Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
Complement or Substitute? How AI Increases the Demand for Human Skills
Citation. Mäkelä, E., Bone, M., Nanji, F., Sehrer, M., & Stephany, F. (2025). Complement or substitute? How AI increases the demand for human skills. SSRN. https://doi.org/10.2139/ssrn.5153230
Summary. This working paper analyzes approximately 30 million online job postings from 2018–2024 to examine how AI-related hiring is associated with demand for human capabilities that complement AI. It separates changes within AI-related roles from spillovers into non-AI roles at company, industry, and regional levels. The authors report greater demand for complementary skills and lower demand for selected substitutable skills as AI prevalence rises, alongside some advertised-wage premiums for complementary capabilities.
Key claims
- AI-related job postings are more likely than non-AI postings to request complementary non-technical capabilities, including analytical thinking, resilience, and digital literacy.
- In AI roles, complementary skills are associated with advertised-wage premiums; the paper reports a 5% premium for resilience.
- Higher AI prevalence in companies, industries, and regions is associated with increased demand for complementary skills and decreased demand for some substitutable skills, including customer service, in non-AI roles.
Evidence & limitations
- The analysis uses Lightcast/Burning Glass vacancy data: 9,998,342 US postings for the internal analysis and approximately 30 million postings across the US, UK, Australia, and New Zealand for external analyses.
- AI roles are defined by whether a posting lists at least one Lightcast AI/ML skill, and the skill labels and posting coverage depend on proprietary data-processing methods.
- The observational associations support evidence about advertised skill demand and wages, not causal estimates of AI adoption’s effect on jobs or worker capabilities; this is a working paper rather than peer-reviewed research.