The OECD uses glass.ai to identify companies adopting AI technologies.

The rapid development of Artificial Intelligence (AI) technologies is having an impact on virtually every part of the global economy. AI is leading to significant growth and productivity gains, but its adoption is not uniform across different sectors and industries. Measuring how firms are adopting and using advanced technologies like AI is critical for understanding the current state of advanced economies and planning for the future, but there is a lack of comprehensive data on firms’ adoption of such technologies.

Typically organisations have been gathering evidence via surveys — for example, the US Census Bureau conducted a big study in 2020 showing that only about 9% of US firms employ AI tools. The data was collected via surveys of 500k US businesses. In the UK, a more recent study by the Department for Digital, Culture, Media and Sport (DCMS) followed a similar approach and aimed to assess the scale of AI activity in UK businesses.

At glass.ai we have developed a new approach that collects evidence of adoption by deep reading the web (millions of websites, press releases, job posts, etc.). This novel approach is much quicker, can gather rich intelligence on millions of businesses, and provides data and insights about adoption that complements the results gathered via surveys. PwC and other consulting firms and corporates have used our AI research capability to uncover case studies and evidence of the adoption of new technologies.

More recently, the OECD, a key pillar of global economic policy, conducted a pioneering study to understand how different types of companies are adopting AI in the UK. Instead of doing a traditional market research study that involves interviewing companies, the OECD used our AI research capability to understand how different companies are adopting AI.

The AI was used to glean insights from multiple sources of data to characterise firms adopting AI. The starting point was a set of AI-related keywords and phrases that the OECD had previously developed. There were four types of data sources: a) Intellectual Property Rights (patent and trademark databases) b) information from company websites, c) online job postings and (d) financial information relating to companies. To be considered an AI adopter, a firm must either have: i) applied for or registered an AI-related patent or trademark, ii) mentioned an AI-related activity/application on their website, news or social media or iii) posted an AI-related job listing. Companies that had applied for AI patents and trademarks for their products and services were termed ‘AI innovators’, not adopters.

Company websites are a rich source to understand the products and services offered by firms. At glass.ai, we applied our world-leading intelligent crawling technology across all UK companies that have a website and social media presence. By machine-reading and understanding the products and services offered by the companies, and by reading thousands of news sources, we were able to identify UK companies adopting AI. We also determined whether companies were end-users of AI or companies developing new AI products and services.

One of the key findings of the OECD study, somewhat unsurprisingly, was that the largest sector adopting AI was the Information and Communication sector (40%) followed by the professional and scientific activities sector (21%). Among the companies that were classified as ‘AI Innovators’, the manufacturing sector had the second highest percentage of companies (after the Information and Communications sector).

In terms of geographies, London and the South had the highest number of firms adopting AI technologies. London had the highest number of ‘AI Core Business’ companies (43%), and companies adopting AI were more likely to be geographically close to universities with AI departments.

Another finding was that the most productive companies in the UK are also the companies that adopt AI the most (labour productivity is defined as revenues divided by the number of people employed). The OECD concluded that more than one-third of all AI adopters in the UK are in the top 10% of most productive companies. Larger firms, on average, are likely to be adopters of AI compared to smaller firms.

In summary, the OECD used our AI research capability to deep-read the web and uncover a variety of patterns of companies adopting AI. This was a purely data-driven exercise, combining a multitude of sources, allowing the methodology to be applied to many other countries, and repeated, to truly understand the dynamics and adoption of emerging technologies.