Elina Mäkelä and Fabian Stephany, in their SSRN working paper “Complement or Substitute? How AI Increases the Demand for Human Skills,” provide a comprehensive analysis of how artificial intelligence (AI) reshapes labour markets by examining 12 million online job postings in the U.S. from 2018–2023. Their study tackles a foundational question in the AI discourse: does AI displace human workers, or does it enhance the demand for particular human capabilities? Through a rigorous two-pronged analytic framework, they assess both “internal effects” in explicitly AI-related roles and broader “external effects” across various sectors.
In addressing internal labour-market dynamics, Mäkelä and Stephany reveal that AI-heavy roles are nearly twice as likely as non-AI roles to require complementary skills such as digital literacy, teamwork, resilience, analytical thinking, and ethics. These roles also tend to offer higher wage premiums—between 5–10%—for candidates possessing these competencies. Conversely, skills considered substitutable by AI, such as basic data processing, summarisation, and routine customer service, are declining in both demand and value within AI-centric positions.
Expanding beyond AI-dedicated roles, the authors identify notable “external effects”: industries, occupations, and regions experiencing AI growth also see rising demand for complementary skills in non-AI roles. Interestingly, these external complementary effects dwarf the negative impacts on substitutable roles—by up to 1.7 times. This net positive suggests that AI momentum in certain sectors generates a ripple effect, pushing workplaces more broadly to value human skills that complement machine capabilities.
The paper’s methodology is meticulous: clustering skills into “complementary” and “substitutable” categories, leveraging logistic and linear regressions to analyse skill prevalence and wage outcomes, and controlling for confounders like education, experience, job structure, and regional economics. The authors further validate their findings with cross-country evidence, replicating the patterns observed in the U.K. and Australia, thus underscoring the global relevance of their insights.
One of the study’s key contributions is the identification of specific skill combinations that are especially valuable across various AI roles—technical proficiency paired with analytical thinking, for example. For data scientists, the presence of resilience and agility can translate into a 4% wage premium. This nuanced perspective highlights that calling for complementary skills is not just a general trend but a strategic priority, especially in AI-integrated professions.
Ultimately, Mäkelä and Stephany conclude that AI adoption does indeed substitute routine tasks, yet simultaneously stimulates demand for uniquely human aptitudes—resulting in a positive net effect on labour demand. They emphasise the urgency of reskilling efforts aimed not just at technical AI capabilities but also at human-centric skills such as ethics, digital literacy, resilience, and teamwork. The full study is available on SSRN.