Final Conference: SkillShift – Future-Proofing the Workforce

Analysing millions of online job advertisements from France, Germany, Italy, the Netherlands and the UK, a group of Pillars researchers designed and implemented novel machine-learning and language models to predict the future pervasiveness of soft, hard-non-digital and digital skills across occupations. This way they identified new technology-related jobs and provided a sound basis for anticipating future labour market trends.

This was combined with an analysis of which tasks these new technologies can perform, in order to find out whether they are more apt to substitute or complement workers. Robots, for instance, are designed more to substitute workers than to complement them, while data-intensive technologies are consistently more complementary to humans, and more pervasive in services than in manufacturing. That said, while the number of sectors exposed to most digital automation technologies is still relatively small, it is expanding fast, since firms need to implement digital technologies rather urgently, given that productivity growth is currently sluggish and that an ageing population is slowing down growth in many economies, especially in Europe. 

In this context, the potential for digital applications, automation, and artificial intelligence (AI) to boost productivity growth is enormous. For businesses, this could mean both increased revenues—with products and services better tailored to consumers’ needs—and reduced costs, as raw materials, intermediate products, labour, and capital are used more efficiently. Moreover, the latest generative AI technology appears to boost productivity most for lower-skilled individuals, with the possible prospect of reduced labour market polarisation.

However, many barriers still hinder productivity-enhancing investment and widespread technology adoption. Uncertainty about demand dampens risk-taking and long-term investment, which points to a role for policy, in the form of well-designed incentives to invest in human and organisational capital. This, in turn, calls for better leadership and management skills—and more ambition.

The Skills Angle

The other key element is skills. Already in short supply, the requirements associated with new technologies could make the shortages get quickly worse. Retraining and upskilling is clearly required across both essential workforce skills as well as various knowledge and occupational specialisms. 

One branch of Pillars’ research found that vocational education, which typically imparts skills early in a person's career, can have a long-lasting positive impact on workers. Using data from the German apprenticeship system, our researchers identified 13,000 narrowly defined skills imparted by apprenticeships and grouped into six categories: cognitive, social, digital, manual, managerial, and administrative. They then merged this dataset with administrative labour-market data covering almost three decades to reveal the economic value of workers’ early career skills over their working life.

Workers who start their career with higher cognitive, social, or digital skills earn significantly higher wages over long-run horizons, partly thanks to skill complementarities: workers who simultaneously acquire cognitive and social skills during apprenticeship are particularly valuable on the labour market, while those with digital skills tend to earn even higher wages. These benefits have risen substantially over the past 30 years, evidencing the value provided by vocational education.

Specifically, our research showed that if you acquire one more month of cognitive skills during apprenticeship, you will earn a 1.3% higher salary about 16 to 20 years after apprenticeship completion, meaning about €500 annually over your working life. 
The biggest payoff comes from social and digital skills. Acquiring one more month of social skills pays about 1.5% wage increase, which is €500 six to 20 years after apprenticeship completion. Digital skills are even better, with a 2.1% wage increase, equivalent to about €800 annually. And the best part, both social and digital skills are transferable across many kinds of occupations. 

Looking specifically at how training can mitigate workers' exposure to automation and help them adapt to technological disruptions, another Pillars research group used granular individual-level data on job tasks across 32 countries, including 17 European ones. A unique measure of automation risk at the individual-level unveiled substantial variation within occupations. Notably, even within narrowly defined occupations, workers who received training have a 4.7-percentage-point lower risk of being automated away, corresponding to a 10% reduction in the mean automation risk across the sample. Additionally, training increases wages by 8% and, significantly, benefits both younger and older workers equally—and, interestingly, training measures are more effective for women, which is good for inclusiveness. These findings underscore the crucial role of training in enabling the workforce to adapt and thrive amidst evolving technological challenges.

A suitable policy environment can have a large impact on the effectiveness of training efforts. For instance, making sure that people have good education skills before they leave school and enter the labour force can facilitate people's retraining later on. Policies for reskilling adults, in turn, should make sure that they work hand in hand with other policies and with what is already available in the local labour market, paying attention as well to the other services that may need to be provided in addition to training. Sectoral work training programmes do this in an effective way and could serve as examples to follow. 

All of this works best when the technologies that are emerging can be identified early enough. This was the task of another Pillars research group, based on the premise that characterising emerging digital automation technologies is critical to understanding the changing patterns of work, firm and industry organisation, as well as labour demand, and thus to formulating tailored policies to mitigate risks while harnessing their benefits. Analysing a large corpus comprising millions of patents and scientific publications related to automation technologies, this group identified emerging technologies and proposed a methodology to combine machine learning methods with state-of-the-art sentence transformers from the field of computational linguistics. In this way, they identified six patterns of technological and scientific development that provide a better understanding of which digital automation technologies are likely to emerge in the near future, and which technologies have already matured. 

They next enlisted the combined expertise from hundreds of sources to make predictions about which automation technologies will become prevalent by 2030, which tasks they will perform, and what skills they will replace. Unsurprisingly, generative AI and large language models show up prominently, but other less newsworthy ones will have a noticeable impact, such as collaborative robots, augmented reality, virtual reality, highly autonomous systems and yes, the not-so-new but ever more pervasive blockchain, for which new applications are constantly being devised. Cloud computing will take an ever more prevalent role, as well as an evolved internet of things. 

The semantic connection between patents and publications documenting emerging digital automation technologies and descriptions of industries and occupations revealed that not only workers performing rote tasks, but also technicians and professionals are becoming highly exposed to digital automation technologies, with managers positioned in the middle of the exposure distribution. 

This suggests that the increasing penetration of artificial intelligence and robots is likely to have a massive impact on regional labour markets, with the skill portfolios of occupations changing as a result of their exposure to such technologies, along with the types of work arrangements. Our researchers argue that regional capabilities play a crucial role in the capacity to cope with the challenges of these technologies. The new skills and jobs that require such skills are more likely to emerge in regions with complementary skills and occupations, which are more likely to seize the opportunities brought by automation, whereas regions without such capabilities will likely experience slow or even negative employment and wage growth, with a growing risk of increased unemployment.

The Policy Angle

All the above results have important policy implications, not only for making labour markets resilient in the face of technological change, but also for enhancing labour market inclusiveness.  

Thus, policymakers need to address three labour market factors simultaneously—job creation, job displacement, and job transformation—to ensure inclusive labour markets during technological transformation. This calls for a comprehensive, coherent and inclusion-focused employment strategy, the success of which relies on leadership at the levels of policy design and policy implementation.

Policy approaches differ across regions, based on levels of economic development, but all instruments for job creation involve collaborative frameworks that stimulate synergies, connectivity and inclusion. Technology-enabled job creation may enhance short-term inclusion of vulnerable groups in the labour market, but it rarely leads to long-term inclusion unless it is closely related to upskilling/reskilling. This calls for a transformation of the education and training system, and development of evidence-based, forward-looking active labour market policies. It also requires effective tripartite co-operation between trade unions, employers' organisations and the government.