Beyond GDP: Firm Size, Sector, and the Structural Determinants of Enterprise AI Adoption a Cross-Country Empirical Study

Shreehari Hulyalkar

shreeharihulyalkar8@gmail.com

Dayananda Sagar College of Engineering

Jayashree K

jayashree-mba@dayanandasagar.edu

Assistant Professor, Department of Management Studies Dayananda Sagar College of Engineering

Contents
Executive Summary

In the current study, an empirical analysis will be performed to determine whether the implementation of AI in businesses depends more on the country’s affluence. This report only uses information from sources which are official and reliable and can be confirmed. There were five other related studies carried out in eight countries in five economic sectors and three company sizes and the results show that there is no significant relationship between GDP per capita and AI adoption rate in a country while firm size, industry sector and especially survey method are good predictors for AI adoption. The single most important predictor identified in this study has nothing to do with economy but rather with the survey methodology as there is a difference of over 50% between the official statistics and surveys of executives concerning the use of AI.

Abstract: -
It is commonly believed that the uptake of AI technologies by firms correlates with a country's level of economic development since developed nations have the resources needed for the successful implementation of AI systems. In order to test this hypothesis, this paper uses secondary data based solely on reports of government statistical offices and international reputable research institutions: Eurostat, the OECD, Statistics Canada, the U.S. Census Bureau, Office for National Statistics in the UK, ICT surveys in Japan, South Korea, Singapore, Australia and China, and GDP per capita from the IMF database. For eighteen countries there is found no statistically significant correlation between GDP per capita and official rates of AI uptake among enterprises (r = -0.10, p = 0.68), thus suggesting that measurement framework, governmental digital strategies, and sectoral structure play a bigger role than income alone. The second hypothesis tested here is whether sector-level rates of AI adoption provided by two different statistical offices are highly correlated. Indeed, they prove to be almost perfectly correlated (r = 0.99, p < 0.001) and exhibit the same 'hourglass' distribution where ICT and professional services sectors are leaders of AI adoption whereas construction, accommodation and transport are laggards. A third analysis demonstrates an existing 34-38 percentage-point difference in terms of adoption rates between large and small firms in both the EU and OECD frameworks. A fourth analysis evaluates whether the survey methodology itself is a factor that causes the difference in the adoption levels by contrasting official statistics on enterprises with executive surveys (IBM, McKinsey, Stanford HAI) with an average difference of 50.3 percentage points between which is statistically significant (Welch's t = 4.41, p = 0.037), despite being much larger than the insignificant influence of national income. Overall, one may conclude that the implementation of AI technology by companies depends significantly on their size and type of activity, although even more on how the term 'use' is defined by the policymakers and survey developers.

Keywords: - enterprise AI adoption, secondary data analysis, GDP per capita, sectoral diffusion, firm size, digital divide, Eurostat, OECD