On the Automation of Job Tasks: Occupational exposure to Artificial Intelligence and Software
While rapid advances in digital technologies transformed the occupational structures and workers‘ skill and task composition over the past decades, much less is known about how Artificial Intelligence technologies (AI) will shape future labour markets. As part of the “ai:conomics” project, researches show that artificial Intelligence (AI) is more likely to affect the job tasks of highly educated employees, while medium- and low-educated employees are more likely exposed to software systems without AI (software). Furthermore, while AI and software can carry out certain tasks, not all tasks within an occupation can be automated. Therefore, the implementation of AI and software cannot be regarded as the exclusive solution to the skilled labor shortage.
These are some key-findings presented in the ai:conomics research project’s fourth policy brief. Using indicators that measure the automation potential of job tasks, the researchers characterize occupations that could be substituted given the patented innovations in their field. They refer to these indicators as relative exposure potentials, as they enable comparisons between occupations. In this study, the relative exposure potential of AI and software were considered separately, leading to more nuanced insights.
These are the main findings:
- Using indicators that measure the automation potential of job tasks, occupations that could be substituted given the patented innovations in their field are characterized (Webb, 2020). It is referred to these indicators as relative exposure potentials, as they enable comparisons between occupations.
- It is shown that while the potential exposure to software is more likely to affect the job tasks of low or medium-educated employees, AI is more likely to have an impact on highly educated employees. This signals the distinct potential of AI in targeting different types of workers than other automation technologies.
- The exposure potential to AI and software is particularly high for workers in the manufacturing industry and in information and communication technologies.
- Occupations with a higher share of female employees appear to have lower AI and software exposure compared to those with a higher share of male employees.
- AI and software have the potential to take over slightly more tasks in occupations with skilled labour shortages, compared to occupations without shortages.
- Nonetheless, the findings imply that even though the observed technologies can potentially automate certain tasks, not all tasks within an occupation can be carried out by AI.
AI and software exposures across the German labour market
Industries and groups of workers that differ in their socio-economic, demographic, or occupational characteristics have dissimilar potentials to be exposed to AI and software.
Regarding the differences between industries, the paper shows that the relative exposure potential is particularly high in manufacturing and construction, while it is particularly low in health and social services. More specifically it shows that employees in ICT, financial and insurance services and professional services have higher exposure to AI, while the opposite pattern occurs for employees in the manufacturing, construction, wholesale and retail trade, transportation, healthcare and social services, where software exposure is higher.
When plotted against qualification levels, the findings indicate that AI exposure increases with qualification levels. Thus, highly qualified employees appear to be most exposed to AI compared to employees with low and medium qualifications who are shown to be more exposed to software. A possible explanation for this is that tasks related to prediction, recognition, forecasting and analytical decision-making are the core of current applications of AI. These rather (non-routine) complex tasks are mainly performed by highly educated workers. On the other hand, occupations that require no formal vocational training (e.g. service staff, cleaning) generally do not require handling large amounts of data and thus are more affected by the implementation of software. This finding indicates that the impact of AI on the labour market is more likely to affect workers across a wider skill and wage spectrum than previous technologies.
Turning to gender-specific differences, the scholars analyzed the distribution of exposure scores across occupations with different shares of female employees. They found that as the share of women within occupational groups increases, both AI and software exposure decreases. Thus, on average, women are less affected by the automation potentials of AI and software than men. A possible explanation for this is that the share of women is particularly high in occupations that often require extensive social, interpersonal, and communication skills (i.e. ‘’human skills’’), where the adoption of digital technologies has so far been relatively limited and where some bottlenecks to automation, even to AI, seem to persist.
Is AI the solution for the acute shortage of skilled workers?
The results demonstrate that AI and software alone could not completely take over the tasks for any of the “bottleneck” occupations and that technological change may reduce but cannot solve the acute shortage of skilled workers. Generally, the use of AI and software seem to play a greater role for some occupations with high shortages relative to occupations with lower or no shortages. However, the scholars observed a significant heterogeneity in exposure scores, particularly for occupations that require highly complex tasks. The exposure scores or geriatric care for example are very low (AI: 23%; Software: 29%), while the exposure potentials for occupations in software development are very high (AI: 88%; Software: 84).
Therefore, the scholars suggest that the observed technologies can play a role in mitigating the shortage of skilled workers. However, the results also demonstrate that technologies alone could not completely take over the tasks for any of the “bottleneck” occupations and that technological change may reduce but cannot solve the acute shortage of skilled workers.
