A focus on wellbeing at work can help employees thrive.

Christopher Pissarides writes:

How is artificial intelligence (AI) changing the structure of the economy, and what does that mean for economic growth?

In an open market economy, when a new technology prompts structural change, market forces will push the economy back to a new equilibrium. The new equilibrium will have different types of jobs, and jobs will be reallocated across sectors - some growing, some shrinking.

The faster the new equilibrium is reached, the better off we are likely to be. But the speed of adjustment itself can be costly, so you have to balance out that speed with the time that it takes you to get there.

Adjustment is usually slow - typically taking decades. For example, the countries of eastern Europe - formerly planned economies - opened up their markets in 1990 and if you look at the countries of the European Union, the former planned economies are the least productive and are typically at the bottom of the EU in terms of GDP per capita. That slow adjustment is due to frictions, in particular:

• Information - what does the technology shock imply for the labour market?
• Location - where are new jobs being created?
• Skills - what skills do workers need to adapt to the new technology?

To understand the relative importance of these frictions for AI, we need to understand the two ways in which it affects jobs. The first is through replacing workers directly. For example, senior lawyers using large language models like ChatGPT rather than having paralegals write reports.

The second is through people leaving sectors or not entering them as young people. This is not because the jobs that they were doing were taken over by AI, but simply because an AI-induced rise in productivity means there's no more demand for the new output that is coming out of those sectors, but there's a lot of demand elsewhere in the economy.

Empirically, it's tricky to distinguish between these two effects. The first is one that obviously attracts the most attention - how many jobs will be taken over? But the second (known as the Baumol effect after the economist who first described it, in 1967) is due to productivity growth.

Subsequently, Rachel Ngai and I have shown that if productivity growth rates across sectors of the economy are different, then you're going to get a movement of workers from the sectors that are growing fast to the sectors that are growing more slowly. That's why you see jobs growth in the health, care, hospitality and other services sectors - the ones that are not experiencing as much productivity growth as elsewhere in the economy.

Reducing frictions to speed AI adoption

There are disagreements about the likely impact of AI on jobs and society. But hardly anyone talks about the second effect although I think eventually it's going to be the dominant one. While some technologists have been very negative about the effects and economists are divided, I am optimistic. I think that whatever happens in the long run with AI, the best we can do now is to embrace it, benefit from it and prosper - improving the quality of our jobs and improving productivity.

To adjust faster we need to identify how frictions are affecting AI adoption. Information and location frictions are important, but the biggest friction is skills - the need to reskill and upskill.

Looking at information frictions first, in our research at the Institute for the Future of Work, we asked workers what could improve conditions now that these new technologies are coming? They first mentioned better communication with managers and subordinates, more transparency about company policy and better social relations with colleagues. Only then do they talk about more time flexibility, including home working and a four-day week, and finally, more money and improved conditions.

Outside London, non-graduates with digital skills are in high demand

The first three are obviously about information - workers are not well informed about technologies and they're worried. Obtaining information about the company's intentions and knowing how to use it is key for these workers. We learned that the best way to communicate internally is through informal contact, not being called in by the boss for a talk - they hate that - but joining them for coffee or lunch breaks.

What about location? Location is also a problem because those who are doing the research and the new start-ups should be concentrated in a small number of places, to benefit from what you might call agglomeration externalities - for example, exchanging ideas or synergies.

There are benefits to this kind of activity from firms being located close together and if you look at the two most successful countries in AI research, the United States and China, all the technical research producing AI is concentrated in three places in both countries: California (Silicon Valley), Boston (Harvard, MIT) and Seattle (Microsoft); and Beijing, Shanghai-Hangzhou (DeepSeek and Ant Group) and Shenzhen (Huawei and Tencent).

In the UK, activity is concentrated in what is known as the golden triangle of London, Oxford and Cambridge. In Europe, it is scattered, which I think is the main reason for the continent's failure to keep up with China and the United States, even though overall investment in the EU plus Britain is at a similar level to China.

The United States is by far the biggest investor: 60% of global venture capital investments that go to AI research are in those three US locations. China invests only about 12% of global resources, as does the whole of Europe - though in the latter, there is no agglomeration.

Keeping your skills options open

But the main friction is skill. In previous industrial revolutions, this was not a problem. In fact, in the very first industrial revolution, workers had been skilled craftspeople and had to downskill to take the jobs in factories.

Now we need new skills. The main ones for the world of AI are related to understanding data dynamics and learning how to communicate the results.

For those leaving the technology sectors to work in other services, such as care or hospitality, communication is very important. They will need empathy rather than Python.

In schools, traditional subjects are still essential. For AI, there will be a shift to STEM (science, technology, engineering and mathematics) subjects. But I also think that schools, and even universities, should not specialise too much. Unless you're really going to work in AI research, it's better not to specialise because technology is changing fast and in unknown directions - what is important today may not be important tomorrow. What's more, AI is getting better at STEM tasks and might take the job away from you.

What is important in school is learning how to learn. I'd recommend acquiring general knowledge in STEAM where the A stands for arts, or even, if we want to include economics: STEEM.

Lifelong learning will become more important than ever. Successful companies make time for employees to learn but workers need incentives to take this up. This is where wellbeing comes in - if workers have good jobs, then they'll want to learn how to do them better.

