A new AI‑powered skill intensity index shows German jobs shifting toward frontier technologies: manual skills still dominate but frontier (including AI) skill content has risen, and occupations with moderate and very large increases in frontier intensity saw stronger employment growth than those with intermediate changes.

Main Finding

The paper introduces the occupational technology skill share (OTSS), a novel skill-based measure that separates manual, digital and frontier (including AI) technologies at the occupation level. Applied to the German labour market, the OTSS shows that in 2023 the average worker’s skill content was 42% manual, 38% digital and 20% frontier. From 2012–2023 there has been a broad shift toward frontier skills, but frontier technologies remain concentrated in specialised occupations. Changes in frontier skill intensity are non-linear: the relationship with employment growth is U-shaped (some intermediate increases in frontier intensity are associated with declines in employment, while large increases are associated with employment growth).

Key Points

• OTSS: a skill-based, occupation-level measure that classifies skills by technology intensity (manual, digital, frontier/AI).
• Method innovation: combines natural language processing, generative AI to standardise skill descriptions, and supervised machine learning to produce structured indicators of technological content.
• Cross-section (2023): average worker skill shares — manual 42%, digital 38%, frontier 20%.
• Distributional pattern: frontier technologies are concentrated in specialised occupations; digital technologies are widespread across occupations.
• Trend (2012–2023): overall shift away from manual and digital skills toward frontier skills across occupations.
• Labour-market effect: the link between changes in frontier skill intensity and employment growth is U-shaped, implying non-linear complementarity/substitution dynamics.

Data & Methods

• Data sources: occupation-linked skill labels enriched and standardised using generative AI; administrative employment data for the German labour market covering 2012–2023.
• Skill classification pipeline:
  • Natural language processing (NLP) to process existing skill labels.
  • Generative AI to produce standardised, human-readable skill descriptions.
  • Supervised machine learning to map enriched skill descriptions to structured indicators of technology intensity (manual / digital / frontier).
• Construction: OTSS computed for all occupations in the German labour market by aggregating technology-intensity-weighted skill shares.
• Analysis: OTSS linked to longitudinal administrative employment data to document dynamics (2012–2023) and to estimate the relationship between changes in frontier intensity and employment growth (empirical finding of a U-shaped relationship).
• Transparency: the method produces standardised descriptions and structured indicators that enable reproducible classification by technology intensity.

Implications for AI Economics

• Improved exposure measurement: OTSS provides a skill-level, technology-intensity measure that refines assessments of occupational exposure to AI and other frontier technologies beyond job-level or task-level proxies.
• Heterogeneous impacts: concentration of frontier skills in specialised occupations suggests uneven distribution of AI benefits and displacement risks across occupations and workers.
• Non-linear labour effects: the U-shaped relationship indicates that small or intermediate increases in frontier skill intensity may be associated with employment declines (possible substitution), whereas larger increases may coincide with employment growth (possible complementarity or reorganisation). Policymakers and researchers should account for non-linearity when forecasting impacts or designing interventions.
• Policy targeting: OTSS can guide targeted retraining and upskilling strategies by identifying occupations with rising frontier-intensity but weak employment prospects, and by highlighting occupations likely to benefit from AI adoption.
• Research directions: apply OTSS across countries and sectors, link OTSS to wages, firm outcomes and task-level data, and use it to evaluate the distributional consequences of AI-driven technological change.