Nice work

Paid work fills up to a third of the waking hours of people with jobs. Hence the growing focus on what influences the quality of that time – what makes a ‘good job’ – as well as how the world of work is changing in the face of AI, net-zero goals and shifting demographics.

‘Good work. Good jobs. Britain working again.’ The UK’s new chancellor of the exchequer, John Healey, put good jobs at the centre of his speech at the Labour Party conference last month.

But what makes a job good? People in work spend up to a third of their waking hours doing it, and the average adult in the UK can expect around 68,000 hours of paid work over a lifetime. Whether those hours are ‘good’ depends on more than pay: the wellbeing gap between people in average and very poor jobs is as large as the gap between people in good and poor health.

A new collection of five Economics Observatory articles, produced with the Huth Initiative for a New Political Economy at the London School of Economics (LSE), examines what makes a job good, where good jobs are found, and how artificial intelligence (AI) and the objective of net-zero emissions of greenhouse gases are changing them. Together, these pieces suggest that good work is designed rather than delivered by economic growth, and that the people doing the work should have a say in that design.

Work is a big part of our lives

Paid work is one of the biggest uses of people’s time. Combining data from the Office for National Statistics’ (ONS) time use survey with national life tables, a typical adult in the UK can expect to spend about one in five of their waking hours from age 18 onwards in paid work (see Figure 1).

This average includes years spent studying, out of work and in retirement. For people aged 25-54, the share rises to 29%; for those in a job, it is 32%.

Figure 1: How adults spend their waking lives

Source: ONS Online Time Use Survey (March 2024); ONS National Life Tables (2022-24); author’s calculations.
Note: expected years from age 18, averaged across all adults including those not in work. Excludes sleep.

Work is also where many adults spend their time with other people. In the United States (for which there are detailed data) people aged 25-54 spend nearly three hours a day with co-workers, compared with about 40 minutes with friends (see Figure 2). In England, people who are unemployed are three times as likely as those in work to say that they feel lonely often or always (15% versus 5%).

Figure 2: In mid-life, Americans spend far more time with co-workers than friends

Source: US BLS, American Time Use Survey (2010-24), via Our World in Data

Yet many people get little from their jobs. Only 10% of UK employees say that they are engaged at work, compared with 12% across Europe and 20% globally (Gallup, 2026).

Growth alone does not make jobs better

Pay is only one part of a good job. In an Observatory article first posted back in July, Francis Green (University College London) sets out six more dimensions to measure job quality: prospects (security and the chance of promotion); working time (including control over schedules); autonomy and skill use; work intensity; the social environment; and the physical environment.

Economic growth can lift pay but it is unrelated to four of the six non-pay dimensions. Better jobs depend instead on how organisations are managed, how they introduce new technologies and whether employees have a say in how things are done.

The next piece in the series, on the geography of good jobs, reinforces this point. Henry Overman and Aadya Bahl (both LSE) explain why the UK’s highest-paying jobs cluster in London, the South East and a few cities. These locations draw in graduates and firms that want to employ them. In 2019, around half of working-age adults in London and Brighton were graduates, compared with 15-18% in Doncaster, Grimsby and Mansfield. High-paying jobs are even more concentrated.

Training more graduates elsewhere will not reverse this on its own. Greater Manchester, for example, keeps only about half of those who study there. The authors argue for coordinated investment in jobs, housing and transport, focused at first on a few places. Fair employment policies can raise pay in lower-paid work, but they cannot by themselves create high-quality jobs in struggling areas.

The move to net zero shows the same pattern in a new industry. Sam Fankhauser (University of Oxford) finds that net-zero industries already employ more people worldwide than fossil fuels (35 million versus 32 million in 2022). Clean energy jobs often pay more, but the premium largely disappears once firm size, unionisation and working conditions are considered, suggesting that green jobs are not necessarily better by nature.

AI could enrich work or hollow it out

AI is the biggest current test of whether work gets better or worse, and its use is spreading quickly. In September 2026, 30% of UK businesses used at least one AI technology, up from 10% three years earlier (ONS, 2026). Workers are incorporating AI quicker than employers – 55% of UK employees reported using AI for work or education (ONS, 2026).

