Automation, the substitution of labour for capital in response to a shift in the relative cost and productivity of these factors of production, is a phenomenon as old as the late modern period.
Workers fearing the adverse effects of structural unemployment on livelihoods as major employment-providing industries have substituted labour for labour-saving technological innovations categorises the early 19th century textile industry in Nottingham as it does many of the occupations and industries under threat from Artificial Intelligence and related technologies (including robotics, drones and autonomous vehicles) today.
The potential impacts of AI on employment is certainly a hot topic at the moment, what with the rise in popularity of ChatGPT and concerns that it will be replacing a lot of people’s jobs – see also this Adzuna blog looking at the boom in AI-related jobs.
The impact of AI on specific jobs
Technological advances have both destructive and creative employment effects. Take one occupation that is seen as being highly susceptible to replacement by AI – legal assistants. This job involves taking in and then summarising large amounts of information that is highly structured and very language orientated, which is suitable for AI given the relative ease at which rules can be created to guide its work combined with its ability to process information at far greater speeds than a person can.
AI will thus reduce the need for legal assistants, assuming the volume of this work that needs performing stays the same, as it is able to perform those automatable tasks far more efficiently and at a lower cost. Legal assistants as a profession likely won’t be wiped out entirely though, just as the spinning jenny didn’t wipe out all employment in spinning yarn – highly skilled spinners could continue their trade as artisans by producing higher quality (real or perceived) hand-made wares.
The increased use of AI might see large numbers of legal clerks replaced, but there will still be certain tasks that they perform that can’t be replaced with AI, for instance client-based interactions.
These tasks will form a much larger proportion of the role – AI will thus not just affect the types of jobs available but also the tasks that make up roles. Work will continue to be centred more on original thought and creativity – something that humans can offer that AI and technology cant (for now) – and away from repetitive, rules-based tasks, both physical and cognitive, that can be performed efficiently by technology.
There will also still be demand for legal workers to guide the AI in performing the tasks that they used to be taken on by legal assistants – these jobs may well be filled by legal clerks who are able to understand and work with AI, just as people were still required to tend to the jennys.
Previous waves of labour-saving technological advancements were more focussed on replacing more physical tasks – the spinning jenny took on the task of spinning yarn, the telegraph took on the task of delivering messages, etc. – AI and other more recent technological advancements have had and will have more of an impact on cognitive tasks – processing large amount of information, copywriting, etc. As such, one of the likely effects of AI will be to accelerate the growth in the skills premium – the difference in earnings between more and less skilled labour.
Improving productivity and creating economic growth
Technological advancements including AI will also lead to the creation of additional employment by improving productivity and creating economic growth – increased overall incomes resulting from these advances will increase aggregate demand and therefore employment – Greg Brockman, CEO of OpenAI, as a result of AI has greater income to spend on hospitality, increasing demand for chefs in San Francisco for example.
PwC produced a report for the then Department for Business, Energy and Industrial Strategy in 2021 looking at the potential impact AI could have on employment in the UK. They estimate that “around 7% of existing UK jobs could face a high (over 70%) probability of automation over the next 5 years, rising to around 18% after 10 years and just under 30% after 20 years”. When considering overall net effects on employment, they suggest the most plausible based on historical trends would be a neutral long-term effect.
The charts below display some of the historical macroeconomic data that they are referring to – the Bank of England’s Millennium of Macroeconomic Data. Firstly, we plot on the right axis total employment in thousands against year-on-year percentage point changes in the employment rate and 5-year running average of the employment rates (to even out year-to-year fluctuations) on the left-axis from 1760-1860. This shows that, despite the significant labour-displacing technological advancements that occurred during this period, the overall employment rate remained broadly level whilst total employment steadily rose.
The second chart, showing total employment and the employment rate across 1861-2011, tells a similar story – a fairly level employment rate and fairly consistent growth in employment. Note that the use of and differences between the two graphs is because of a discontinuity in the dataset – it is still a work in progress, representing best guesses not official estimates.
In the face of the dramatic technological advancements across the 18th, 19th and 20th centuries, employment rates held fairly constant with the productivity and growth benefits allowing for unprecedented population, employment and income/ standard of living growth. So it seems fair to assume that AI and related technologies will have similar effects.
AI is further reducing the range of tasks that humans have a competitive advantage over technology in – the continual upskilling of the labour force we have seen since the industrial revolution has been necessitated by the need for humans to offer value over capital alternatives to their labour to incentivise their employment. Those that are left behind in terms of skills will continue to face serious disadvantages in the labour market, with both the magnitude of these disadvantages (gap in ability to find employment and quality of employment) and the proportion of individuals that are being left behind (the inflation in skills requirements outstripping increases in the prevalence of skills across the population as a whole) likely increasing.
Could AI inflict employment Armageddon?
This is what is perhaps what is driving the public fears about AI’s ability to inflict employment Armageddon, and one of the reasons for concern amongst policy makers regarding the potentially negative effects of AI. We should be acutely aware by now of the negative impacts on all of us of the existence of large inequalities in a society – the populist wave that rose up since the Financial Crisis has divided peoples and damaged democracies previously seen as strong.
The aforementioned PwC report shows that the net effect on employment will be most positive for professional and management occupations such as Nurses, medical practitioners and CEOs, whilst the most negative effect will be for administrative and process-oriented roles call centre and customer service workers, with a clear positive relationship between estimated effects on employment and average earnings. Differences in employment effects across occupations and industries naturally translate to differences in employment effects by geography given variations in the structure of the economy by region. PwC produce the following graph of the estimated net employment effect of AI by region.
According to this, AI will likely have a more positive net effect on employment in areas such as London and the South East and a more negative net employment on employment in parts of Northern England and the Midlands i.e., potentially worsening differences in employment between already more and less advantaged parts of the country.
So overall, it’s reasonable to assume that AI and related technologies will have an overall neutral effect on employment whilst boosting productivity, growth and incomes. However, the distributional impacts, across occupations, industries and areas, will be significant and will have significant negative impacts on certain groups which need to be mitigated – the further increases to the skills premium that AI will likely have means that our education system needs to do a better job of addressing inequalities in skill acquisition and promoting social mobility to avoid the adverse societal impacts associated with an already highly unequal economy becoming even more so. For a recent detailed summary of the research and discussion around the potential impact of AI on the labour market, see this House of Commons briefing.
Want to learn more about how AI is impacting the job market? Speak to our Adzuna Intelligence team to get all the insights
Read more: Speaking of AI; OpenAI and GenerativeAI jobs are booming
Written by Daniel Muir, a Research Economist (Fellow) at IES.
IES is an independent, apolitical, international centre of research and consultancy in public employment policy and HR management. It works closely with employers in all sectors, government departments, agencies, professional bodies and associations. IES is a focus of knowledge and practical experience in employment and training policy, the operation of labour markets, and HR planning and development. IES is a not-for-profit organisation.