State of AI
Methodology

How this report was built.

Dataset, prior use in the published AI-workforce literature, the Austrian sample, the five-step role taxonomy with Fleiss' κ = 0.84 inter-rater reliability, and the broader research programme behind the measurement standards used here.

A · Dataset

Data spine: Revelio Labs Workforce Intelligence

The data spine of this report is Revelio Labs Workforce Intelligence, an individual-level workforce panel that standardises public professional records (LinkedIn, XING and equivalent networks) into a globally comparable position-history schema. The data are accessed under academic licence via Wharton Research Data Services (WRDS) at WU Vienna.

Each position record carries a standardised occupational classification drawn from Revelio's v3 role universe of approximately 17,000 fine-grained categories, plus a machine-imputed annual salary in EUR (ECB annual averages), machine-predicted binary gender, education, seniority on a 1–7 scale, and a resolved corporate parent linking each position to a firm. The panel is observed at quarterly frequency from 2018 onward and is constructed at the unique-worker level, supporting both cross-sectional snapshots and within-worker longitudinal analysis.

Data use

All figures published in this report are aggregated statistics derived from Revelio Labs individual-level employment data, accessed under academic licence via Wharton Research Data Services (WRDS). The underlying user-level records are not redistributed; this site publishes derived headcounts, shares, medians, and longitudinal trends only. Aggregations are based on the Vienna University of Economics and Business (WU Wien) WRDS subscription; reuse of the figures should cite this report and Revelio Labs as the upstream data provider.

B · Use in published literature

Prior use in the published AI-workforce literature

Individual-level workforce panels of this type have become a standard empirical data source in the published AI-workforce literature. Babina, Fedyk, He and Hodson (2024, Journal of Financial Economics) use Revelio to identify firm-level AI investment among S&P 500 firms; Tambe (2025, Management Science) studies AI reskilling on similar individual-level data; Marchetti and Puranam (2026, Strategic Management Journal) use it to study firm cultures; and Cheng (2025, Information Systems Research) leverages a parallel platform-data spine to study AI labour shocks.

Earlier waves of the AI-skills literature establish the underlying empirical strategy: Alekseeva, Azar, Giné, Samila and Taska (2021, Labour Economics) and Acemoglu, Autor, Hazell and Restrepo (2022, Journal of Labor Economics) document the AI wage premium and labour-demand effects of AI exposure.

The Revelio data have been validated independently against external benchmarks. Cai, Chen, Rajgopal and Azinovic-Yang (2024, Review of Accounting Studies) show that Revelio's firm-level aggregates correlate strongly with hand-collected SEC disclosures, and Liang, Lourie, Nekrasov and Shevlin (2025, Journal of Business Finance & Accounting) demonstrate close tracking of official labour statistics across industries and over time. The occupational exposure measures used in the brain-drain and complementarity chapters anchor on Eloundou, Manning, Mishkin and Rock (2024, Science) and Felten, Raj and Seamans (2023, Strategic Management Journal) — both built on the O*NET task backbone also referenced by the OECD AI Skills Indicator and the ILO/NASK Refined Global Index of Occupational Exposure to AI (2025).

C · The Austrian sample

Sample, segment, and coverage

The Austrian sample drawn from this panel comprises 2.86 million position records covering 1.23 million unique workers across 188,265 firms over the 2018–2025 period. The primary segment is austria_located, defined as every position physically located in Austria and conceptually comparable to Eurostat's place-of-work employment definition. Brain-drain and diaspora analyses use the total segment of all Austrian-trained workers, regardless of current location. Cross-country benchmarks are constructed from direct WRDS aggregate queries covering 38 European peer economies.

Coverage: Revelio's Austrian coverage is approximately 22.5 percent of Eurostat official employment, reflecting LinkedIn and XING penetration in the DACH region. Within-country trends are robust; absolute cross-country headcount comparisons should be read with this coverage caveat. Time period: 2018–2025, with 2025 figures preliminary owing to Revelio's profile-backfill lag.

Salary metric: median machine-imputed annual EUR salary (ECB annual averages); aggregate medians align with published Austrian salary surveys, but individual-level salary uncertainty remains high. Gender classification: machine-predicted by Revelio at ~95% aggregate accuracy; binary; all figures are machine classifications, not self-reported identities. Brain drain is defined empirically as the next observed position being located outside Austria, not as permanent emigration.

Three slices of the May 2026 refresh

Three views of the same Revelio Labs refresh, each shaped to its question.

  1. The Austrian workforce panel — Chapters 1, 2, 5, 6. Every position physically located in Austria, 2018–2025. Filter: country = 'Austria', no firm or role gate. · 2.86M positions · 1.23M unique workers · 188,265 firms · ~22.5% Eurostat coverage.
  2. The AI career trajectory panel — Chapter 3 (Brain Drain). Every Austrian-trained AI worker's full career path, including the ones who left. · 19,037 AI users · 131,947 positions across 80+ destination countries.
  3. The cross-country benchmark — Chapter 4 (Benchmarking). Direct WRDS aggregate queries for 38 European countries — tier counts, density, salary, share. · 38 countries × 8 years × 4 tiers.
D · Measuring AI workers

A five-step construction pipeline

We identify AI workers by mapping each position's v3 occupational code into a three-tier role taxonomy: Build, Enable, and Integrate. The taxonomy is constructed through a five-step pipeline applied identically to the Austrian sample as in the authors' broader research programme:

  1. STEP 01
    Revelio v3 base taxonomy

    Start from Revelio's standardised occupational classification system: ~17,000 fine-grained role categories mapped consistently across firms and over time.

