Work in the age of AI

Fears about the impact of new technologies on jobs are age-old – and with good reason. Truly transformational technologies reshape and destroy work – but also boost productivity, fuel growth and create many more jobs. That, at the least, is what the history of the last 300 years, since the start of the industrial revolution, shows.

So past technologies have, in aggregate and over time, increased employment. But in thinking about today’s hot technologies most of us incline to the belief that ‘this time it’s different’ – that the latest technology is so transformational that it will destroy more jobs than it creates.

Could AI be that technology?

Many people think so. Geoffrey Hinton, the Nobel prize-winning computer scientist who has been called the godfather of AI believes it will increase unemployment. Dario Amodi, CEO of Anthropic, has said that AI could eliminate half of white collar entry level jobs in the next five years. Bill Gates goes further and has said that advances in AI will, over time, mean that humans will no longer be needed “for most things” in the world.

AI has been around for a long time, but today’s frenzy really took off with the launch of ChatGPT in November 2022. We can see the effects of the AI boom in tech shares, in the behaviours of corporates and in the surge in spending on tech and data centres in the US. So almost three years in, is AI starting to destroy jobs?

Unemployment has certainly risen in the US and UK since 2022, albeit from low levels and consistent with softer economic activity. Weakness has been particularly apparent in the market for new graduates, which has been widely attributed, at least in part, to the impact of AI on entry-level jobs, especially in finance, technology and business services.

A recent Stanford research paper found that employment of younger workers in AI-exposed sectors of the US economy, including software developers and customer service representatives, had contracted by 6% since the introduction of ChatGPT while employment among older workers had risen by 7%.

We can see similar pattern in UK job vacancy data. My colleague Tom Avis has found that the number of job ads for the 20 UK sectors most exposed to AI have fallen by 47% since mid-2022 while ads for the least exposed sectors fell by 26%.

Yet it’s too early to say that we are on the verge of mass job destruction.

It is impossible, at least using whole-economy data, to prove causation, especially in a weakening job market. Poorer job prospects in sectors that are exposed to AI could reflect the wider economic slowdown, the impact of old-fashioned competition or a search for cost savings. Employment in call centres, travel agents and administrative and back office functions has been shrinking for years, with older technologies, such as robotic process automation or even the internet, the main drivers. They, rather than AI, probably still are. It is not yet clear that AI is driving widespread organisational change. Earlier this year, for instance, MIT reported widespread disappointment among US companies with the performance of AI pilot schemes.

Job destruction and creation are, in any case, an essential part of a dynamic economy. Job stasis, in which sectoral employment is forever fixed, would mean economic stasis.

In the last 20 years the UK has created an additional 4.9m net new jobs, comprising 2.2m jobs destroyed and 7.1m created. Large though they are, these numbers understate the true scale of change, since they relate to the net change in jobs under about 450 different job classifications but exclude job gains and losses within classifications.

It’s also worth remembering that, AI, like previous technologies, will create jobs. The Stanford paper mentioned earlier have found evidence of AI-related job growth in the US, for instance in cyber-security, AI ethics and machine learning engineering.

Even if we could accurately measure the scale of the AI effect on jobs, it would almost certainly be lost in the wider sweep of gains and losses.

History says that tech creates jobs. But AI’s growing capacity, and the possibility that it might reach beyond human intelligence, have stoked fears of mass unemployment. For now AI is nibbling at the edges of the job market. Only when AI is embedded in the workplace will we know whether this time it’s different.