Research and analysis

Understanding the UK AI labour market: 2020 Executive Summary

Published 18 May 2021

About this research

This report presents findings from research into the UK Artificial Intelligence (AI) labour market, carried out by Ipsos MORI, in association with Perspective Economics, Warwick University, and the Queen’s University Belfast, on behalf of the Department for Digital, Culture, Media & Sport (DCMS).

The research aimed to create a set of recommendations on policy areas that the government and industry should focus on, to bridge skills gaps in the sector. It involved:

• A survey of 118 firms and public sector organisations, including those whose core business was developing AI-led products or services, and others in wider sectors developing or using AI tools, technologies or techniques to improve their products, services or internal processes;
• A total of 50 in-depth interviews with firms, public sector organisations, recruitment agencies, employees and aspiring employees, universities and course providers;
• Analysis of AI job postings on the Burning Glass Technologies database; and
• A roundtable discussion with stakeholders from across government, the private and public sector to validate the findings.

Key findings

Working in AI and Data Science

The majority of surveyed firms were micro or small in size; 41% had 1-9 employees, and 48% had 10-49 employees. Only 9% had 50 or more employees. This is reflective of organisations typically seen in the sector. In the last 3 years, the most common uses of AI were: predictive machine learning (77%), regression for machine learning (70%) and classification (69%). The top 3 machine learning techniques used were: deep learning (71%), clustering (69%), and spatial data analysis (63%).

Career pathways

Career pathways into AI roles were diverse and lacked a clear structure. Some staff had entered their career directly from university having studied a related course, while others had not considered a career in AI when they went to university and it played no role in their course choice.

There was also evidence that vocational qualifications played only a small part in helping individuals progress into AI roles. Surveyed firms were unlikely to hire staff through an internship (11% of surveyed firms had hired some staff in AI roles through this route) or as an apprentice (3%). Although the survey did not specifically capture the reasons behind this, apprenticeships are currently not an established pathway into AI, and it is possible that recent developments in this sector could help to address this over time.

Refer to footnote [1].

Diversity

As elsewhere in the report, it is important to be aware that the sample comprised mainly small businesses and AI teams: the average (median) size of an AI team discussed in the survey was 4 people and the mean size was 9 people. Some firms had workforces that were very diverse but this was not the case across all the surveyed firms:

• Over half (53%) did not employ any females in AI roles. Overall, only a quarter (24%) of the surveyed firms’ workforce was female, but this was greater than in the cyber sector (15%) [2];
• Two in five (40%) firms did not employ any staff from an ethnic minority background in AI roles but overall 27% of the surveyed firms’ workforce was from an ethnic minority group. People of Asian ethnicity were thought to be well-represented in the sector, whereas some employers identified that people from African and Caribbean ethnic groups were under-represented.
• A similar proportion (41%) did not employ any non-UK nationals in AI roles but 32% of the surveyed firms’ workforce were non-UK nationals.

Employers wanted their workforce to be more diverse, but there were various barriers to achieving this, including:

• Lack of diversity among the individuals who were applying for AI jobs;
• Employers prioritising finding candidates with the ‘right’ skills, attitudes or cultural fit when recruiting, over improving diversity;
• Over-use of informal networks and word-of-mouth at the recruitment stage, which hindered candidates from a different background from being selected; and
• High competition for candidates from under-represented groups, due to growing focus on achieving diversity within the sector. This was a barrier to diverse recruitment. Smaller businesses found it difficult to compete against higher salaries offered by larger firms, and so employees were often poached.

The fast pace of change within the industry was a hindrance for women attempting to return to the workforce after a career break or maternity leave. This could be contributing to the gender pay gap for women.

Current skills and skills gaps

Technical skills gaps were a concern for many firms. A third (35%) said that existing employees lacking technical skills had prevented them from meeting their business goals, and 49% said that job applicants lacking technical skills had done the same. Some employers said that it had restricted or slowed their growth, or prevented them from moving forward with projects. Combining these results indicates that 62% of firms had faced problems with technical skills gaps, which was similar to the cyber sector (64%).

Technical skills gaps were reported in: understanding of AI concepts and algorithms (55%), programming skills and languages (52%), software and systems engineering (52%), and user experience (51%). However, many employers also faced issues with non-technical skills, including in communication, awareness of potential bias around the organisation’s use of AI, and awareness of privacy or ethical issues.

Recent reports are unanimous that, globally, the gap between demand and supply is significant and growing. In the survey, two-thirds of firms (67%) expected that the demand for AI skills in their organisation was likely to increase in the next 12 months, as a result of both COVID-19 and also other expected changes. Most interviewees in the qualitative element expected the demand for AI skills to continue to outstrip supply. This was in spite of a predicted increase in the supply of these skills, due to increasing interest in and awareness of AI in society. Similarly, recent research by Microsoft [3] found that more than a third (35%) of UK business leaders believed there would be an AI skills gap in the next 2 years.

Training

Three in five surveyed firms (62%) reported that employees in AI roles had undertaken internal or external training in the last 12 months to improve their knowledge and skills. This was similar to the UK economy overall (61% in 2019) [4], but higher than in the Cyber sector (24%) [5]. Only a quarter of firms reported offering training on ethics in AI (24%).

Much of this training was informal or on-the job. Many employers did not expect to provide formal training to their staff in the technical aspects of AI, as they felt staff should already have the appropriate technical skills when recruited or take responsibility for keeping these skills up to date. Self-directed learning was common in the sector.

