How do different sectors engage with AI?

AI has the potential to boost productivity and drive economic growth, but it does not affect all sectors equally due to the different nature of their activities. To unlock its full benefits, we need to unpack and understand the factors influencing AI engagement, including the demand for talent necessary to develop and maintain AI technologies, the development of innovative AI applications, the relevance of AI to various tasks in different occupations and sectors (and the barriers to its implementation), and the actual extent of AI use. 

To assess the extent to which AI affects the activity of economic sectors, also known as AI intensity, an OECD study focuses on four key indicators:

1. AI talent: proxied by the share of AI-related online job vacancies, reflecting the demand for AI-skilled workers.
2. AI innovation: measured by the share of AI-related patents, proxying the creation of new AI technologies and tools.
3. AI exposure: a novel measure of how tasks in various sectors are potentially affected by AI, accounting for barriers such as cost or regulation.
4. AI adoption: the actual share of firms using AI in their operations.

Together, these indicators provide a comprehensive view of the state of AI across sectors. They allow us to better understand the drivers behind sectoral leadership in AI intensity, while also pointing to the reasons some sectors do not have the same level of AI intensity. 

As shown in Figure 1, certain sectors, such as IT services, media and telecommunications, rank highly across all dimensions of AI intensity. These sectors are not just adopting AI – they are actively shaping its future through investments in talent and innovation.

The scientific R&D sector also demonstrates remarkable AI intensity, underscoring AI’s growing role as a foundational tool for modern research. This is unsurprising, considering the innovative capacity of this sector and its highly skilled labour force. Some of the outputs from this sector continue to be the basis for further developments of AI technologies.

Not all sectors are uniformly integrating AI. For example, the pharmaceutical sector shows high levels of AI talent but lower levels of AI innovation, potentially relying more on existing AI applications that are adopted and adapted for use by AI talent. Furthermore, actual AI adoption in this sector is not the highest, possibly suggesting a potential of AI to increasingly spread as the demand for specialised human capital further translates into relevant applications.

Computer and electronics manufacturing, a sector at the core of the ICT revolution, shows strong innovative efforts, high demand for AI talent and significant potential for AI applications. However, it has yet to achieve widespread adoption, suggesting untapped opportunities and possible implementation challenges, such as integrating AI into existing legacy automation systems to enhance predictive maintenance or real-time decision making.

Other sectors, such as the manufacturing of transport equipment or of other machinery and equipment, tend to exhibit overall medium AI intensity, with notably similar scores along the different AI intensity indicators.

Industries such as food, textiles, wood and paper, and construction, exhibit low AI intensity. This does not necessarily mean these sectors are being left behind. Instead, it may reflect specific technological needs or barriers that are unique to these industries that need to be addressed. For instance, the Construction industry is limited in its potential AI intensity due to the physical nature of many of  its activities. While planning shifts or raw material usage with software might benefit from AI use, pouring concrete is unlikely to benefit from AI today. The situation could change in the future with advances in automation unlocking greater potential for AI.

These findings challenge the notion that digital transformation and AI adoption always go hand in hand. While there is clearly a relationship between digital and AI intensity (e.g. the construction sector exhibits low levels of both), the analysis shows that being digitally advanced does not automatically translate into AI leadership. For instance, manufacturing of transport equipment, a highly digital-intensive sector, exhibits relatively low AI patenting activity and exposure to AI. Manufacturing of wood and paper also exhibits low AI intensity albeit being a medium-high digital-intensive sector. This suggests that AI diffusion might have its own characterising features and require its own distinct strategies and capabilities.

For business leaders, understanding where their sector stands in AI intensity can help benchmark adoption efforts and identify opportunities for competitive advantage. Sectors with high AI potential – as proxied by the AI exposure indicator – but lower current adoption (such as computer and electronics manufacturing) may represent an opportunity for companies to leap ahead of competitors by strategically investing in AI.

For policymakers, the findings not only highlight policy-relevant heterogeneity across different sectors, e.g. with respect to the abovementioned differences between potential and actual adoption, but also pave the way for future analyses that – leveraging the AI intensity taxonomy in combination with other data sources – may help better understand the implications of AI for economic outcomes. 

AI adoption is reshaping industries in ways that go beyond the simple division of “tech” versus “non-tech” sectors. Instead, different sectors are finding their own unique paths in the age of AI, driven by their specific needs, capabilities and challenges. Understanding these patterns is critical for anyone trying to navigate the AI revolution – whether leading a business or designing policy.

Going beyond sectoral analysis, complementary OECD work has further explored the different facets of AI described above with detailed microdata, focusing for instance on emerging trends in AI skill demand, on drawing a portrait of AI adopters across countries, or on exploring what technologies are at the core of AI. A recent OECD policy brief provides key takeaways and policy recommendations based on these and other OECD studies that have analysed AI adoption based on microeconomic data.