When picking a skill to learn, prioritize those that require hands-on work and on-the-spot judgment — these are the hardest for automation to replace.

According to U.S. estimates, roughly 47% of jobs carry a relatively high risk of being taken over by computers. Later studies that evaluated individual tasks instead yielded a much lower percentage. Three core challenges currently prevent machines from handling certain jobs: manual dexterity, creative problem-solving, and social interaction. Hence, roles that demand physical work in unpredictable environments and real-time decision-making are far less likely to be automated, while jobs with rigid processes and heavy information processing carry the highest risk.

No cost at all. Before making a choice, run through the skills you’re considering using this criterion as a filter.

This study assessed 702 U.S. occupations to estimate the likelihood of each being automated — the result showed that about 47% fell into the high-risk category. The three aforementioned challenges form what the authors call “engineering bottlenecks” that limit automation potential. These include sensory perception and physical manipulation (finger and hand dexterity, working in cramped or awkward spaces), creative intelligence (the ability to devise novel solutions), and social intelligence (interpersonal interaction). Consequently, occupations requiring hands-on work in unstructured settings and on-the-fly judgment are consistently less prone to automation, rather than this being a random trait of specific jobs. Roles with standardized workflows and information-centric tasks face the greatest automation risk.

This research is somewhat controversial. It originated as an unpublished draft paper from 2013 that measured the overall automation potential of entire occupations. Later studies that evaluated individual tasks instead produced a significantly lower high-risk percentage. The paper also does not specify a timeline for when automation might occur. When using these findings, avoid fixating on the 47% figure; instead, focus on the three key bottlenecks it identifies. Jobs involving on-site installation, maintenance, and caregiving are far more resistant to automation than purely clerical roles for precisely these reasons. Therefore, when deciding which skill to acquire, prioritize those that demand physical work in non-standard environments.

Frey CB, Osborne MA (2017). The future of employment: how susceptible are jobs to computerisation? Technological Forecasting and Social Change, 114, 254-280. https://doi.org/10.1016/j.techfore.2016.08.019