
We measure how artificial intelligence is changing the skills employers ask for, which firms actually adopt it, and what both mean for workers in Luxembourg and the Greater Region.
Which firms adopt artificial intelligence, and how far do the skills of their workforce shape that decision? How does AI change the skills employers demand, and when does it replace tasks or complement workers’ expertise? We study how these changes unfold within and across occupations, what they mean for employment, wages and productivity, and how the benefits and costs are distributed across workers and firms.

Measure
We use large language models to analyse millions of job advertisements and company websites across Belgium, France, Germany and Luxembourg, tracking changes in skill demand and the AI firms actually use, rather than their potential exposure to it, over time.
Explain
Our establishment surveys capture how firms deploy technology and organise work. Linking survey and text-based measures to administrative employer–employee records allows us to study the consequences for firms and their workers.
Advise
We translate findings into policy briefs and advice for ministries, EU institutions and social partners on reskilling, AI adoption and the organisation of work. Where licensing permits, we publish our indicators for others to use.

We build our own measurement infrastructure. Where licensing allows, we publish it.
Online job vacancies
Lightcast postings for Belgium, France, Germany and Luxembourg, 2018–2023, with skills extracted by LLM and reweighted to Labour Force Survey hiring flows.
___
MAP-AI firm indicators
Realised AI adoption, ecosystem role and technology type, built from firm website text 2016–2024 and linked to business registers. Released openly under CC BY 4.0.
Download the MAP-AI firm indicators (CC BY 4.0)
___
BIZA establishment surveys

The job-postings and firm-website projects share a skill taxonomy and a method: language models reading unstructured text that already exists, rather than new survey questions.
Both arrive at the same conclusion from opposite directions. Data skills are the broad foundation; AI skills are rare but decisive. Skill supply, not technology, is the binding constraint.
The BIZA surveys supply what web data cannot: firms stating what they use, and administrative records showing what then happened to their workers.
Why build all this? Because judging whether innovation policy works, as our own policy work argues, requires linked firm-level data that does not otherwise exist.






___
OTSS skill shares
Occupational technology skill shares splitting each occupation’s skill content into manual, digital and frontier technologies, at 3- and 5-digit occupation level. Released openly under CC BY 4.0, with an interactive dashboard.
Three projects, one programme. Each builds a new way to measure AI and skills, and each feeds directly into policy advice.