever new opportunities emerge.
The main takeaway is restraint. Forecasts about AI and jobs should be taken seriously, but not treated as definitive. The direction of change is real, but the scale and timing remain uncertain. Past technology shifts did not eliminate work. They reorganized it. There is little reason to assume this one will be fundamentally different.
The labor market will adjust, as it always has. The open question is how smooth that adjustment will be, and who will be left behind in the process.
