2024 was the year in which AI made a breakthrough in businesses, with some quick on the uptake. In the EU27, 13.5% of enterprises with 10 or more employees were using AI in 2024, a rate of usage 60% higher than the year before and almost double in some countries.
This surge has been supported by the arrival of accessible generative AI (GenAI), which brought AI capabilities within the reach of all. But is this enough to avoid another digital divide?
Gen AI… some escaping the pack?
On the face of it, generative AI could be a great leveller, provided it is deployed efficiently and safely across regions and businesses. The use of predictive AI for recommendations or automation – e.g. to optimise production and maintenance, reduce energy and material consumption, or anticipate market fluctuations and enable dynamic pricing – requires complex models, high computing power and deep expertise.
GenAI is comparatively more accessible, allowing businesses to leverage pre-trained models and experiment without advanced technical skills. GenAI also expands the range of applications that could be tailored to local needs and business models, for example by helping firms unlock dormant data and streamline administrative tasks.
Yet technological change always brings winners and losers. New gaps often layer onto existing ones, deepened by previous technological waves. Some firms are slow on the uptake, holding on to traditional methods and technologies, or lacking resources and skills to transform. As a result, divides appear across industries, firms, populations and places, as some flourish while others are left behind.
Delayed adoption is hard to overcome. First movers — who set industry standards, build reputations, and consolidate networks — can quickly gain an insurmountable advantage.
AI compounds this difficulty by enabling businesses to use larger volumes of data — or even generate new data — leading to more accurate and high-performing models.
These models, in turn, produce even better data, reinforcing the cycle. Moreover, while pre-trained platforms enable proof of concept, deep GenAI integration, fine-tuning and implementation at scale still pose major challenges, remain complex and highly context-specific, and often imply external expertise and costs.
These divides are already playing out in AI adoption today, with some firms and regions surging ahead and others falling behind.
Regions are entering the race but the most innovative have taken the lead:
AI adoption rate by region and innovation performance (%), EU27, 2023 and 2024
Source: Kergroach S. and J. Heritier (2025 forthcoming), “Early Divides in the AI Transition”, based on Eurostat (2024), Artificial Intelligence, by NACE Rev. 2 activity and NUTS 2 region [isoc_r_eb_ain2], and EC Regional Innovation Scoreboard 2023.
Note: Adoption rates refer to the percentage of companies (10 or more employees) using AI in each region (NUTS2 level regions, except for Austria and Belgium NUTS1). All activities. The level of innovation performance of regions is a composite index calculated from a set of performance indicators in 2023.
Quick off the line
In leading countries such as Korea, Denmark, and Sweden, AI adoption rates among enterprises exceed 25%, across all firm sizes and sectors. Innovation hubs such as Midtjylland (Denmark) and the Brussels-Capital Region (Belgium) have adoption rates exceeding 30%.
In knowledge-intensive sectors, such as information and communication services, as well as professional, scientific, and technological services, Finland, Sweden, and Denmark report business adoption rates exceeding 66% and 45%, respectively. Unlike earlier technologies that primarily impacted low-skilled jobs, AI can perform expert, knowledge-intensive tasks.
On the other hand, with adoption rates of 5% or less, Turkey (4.4%), the Western Transdanubia region in Hungary (3.3%), and the accommodation and food services sector (OECD median of 5.4%) are at the rear. Small firms are also lagging, with an OECD median adoption rate of just 8.5%.
Regional disparities are also evident within countries and within sectors. In Spain, only 11.4% of firms in the IT services sector in the Rioja region use AI, and 19% in Cantabria – well below the European average of 48.7%. By contrast, Extremadura, an emerging European innovation hub in ICT, renewable energy, and zero-waste software, reports an impressive 90.5% adoption rate, as it establishes itself as a data centre hub for generative AI and cloud computing services.
With GenAI in particular, time savings could be substantial in 77% of jobs in Greater London (UK) compared to 16% in Guerrero (Mexico). Capital cities stand out in intra-country differences, suggesting a growing urban-rural divide as GenAI spreads.
Navigating the next turn
The initial stages of AI adoption have passed, and we are now entering the phase of innovation diffusion when early adopters form and reap exponential benefits. Policy makers must work quickly to ensure it delivers widespread benefits, manage risks and prevent regional divides from deepening.
Regions’ industrial structure will dictate the pace of change, along with the effectiveness of local networks and innovation actors to integrate different AI technologies into tailored, reliable business solutions.
Local governments will therefore play a key role in facilitating co-operation, including across borders, in promoting data sharing, and stimulating uptake through standards, skills programmes, and infrastructure investment. In doing so, they must also manage new risks arising from unethical and untrustworthy uses of AI. Issues of security, protection of privacy and human rights, industrial sovereignty and geopolitics have become integral parts of their transformation agenda.
To get to the front of the pack, Catalonia (Spain), Quebec (Canada) and Scotland (UK) have already adopted comprehensive Artificial Intelligence Strategies to strengthen local AI ecosystems, drive the development of social and business applications, and reinforce public-private collaboration and cross-border linkages.
The AI race isn’t just about speed — it’s about inclusivity. Policy makers must act now – alongside businesses and communities – to bridge divides, build trust and to unlock AI’s full potential for everyone.
To find out more about the new geography of Generative AI, see our flagship publication Job Creation and Local Economic Development 2024: The Geography of Generative AI and forthcoming report on “Early Divides in the AI Transition“.
OECD work
Kergroach, S. and J. Héritier (2025 forthcoming), Early Divides in the AI Transition, OECD Regional Development Papers, OECD Publishing, Paris.OECD (2024), Job Creation and Local Economic Development 2024: The Geography of Generative AI, OECD Publishing, Paris, https://doi.org/10.1787/83325127-en.
OECD (2021), The Digital Transformation of SMEs, OECD Studies on SMEs and Entrepreneurship, OECD Publishing, Paris, https://doi.org/10.1787/bdb9256a-en.
Sandrine Kergroach is Head of SME and Entrepreneurship Performance, Policies and Mainstreaming unit at the OECD Centre for Entrepreneurship, SMEs, Regions and Cities (CFE). She leads the work on innovation, internationalisation and the scaling up of SMEs and start-ups, their productivity and ESG performance. She supervises activities related to policy monitoring, the development of data infrastructure and the OECD SME and Entrepreneurship Outlook. She also leads efforts for mainstreaming SME&E policy considerations. Sandrine holds a Doctorate in Economics (TU Berlin), a Master in Strategy and Management (Paris Dauphine-PSL), a Master in Modern History (Paris Sorbonne) and a Bachelor in Applied Economics and Statistics (Paris Dauphine-PSL).
Julien Héritier is an AI consultant, specialised in the digital transformation and based in Paris (France). He carried out projects on artificial intelligence and information technology for large players in the energy and automotive sectors, and contributes to develop generative AI solutions for the private sector. Julien holds a Master degree in applied economics from Paris I and a Master in economics with a specialisation in data science from Paris-Saclay.