    17,000 → candidate set
  2. STEP 02
    Skill- and keyword-based filtering

    Identify the candidate universe of AI-related roles using a combination of skill endorsements and title keywords drawn from prior AI-workforce research (Alekseeva et al. 2021; Acemoglu et al. 2022; Babina et al. 2024).

  3. STEP 03
    Ensemble coding

    Combine rule-based classifiers with LLM-assisted coding to produce an initial tier assignment (Build / Enable / Integrate) for every candidate role.

  4. STEP 04
    Independent expert validation

    Two to three independent academic experts in AI and data science (external to the author team) manually review every role. Disagreements are resolved through discussion.

  5. STEP 05
    Inter-rater reliability

    Three-coder agreement on the resulting classification reaches Fleiss' κ = 0.84, within the "near-perfect agreement" range (Landis & Koch 1977).

    Fleiss' κ = 0.84 · "near-perfect" (Landis & Koch 1977)

The three tiers

  • Build — 25 roles that create AI capability from first principles (AI Research Scientist, Machine Learning Engineer, Deep Learning Engineer, Generative AI Engineer).
  • Enable — 60 roles covering infrastructure, deployment, and governance that keep AI systems running at scale (MLOps Engineer, AI Solutions Architect, Cloud Platform Engineer, Data Governance Specialist, AI Quality Analyst).
  • Integrate — 257 roles that embed AI in products, decisions, and business processes (AI Product Manager, Data Scientist, AI Strategist, AI Designer, Analytics Lead).

"Core AI" covers the ~120 roles flagged include_in_core = 1 across Build / Enable / Integrate. "Full AI" adds the outer Adjacent ring (BI, domain analytics, decision support) for ~370 roles in total.

Caveats
  • Revelio is not a census — coverage varies across countries and over time.
  • 2025 is preliminary; year-over-year comparisons involving 2025 should be read with Revelio's collection lag in mind.
  • Machine-predicted gender is binary — non-binary identities are not separately represented.
  • Salary data are imputed medians; reliable at the cohort level, not for individual comparisons.
  • Brain drain is defined as the next observed position being outside Austria; it is not a permanent emigration measure.
E · Research programme

Part of an ongoing research programme

The dataset, construction pipeline, and role taxonomy applied here are not built for this report. They are the same instruments the authors use in an ongoing programme of research on AI workforce composition, firm value, and worker-level career adaptation under generative-AI exposure. The Austrian application requires no recoding of the role universe and no re-validation of the taxonomy: the five-step construction pipeline, the expert-validation protocol, and the inter-rater reliability evidence carry over directly.

This study therefore inherits the same measurement standards as the authors' broader workforce-AI research and applies them to the Austrian labour market.

Meet the authors →
References

References

  1. Acemoglu, D., Autor, D., Hazell, J., & Restrepo, P. (2022). AI and Jobs: Evidence from Online Vacancies. Journal of Labor Economics, 40(S1).
  2. Alekseeva, L., Azar, J., Giné, M., Samila, S., & Taska, B. (2021). The Demand for AI Skills in the Labor Market. Labour Economics, 71.
  3. Babina, T., Fedyk, A., He, A., & Hodson, J. (2024). Artificial intelligence, firm growth, and product innovation. Journal of Financial Economics, 151, 103745.
  4. Cai, W., Chen, J., Rajgopal, S., & Azinovic-Yang, L. (2024). Workforce data quality: A comparison of Revelio Labs aggregates with hand-collected disclosures. Review of Accounting Studies (forthcoming).
  5. Cheng, Z. (2025). Skill-biased technical change, again? Online gig platforms and local employment. Information Systems Research, 36(3), 1354–1374.
  6. Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Science, 384(6702), 1306–1308.
  7. Felten, E., Raj, M., & Seamans, R. (2023). Occupational, industry, and geographic exposure to artificial intelligence. Strategic Management Journal, 42(12), 2195–2217.
  8. International Labour Organization & NASK National Research Institute. (2025). Refined Global Index of Occupational Exposure to Generative AI. ILO Working Paper Series. Geneva: ILO.
  9. Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159–174.
  10. Liang, C., Lourie, B., Nekrasov, A., & Shevlin, T. (2025). The gender position gap and firm performance. Journal of Business Finance & Accounting, 52, 2464–2491.
  11. Marchetti, A., & Puranam, P. (2026). Are less hierarchical firms organized around stronger cultures? Strategic Management Journal, 47(2), 463–493.
  12. Tambe, P. (2025). Reskilling the workforce for AI. Management Science, 72(1), 515–537.