AI and robots should create as many jobs as they displace in the long run
17 July 2018
Our latest UK Economic Outlook report projects relatively subdued UK growth over the next couple of years, in part due to the drag on business investment from ongoing uncertainty around Brexit. This will also keep UK house price inflation at modest levels over the next few years, with average property prices falling in London in the short term.
Looking beyond the next few years, however, the key longer term economic challenge for the UK is to boost productivity growth, which been anaemic ever since the financial crisis. Our past research suggests that Artificial Intelligence (AI) and related technologies like robots, drones and driverless vehicles could play a major role here, potentially boosting UK GDP by around 10% by 2030 if investment in these areas can be sustained at high levels.
A perennial concern about technologies like AI and robots, however, is that they will boost productivity at the cost of displacing many human workers. There is an element of truth in this, as indicated by our past research suggesting that up to 30% of existing UK jobs could be at potential high risk of automation by the mid-2030s.
But is not the whole truth for two main reasons, which we explore in detail in an article in our latest UK Economic Outlook report.
First, just because a job has high potential to be automated does not mean that this will definitely happen as there may be a variety of economic, political, regulatory and organisational factors that could block or at least significantly delay automation. Based on our probabilistic risk analysis, our central estimate is that only around 20% of existing UK jobs may actually be displaced by these technologies over the 20 years to 2037, although this still represents around 7 million jobs so it remains a very significant number. We refer to this as the ‘displacement effect’.
Second, and more importantly, AI and related technologies will also boost economic growth and so create many additional job opportunities, just like other past waves of technological innovation have done from steam engines to computers. In particular, AI systems and robots will boost productivity, reduce costs and improve the quality and range of products that companies can produce. To stay competitive, firms will ultimately have to pass most of these benefits on to consumers in the form of lower (quality-adjusted) prices, which will have the effect of increasing real income levels. This means that households can buy more with their money and, as a result, firms will need to hire additional workers to respond to the extra demand. We refer to this as the ‘income effect’, which offsets the displacement effect on jobs as summarised in the flow diagram below:
Our new research presents a PwC estimate of this income effect for the first time, with the conclusion that AI and related technologies should create broadly as many jobs as they displace in the UK over the next 20 years. In absolute terms, around 7 million existing UK jobs (c.20%) could be displaced, but around 7.2 million additional jobs could be created. The net impact therefore seems likely to be broadly neutral on jobs, but strongly positive for overall GDP given the boost that these technologies will provide to productivity.
As the chart below shows, however, the net impacts on jobs could vary widely by industry sector. We estimate that the most positive effect of AI will be seen in the health and social work sector, where employment could increase by nearly 1 million, a net rise equivalent to just over 20% of existing jobs in the sector. On the other hand, the number of jobs in the manufacturing sector could be reduced by around 25% due to further automation, representing a potential net loss of nearly 700,000 jobs.
In general, the industries that are likely to see a positive net effect on jobs are those requiring highly ‘human’ skills and those needing highly technical digital skills. For example, AI systems could replace many of the more routine aspects of a teacher’s job such as marking homework and would allow online learning that is personalised to a student’s abilities. But a ‘human touch’ would still be needed for personal coaching, supervising group work and teaching subjects like art, music, sport and drama. Healthcare will also continue to need a human touch, albeit one enhanced by AI systems and robots in many cases.
On the more technical side, specialist skills will be required to develop, sustain and explain innovations in AI. For example, as prediction becomes cheaper and more accurate due to machine learning, so expert human judgement to assess what action to take in the face of a given set of predictions will become more valuable.
The sectors where we expect to see net job losses from AI-related automation are those involving a high degree of repetitive and routine tasks. Automation has been taking place in the UK manufacturing sector for around 250 years and our analysis suggests this is likely to continue. What is new about AI, however, is that it is out-competing our minds as well as our muscles. So routine data analysis tasks in services sectors will also increasingly be automated and so could manual tasks like driving as autonomous vehicles roll out across the economy.
How can government policy maximise the gains and minimise the costs of AI?
While our central estimate is that the net effect of AI on jobs will be broadly neutral, there are many uncertain factors that could tip the balance towards a more optimistic or pessimistic scenario. Government can play a key role here in both mitigating the costs associated with the displacement effect and maximising the benefits from the income effect, so tipping the balance towards a more optimistic outcome.
The AI Sector Deal announced earlier this year is a great start here, provided it can be implemented in full. Government should also work with other stakeholders such as business, unions and educational providers to foster a flexible and adaptive workforce by supporting national retraining schemes for displaced workers and building ‘STEAM’ skills (where the ‘A’ stands for art and design as a key input to innovations like the iPhone).
Finally, we should not forget that even if the net effect of AI on jobs in the UK is neutral or even slightly positive, there will still be considerable disruption and redistribution. Government should strengthen the safety net for those who find it hard to adjust to technological change and use pilot schemes to start to build the evidence base on possible solutions ranging from lifelong learning accounts to more radical ideas such as universal basic income. Only in this way can the huge potential benefits of AI and related technologies be spread as widely as possible across society.