Why the Same Exposure Travels Differently Across Countries
The European workplace data show that generative-AI adoption is not only a property of workers or occupations; it is also shaped by the environments in which those workers do their jobs. This chapter focuses on the country-level conversion problem: why a similar relationship between occupational exposure and actual use can be steep in one national labour market and weak in another. The answer is layered. Digital work creates the basic possibility, while training systems and the composition of jobs appear to help determine how effectively that possibility becomes adoption. None of these comparisons, however, turns a country into a simple winner or proves a single national policy caused the difference.
The first threshold is a digital workplace
A generative-AI tool is software, so it needs a software-shaped workplace in which to be used. The European study measures digital intensity as the share of workers who report using a computer almost all or all of the time. This is a relatively simple indicator, but it captures an important threshold. If a worker’s daily tasks are rarely performed through a computer, an occupation’s technical exposure to language models may not provide an immediate route to adoption. The potential is present in a task description, yet the infrastructure for acting on it is thin.
Digital intensity is not the same as a fast internet connection, a company’s AI budget, or a worker’s digital skill. It is a measure of how deeply computer-mediated work is already embedded in a country’s jobs. That makes it a prerequisite rather than just another characteristic to add to a long list. A worker cannot easily integrate a text-generation or recommendation tool into a workflow that has no regular digital interface, regardless of how susceptible the underlying task may be.
The country evidence reflects this threshold. Digital intensity is the dominant raw correlate of variation in the country-specific exposure–adoption slope, with a correlation of 0.88. The slope is the country-level version of the question developed earlier: as occupational exposure increases, how much does the probability of reported use increase? Countries with more regular computer use tend to have steeper relationships between technical exposure and workplace adoption.
A country slope is therefore not a score for national enthusiasm. It is recovered from how adoption varies across occupational exposure within that country. This makes the comparison more revealing than a simple table of national averages, but also more sensitive to the amount and distribution of data available in each setting. A steep estimate can mean that exposed occupations are converting potential particularly well; a flat estimate can mean that adoption is low across the occupational distribution, that exposure is not finding a practical route into work, or that the measure is too imprecise to distinguish the pattern. The slope is a disciplined summary of a relationship, not a complete diagnosis of its cause.
The distinction between the average level of use and the slope sharpens the interpretation. The European study finds a strong positive relationship between a country’s mean adoption rate and its exposure–adoption slope, with a correlation of 0.89. That does not mean every worker in a high-adoption country uses AI more equally. It suggests instead that higher national uptake is driven especially by workers in occupations where the technology is more relevant.