The discussion of the impact of AI on entry-level white-collar jobs has become somewhat cliché now, and with good reason: junior hiring is down 14% to 29% across the US, UK, Canada and Australia.

The most cited research explores whether generative AI is the main cause, which keeps putting the AI displacement story in the headlines.

But a new paper argues that bad news for entry level roles are likely misattributed to AI - something else could be driving what seems like its consequences for junior workers.

Economists Peter Lambert and Yannick Schindler point out that the occupations highest on AI exposure are broadly also the ones with highest exposure to the availability of working from home arrangements: software, finance, professional services, or accountancy.

When they include both shocks in their analysis, the WFH effect remains statistically significant; the AI effect does not .

What drives this is what the office does for workers still building up their human capital; the acquisition of expertise and know-how about their jobs.

Randomised trials of in-person versus remote work find office workers about 18% more productive than home workers, with two-thirds of the gap there from day one and the rest opening up because office workers learn faster.

Sitting near teammates raises the digital feedback engineers get on their code by 18%.

Some workers and companies are more productive when working from home is an option: hybrid work can be a big win for workers already established in their careers. They had their formative years in the office, and they now have the senior colleagues, the network and the tacit knowledge that let them be productive from home.

But there is now a cost that is borne almost entirely by the cohort that hasn't yet been given the same opportunities to build up their careers.

There is an uncomfortable implication in all of this. The workers with the most to gain from time in the office are also the ones with the least say in whether they get it. The current arrangement is largely determined by workers and companies that have adapted to hybrid forms of work, not those who may need the ability to acquire work experience the most to benefit from automation.

The argument from Part I of this puts this in sharper relief: AI struggles most with the parts of work that depend on tacit knowledge, the contextual judgement that has to be transferred in person.

https://britishprogress.substack.com/p/is-ai-the-new-breadwinner

Younger workers building human capital under hybrid are accumulating less of the kind of knowledge AI is worst at substituting for, and so will face a higher automation risk over time, as the technology evolves.

If the UK is to be prepared to capture and distribute the gains this new technology will bring, it needs to give young workers a better chance of harnessing these benefits.