To learn more about how AI is affecting work, the next piece in our series – by Kate Vredenburgh – explores the use of AI within IT and data science jobs. She finds that workers using AI are more productive and could spend more time on tasks they find interesting. But many also fear becoming a ‘babysitter for AI’, checking output rather than producing it, even though no manager has taken any power away.

For Kate, what’s at stake is workers’ standing – being the author of their work and the person who answers for it. Keeping a ‘human in the loop’, the main safeguard in the European Union’s AI Act, can leave a worker with oversight but no authorship.

Of course, autonomy in work was falling before AI arrived. UK workers have less control over their day-to-day tasks than they used to. In 2012, 44% of employees had a great deal of say over what they did and how they did it; by 2024, the share had fallen to 34%. The share reporting real influence over wider decisions affecting their work is now lower than in 1992 (Gallie et al, 2025).

This matters because people’s sense that their work is meaningful is more closely linked to control over one’s work, skill use and supportive managers than to pay (Zhou et al, 2025).

It is also true that AI can be used both to collaborate or delegate. How people use it varies. Data from AI company Anthropic on conversations with its Claude chatbot distinguish between ‘augmentation’, where a person works with the AI and iterates on its output, and ‘automation’, where a person hands over a task with little further input. In November 2025, the split was close to even: 52% of conversations were augmentation and 45% automation. When businesses connect AI directly into their systems, around three-quarters of use is automation (Anthropic, Jan 2026).

The design of the tool matters as much as the task. A typical chat conversation that produces a blog post involves 13 rounds of back-and-forth; a typical session producing the same output in Claude Code, an agentic tool that can work on its own for hours, involves a single prompt. Most of the gap in autonomy between the two comes from people delegating the same tasks more, not from doing different work (Anthropic, June 2026).

Workers may not feel the cost straight away. In a survey of Claude users, those who delegate the most are the most optimistic about their future pay and job prospects – although, as the report notes, skills can erode even as they become more valuable.

This difference might be showing up in the labour market. In the United States, employment of people aged 22-25 in AI-exposed jobs is 19% below where it would have been had it kept pace with less-exposed jobs. The declines are concentrated in occupations where AI mostly substitutes for human tasks, while employment is flat or rising where it mostly complements workers (Brynjolfsson et al, 2026).

As Kate points out in her article, augmentation keeps the worker as author, while automation risks leaving them with oversight alone. Which of these prevails will depend on how employers design jobs and tools, and on whether workers have a voice in those decisions.

If AI displaces workers, where should they go?

People who lose their jobs to AI lose more than a wage. In her Observatory article, Anca Gheaus (Central European University) explains that work provides goods that money cannot buy: excellence, a contribution to society, community and recognition. Caring for others offers all of these, and it cannot, and should not, be fully automated.

The need is large. Unpaid housework and care already take up slightly more of the average UK adult’s life than paid work (8.1 years compared to 7.8 – see Figure 1). England's adult social care sector has around 96,000 vacancies on any given day (Skills for Care, 2026). And health and education are among the sectors where workers are most likely to find their jobs meaningful (Zhou et al, 2025).

But care work today is often rushed and micro-managed, which harms carers and the people they look after. Anca argues that moving displaced workers into care would only work if care were to become good work, with manageable workloads and proper training. Paying for enough care of that quality would mean sharing the gains from AI more widely.

Good work is a political choice

The five articles in this collection reach a set of clear conclusions. Growth raises pay but does not reliably bring autonomy, job security or manageable workloads. High-paying jobs cluster, and their benefits do not spread far by themselves. And AI can make a job richer or emptier depending on how it is introduced.

In each case, the quality of work is shaped by decisions – how managers design roles, which tools firms adopt and what governments regulate and fund. That makes good jobs a question of political economy as much as economics: who makes those decisions and who shares the gains?

Francis Green and Kate Vredenburgh both argue that workers should have a say in how new technology enters the workplace, while Anca Gheaus points out that the gains from AI must be shared if care is to become good work. With a third of working people’s waking hours at stake, these choices deserve as much attention as the pace of growth itself.