Artificial Intelligence (AI) represents a general-purpose technology with significant potential to transform economies and promote broad-based economic growth – comparable to the revolutionary impact of electricity and personal computers.
In the European context, AI adoption is particularly critical as the region has suffered from low productivity growth in recent decades, widening the notable productivity gap compared to the United States.
A comprehensive IMF study analyzes AI’s impact across 31 European countries, using econometrically justified parameters to model multiple scenarios.
The authors focused primarily on identifying cross-country differences in productivity growth, considering variations in AI adoption rates according to countries’ economic indicators and regulatory frameworks.
Key findings: European average productivity growth limited to 1.1%, Estonia ranks among the mid-range countries
The authors’ analysis predicts an EU average productivity growth of 1.1% due to AI, with the greatest impact in wealthier countries like Luxembourg, Norway, and Switzerland. The lowest impact is expected in lower-income countries such as Romania.
The study places Estonia in the “middle range.” According to the preferred scenario, AI adoption could increase Estonia’s total factor productivity by approximately 0.8-0.9% over five years, falling below the average of 31 European countries at about 1.1%. The main reason is Estonia’s relatively moderate wage level, which reduces companies’ motivation to invest in AI that replaces or augments labor.
Estonia’s strength lies in its IT sector’s share, but the volume of traditional financial services and other white-collar sectors with high AI exposure is small. Therefore, sectoral “AI exposure” remains more modest than in Nordic countries, which together with moderate wage levels explains why Estonia doesn’t rise to the top of the income scale. According to Figure 7, both industrial structure and wage calculations give Estonia a small plus, but the overall position remains slightly below the European average.
 will do something, do it better, or do more.
The authors of this study rely heavily on the work of MIT professor Daron Acemoglu to divide the economy into certain work segments and then derive the “AI exposure” of each segment, which simply means how much AI capabilities overlap with the (human) capabilities needed to perform the work.
This study, in its radical “austerity,” argues that AI won’t really take away jobs, and there are several reasons - ranging from work segments not being coverable by AI to national and EU regulations precluding the automation of work segments.
It’s true that the authors themselves acknowledge that Acemoglu is ultra-conservative regarding AI availability. The more accessible AI solutions become, the greater AI’s impact will be. So in reality, the results may be significantly “better.”
However, the greatest impact of this study is that it articulates surprisingly clearly the main challenges that may significantly slow down AI adoption:
- Economic structure - i.e., how large a proportion of jobs can actually be performed with AI assistance;
- Labor costs in output - the lower the labor costs, the less interest there is in seriously using AI and automating work;
- Regulatory restrictions - the more human labor is a prerequisite for performing work, the less AI can be used.
Economic structure is something that is very difficult to change in the short or medium term. It can be gradually shifted through private and public sector cooperation, but it truly takes years.
Wage growth is also something that is somewhat difficult to manage within “Estonia’s boundaries.” Yes, it can be nudged a bit with minimum wage setting, public sector wage policy, and some other levers, but wages are more broadly affected by larger trends beyond Estonia, such as the state of the EU economy, EU inflation, and interest rates.
Regulatory restrictions, however, are an area where policy-making can begin to make a difference in the medium term. The focus should be on auditing existing regulations and identifying requirements that may significantly limit AI application in specific work. A somewhat simplified example: must building energy efficiency necessarily be audited by a human, or could an AI agent do this if the data is available?
An AI-friendly regulatory environment may also attract additional investments to Estonia. By creating the necessary framework for AI agents or other more complex AI systems, many companies may provide services using AI specifically in Estonia.
But in summary - yes, the IMF study is a cold shower for all AI aficionados, but also a necessary wake-up call. If we start systematically addressing these pain points, there’s a greater chance that Estonia’s economy will grow somewhat more than 0.8% with the help of AI.