Recruitment and retention

Two-thirds of surveyed firms had tried to recruit in the last 2 years. This was similar to the cyber sector (68%). Use of informal recruitment channels and word-of-mouth was common: 2 in 5 (42%) had used word of mouth or industry networks, and 1 in 5 (22%) had partnered with a university.

Among those that had vacancies in the last 2 years, 69% said that at least one vacancy was hard to fill. This was higher than in the cyber sector (57%) and in the Information and Communications sector as a whole (39% in the 2019 Employer Skills Survey). The bulk of hard-to-fill vacancies were among middle-management and other senior roles, which required 3 or more years of experience. Barriers to filling vacancies included: applicants lacking technical skills and knowledge (65%), work experience (40%), and industry knowledge (25%). This tied in with the findings on skills gaps, which identified poor commercial awareness among employees working in AI-related job roles. Other reasons for hard-to-fill vacancies were salary demands being too high (25%) and location or poor transport links (17%).

Job vacancies

The job postings search for AI and data science roles identified 110,494 postings in scope [7] from January to December 2020. The annual number of job postings had more than doubled since 2014, reflecting strong growth in demand for these roles among employers. Despite the challenges faced by employers in 2020 due to the COVID-19 pandemic, 2020 was the highest year to date for the number of online job vacancies related to AI and Data Science, with an increase of 16% from 2019 levels.

More than half of these job posts were in London and the South East. London was a major hotspot for AI and data science roles, with 36,715 roles identified. This was followed by Cambridge (5,453), Manchester (3,619), Bristol (2,505), Edinburgh (2,365), Oxford (2,311) and Birmingham (2,095). In relative terms, Cambridge had the highest concentration of demand for AI professionals.

A quarter (24%) of job postings were for Software Developers and Software Engineers, reflecting the breadth of technical and non-technical skills within these roles. The top 3 skills requirements mentioned in job descriptions included: Software Development Principles, Scripting Languages, and knowledge of Machine Learning.

Employers placed a very strong emphasis on applicants having Bachelor’s degree or higher qualifications. In total, 91% of roles required at least a Bachelor’s Degree or higher. Half (50%) of the job postings requested a background in Computer Science, and a quarter (25%) requested an Engineering (i.e. software) background. Mathematics was the third most requested subject (14%) followed by a background in Business Administration (8%), and Statistics (7%).

The mean advertised salary was £54,800 for an AI and data science related job posting (with a median value of £50,000).

The education sector (primarily Higher Education institutions) and financial sector (excluding insurance) provided the greatest demand for AI and data science professionals. Analysis showed there were a wide range of sectors employing AI and data skills (e.g. within scientific research, retail, legal services, health, consultancies, and the public sector) which suggested that AI skills can be deployed across the whole economy to improve national productivity.

Conclusions

The following recommendations are based on the evidence generated from all elements of this study. It will require engagement from government, AI firms and other employers, education institutions and recruitment agencies to take them forward:

• Increase diversity in the AI workforce, particularly among women, a wider range of ethnic minorities, and people from poorer backgrounds within the UK. Attracting global talent can also support increasing diversity in the workforce.
• Improve the talent pipeline through education, student employability and diversity. Increasing the talent pool and ensuring a future pipeline is key to the success of the AI sector within the UK. Entry into the AI sector from a diverse range of people needs to be encouraged, and can be achieved by increasing the levels of awareness about AI in general and the career opportunities in this sector. The talent pipeline can be further bolstered by ensuring that graduates have the skills required by employers. There was some evidence that employers felt that new graduates were unable to apply their skills to real life situations and/or had sufficient soft skills – Industrial Funded AI Masters have been one way of providing undergraduates with work experience to increase their employability.
• Create more opportunities for those not currently working in AI to convert to a career in AI and raise the levels of awareness of these opportunities. AI conversion courses [8] have been set-up to meet this demand and a new apprenticeship scheme has recently been launched. However, more thought should be given to how people at different life stages can convert to a career in AI.
• Encourage small firms to broaden their recruitment practices and provide support to small firms/employers located outside ‘hot spots’ to recruit and retain staff: small firms preferred word-of-mouth and networking to recruit their employees; this was a cost effective method of recruitment but meant that their talent pool of candidates was restricted. Employers who were located outside of the AI ‘hotspots’ found the recruitment and retention of staff particularly challenging, and there is a need to explore how to support these firms.
• Firms need employees to have a range of both technical and soft skills so that they can communicate effectively with management, other team members, internal stakeholders and clients about the AI product, its application and the benefits or limitations.
• Identify the AI skills required by different sectors: focus needs to shift towards thinking about the AI skills required in each sector, and how academic organisations could expand their courses to ensure that students gain the correct skill set.
• Increase ethics and bias training and make the case that it is in firms’ commercial interests to avoid flaws related to bias and ethical issues in their products.

---

1. The survey data has not been weighted, and so cannot be taken as being representative of the AI labour market. Where figures do not add to 100%, this is typically due to rounding of percentages or because the questions allow more than one response.
2. Cyber Security Skills in the UK Labour Market 2020
3. Microsoft report ‘AI Skills in the UK’
4. Employer Skills Survey, Department for Education, 2019
5. https://www.gov.uk/government/publications/cyber-security-skills-in-the-uk-labour-market-2020
6. Employer Skills Survey 2019 data tables, figure calculated from Table 1: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/925670/ESS_2019_UK_excl_Scotland_Data_Tables_Controlled_v03.01_FINAL.xlsx
7. Further details can be found in the Technical Report.
8. Postgraduate conversion courses in data science and artificial intelligence, Office for Students