So, provide good jobs, ones in which employees are able to strike a balance between work and family. I am convinced that if you provide jobs that improve the wellbeing of the workforce and at the same time you let them "own" the training, put it up online or offer alternatives, then they are much more likely to take up the offer. The company will be able to implement the new technology faster and the workers taking advantage of it will mean productivity rising faster.

Mary O'Mahony writes:

How do digital skills vary across the UK? How are these skills developed through the education system and training programmes? And what are firms doing to meet their demands for those skills?

At The Productivity Institute, we divide digital skills into three groups:

• Developer skills - programming, advanced AI and data analytics skills.
• User skills - the kind of digital skills needed to use software such as Salesforce.
• Basic skills - such as being able to use Excel, skills that are still very much in demand.

Looking at the share of digital technical skills in job adverts in 2022, we find that the demand for basic and user skills is spread out across the country. But the demand for developer skills is concentrated in the golden triangle of London, Oxford and Cambridge.

We also find that demand for basic and user skills has been growing steadily for some time, but demand for developer skills is different: it's growing in London and the south-east but few other places.

For around two in five adverts, we could identify whether the hirer wanted graduates or non-graduates. Looking at the share of job adverts wanting developer skills, we find that in London, a high proportion wanted graduates, whereas the demand for non-graduates with similar skills is more spread out. This shows that firms across the country need these skills - and they’ll often use non-graduates instead of graduates to fill those posts.

Looking at the earnings of people with various types of skills, there's a big wage premium for graduates over non-graduate developers. But that premium has been declining over time - and earnings for non-graduates with user or basic digital skills have been rising rapidly.

Beyond the golden triangle

On the supply side, we analyse Longitudinal Educational Outcomes data and for any one year, we calculate the proportion of those graduating with STEM or digital skills at three different levels: schooling, further education and university. The kind of skills are maths and logical skills; computing and programming skills; science and engineering; and applied digital skills. We look at those who do well: A* to B at A-level or a 2.1 degree or above.

The data allow us to map the supply of digital skills across England. The highest concentration is in the golden triangle and in Leamington Spa, a hotspot for the games industry and close to the University of Warwick. Then there are some other areas where there's a lower concentration, but the supply is still above average, such as Birmingham and Manchester. But large areas of the north-east, Yorkshire and the East Midlands have below average supply.

Technology is changing so fast, that becoming too specialised in certain skills is risky

The preliminary results of putting demand and supply together can be summarised as follows:

• The high supply/high demand areas are mainly in London and the south-east, plus a few more places like Birmingham and Manchester.
• The high supply/low demand areas are mainly rural areas.
• Then there are areas where there's a high demand by firms for digital skills but the supply is low. These areas have to rely on people already within the area to provide those skills. They are where skills shortages really bite, and include cities such as Newcastle, Nottingham, Portsmouth and Sheffield.
• Finally, there are low supply/low demand areas in the north-east and the north Kent coast, where low skills are just one of many issues.

How do firms get the skills they need in the face of shortages? There are two ways: through the education system and through firms providing training.

Schooling is very important. You need a good foundation of STEM skills and good literacy skills in schools. In places like Hull, Lincoln, Liverpool and Sheffield, the share of pupils achieving high GCSE grades in maths, science or ICT (or studying these subjects in further education) is about 75% of the London level.

There is a huge variation across the country, and this is an important area for public policy. There is much evidence of the importance of digital skills below the degree level, which highlights the importance of further education colleges.

Supporting employees in work and life

In the meantime, firms are demanding digital skills. The Department for Education's Employer Skills Surveys (ESS) show a long-term downward trend in expenditure on training by firms and average number of days per trainee. But there is a recent positive trend in online training: one ESS statistic says two-thirds of employers in England had arranged some online training in 2022 compared with just over half in 2017.

There's a well-known argument that firms will under-invest in general skills training because when they invest, the worker could leave for a competitor. But online programmes have an advantage over traditional training in that courses are less costly and they reduce the opportunity cost of the time a person spends away from the workplace in training.

Another way that firms can gain digital skills is to use digital platforms - that's a cost-effective way of getting the talent you need. Around half of firms outsource some IT services, and this has been growing rapidly. Many on such platforms are young people who are gaining experience by doing this work and then getting full-time jobs.

Digital skills and attitudes to work

How are digital skills affecting values and work practices? This is something that social psychologists are emphasising. It's often young people who have the most advanced and most up-to-date digital skills - so what happens if there's a problem with young people?

Commentators Jonathan Haidt (2024) and Jean Twenge (2023) have both warned that digital technology has caused some change in the attitudes of young people and the way that they work. This work suggests that there is an issue with young people having more mental health problems than previous generations and something needs to be done.

Our own work suggests that Gen Z (born 1995 to 2012) are much more likely to value leisure time, which may be a good thing. But one of the most important transitions that is going to happen over the next 30 years is the ageing society. And if young people are not so engaged in work, that's a problem, as they are the ones with the digital skills that the economy needs.

---

21 October 2025     Paper Number CEPCP712

Download PDF - Skills in the age of AI

Download Press Release

This CentrePiece article is published under the centre's Labour programme.

This publication comes under the following theme: Labour market dynamics, Work and wellbeing