Purpose The purpose of this paper is to identify workforce skills that protect an occupation from elimination due to automation technology. Design/methodology/approach The authors apply a Gaussian process (GP) classifier, based on the level of non-automatable work activities in an occupation, to USA and Colombian occupational datasets. Findings The authors find that communication, interpersonal relationship management and decision-making skills are most important in occupations that are resistant to automation. Research limitations/implications The results are based on work activities data from the Occupational Information Network (O*NET) database developed for the USA labor market. This dataset does not capture significant differences in work activities, where they exist, for the same occupation between the two countries. The findings are also limited to Colombia. Readers should be careful to extrapolate the findings outside of this geography. Originality/value The authors discover that automation is likely to be a global phenomenon that can only be slightly mitigated by cultural and political factors.