PARIS - Building on recent OECD work, this paper analyses the skills sets (“skills bundles”) demanded in artificial intelligence (AI)-related online job postings. The analysis uses Burning Glass Technologies’ data for the United States and the United Kingdom and finds that skills related to the open source programming software Python and to machine learning represent “must-haves” for working with AI. Employers additionally value specialised skills related to robotics, AI development and applying AI.

A comparison of the periods 2013-15 and 2017-19 shows that the latter two have become more interrelated over time, with “neural network” skills connecting both groups. Network analysis relating AI skills to general skills highlights the growing role of socio-emotional skills; and of skill bundles related to programming, management of big data and data analysis. Key results hold for both countries and time periods, though differences emerge across occupations and industries.

Executive Summary This paper analyses the set of skills demanded by employers in the United States and the United Kingdom in online job postings related to Artificial Intelligence (AI). It relies on Burning Glass Technologies’ (BGT) online vacancy data for the period 2012-19 to provide first-time evidence about the relationships between the different AI skills that are demanded on the job, and between AI and non-AI skills.

AI skills are identified as in Squicciarini and Nachtigall (2021[1]), using a list of AI-related keywords of Baruffaldi et al. (2020[2]), whereas non-AI skills are any type of cognitive and socio-emotional skills that workers are required to be endowed with.
The evidence provided in this paper contributes to the OECD’s programme on AI in Work, Innovation, Productivity and Skills (AI-WIPS), supported by the German Federal Ministry of Labour and Social Affairs (BMAS).

This work aims to inform industrial, innovation, labour and education policies, by shedding light on the set of workers’ skills that are demanded on the job in relation to the development and adoption of AI.

It further provides evidence about those skills that are central to the deployment of AI, and the way these focal skills relate to other cognitive and socio-emotional skills. This helps prioritise interventions and informs the design of policies fostering the development of the human capital needed for AI to become the economically and societally enhancing technology countries hope for.

The main findings of the analysis and their implications for policymaking are as follows.

 Employers in the United States and the United Kingdom require relatively similar skill bundles for AI-related workers, and these remain largely the same across both periods considered (2013-15 and 2017-19). This may mean that, while AI is rapidly evolving, the skills required for AI-related workers are relatively stable and are expected to be also needed for the AI talent of the future.

 AI-related workers are technically skilled people, who need to exhibit a set of AI- related skills regardless of whether their job entails developing or adopting AI. All AI workers should be endowed with Python and machine learning (ML) skills, as they are by far the most frequently demanded skill pair, forming the foundation for AI-related workers’ skill profiles. Other important AI-related skills include data mining, cluster analysis, natural language processing and robotics.

 Three main bundles of AI-related skills emerge, which relate to ‘developing and advancing AI’, ‘AI applications’ and ‘robotics’ but employers also often require additional specialisations.

 Skills associated with AI applications and with developing or advancing AI have become more interrelated over time, with neural network (i.e. algorithms processing information in a way that mimics the way the human brain operates) becoming more central, connecting both groups.

 Expanding the scope of the analysis beyond AI-related skills, by also including other technical and socio-emotional skills in AI-related jobs, allows to further separate these specialisations (i.e. the bundles) into ‘programming and software related skills’, ‘the management of big data’, ‘data analysis tools and broader analytical skills’ and ‘socio-emotional skills’.

 Programming languages (beyond Python) are important in both countries, namely Java, SQL and C++, although the latter seemingly loses importance over the time period considered. AI-related jobs further require competencies related to big data and, in most recent years, data science, which is an inter-disciplinary field using scientific methods to extract knowledge and insights from data. The demand for data science-related abilities points to the need of AI workers to comb through vast amounts of collected data, to help identify business opportunities and to optimise product and process development.

 In both periods, a significant number of socio-emotional skills are demanded in combination with more cognitive skills. They can broadly be denoted as communication, teamwork and problem solving skills, which appear together with creativity and writing. When considering differences between sectors, we find that socio-emotional skills are particularly important for AI-related jobs in business services and education.

 Although the composition of the socio-emotional skill bundle remains basically unchanged over time when considering the top-30 skills, looking at the top-50 skills demanded reveals that communication skills gained in relative importance and that there is the need for AI talent to be further endowed with presentation skills and for being detail oriented.

 Communication-related skills are especially key in the United States as compared to the United Kingdom, possibly because there is relatively greater demand for AI- related managers in the United States. Results reflect the need to communicate within the team involved in the development and adoption of AI, as well as communicating among the different parts of the firm or institution developing or adopting AI, for AI to be correctly deployed.

 Of AI-related jobs, 87% relate to the professionals occupation, and the networks for professionals looks very similar to the one for all occupations combined. In the case of managers, however, socio-emotional skills are much more important (12 out of the top-30 skills, as compared to the 5 socio-emotional skills found across all occupations). When it comes to these socio-emotional skills, high centrality is observed again in the case of communication skills, problem solving and creativity, but for managers in the AI field, also presentation skills, planning, budgeting and business development are important.

 From the different occupational groups, professionals, craft and related trade workers and plant and machine operators and assemblers stand out in terms of the relatively higher number of different AI skills sought out of total skills demanded. This may reflect the need to be more specific about AI-related skills in job postings in occupational groups where AI skills cannot be taken for granted.

 Companies looking for AI-related managers or technicians generally also try to hire AI-related professionals, that is, recruitment related to these different occupational categories often happen jointly in the same organisation. This may indicate that professionals are crucial for the development and implementation of AI (which also represent the bulk of AI-related personnel sought) and that complementarities exist between the work of professionals and those of managers and technicians.

 Demand for AI-related jobs appears very concentrated geographically. Although the number of AI-related vacancies has increased across all regions in both countries, also as a share of total vacancies, a majority of those jobs continues to be located in London in the United Kingdom and in California in the United States, with little change observed over time.

For the full report, visit: https://www.oecd-ilibrary.org/docserver/2e278150-en.pdf?expires=1632324174&id=id&accname=guest&checksum=0B5C2FF3F22D1A982DB7E9AF606B251B