Who’s winning the power battle between workers, automation and AI?

Afternoon all,

Worried that AI is thinning our language and destroying our ability to learn? Well I have good news – not yet. It was a pleasure to host the launches for two fantastic books this week, both rich in humanity, story-telling and knowledge. Soumaya Keynes and Chad Brown have written the definitive guide to trade that will take you from zero to ready to win the trade war we inevitably find ourselves in. Sarah O’Connor has got her shoes dirty looking beyond the hype and telling the tales of how the nature of work is changing – from translators to Amazon workers. Both books are a call to arms for policy-makers to face these big economic trends more strategically, and to do so with agency not inevitability. Inspired by the books, more on trade, AI and worker power in this week’s top of the charts.

Have a great weekend,

Ruth

Chief Executive
Resolution Foundation

Coin toss. What do you do when you’re short on international trade data? Follow the coins. The authors of this paper use data on hundreds of thousands of coins found across Europe, North Africa and the Middle East to track the evolution of trade routes in the first millennium. The ancient coins and novel methods support what has long been theorised: that the fifth century ‘fall’ of Rome didn’t result in the immediate demise of Roman institutions and commerce. The big disruptor to Mediterranean trade was the rise of Islam in the seventh century, which led to the collapse of north-south Mediterranean crossings, even as south-south trade thrived.

Novel niches. One of the less well understood parts of the automation puzzle is how and when new work is created. This paper digs into US census data to understand how the creation of novel roles differs from just more existing work. Around one-in-five American workers today hold jobs that didn’t exist before 1970 (from “Artificial Intelligence Specialist” in 2000 to “conference planner” in 1990). The authors find that these new roles are genuinely distinct: they are disproportionately filled by younger, more educated workers and command significant wage premiums that persist beyond workers’ initial entry into these fields. Not everyone has benefited equally, with White workers and women among those most likely to have these new job titles. What does this mean for us as we face a new wave of AI-driven automation? We do often invent new corners of human expertise as old ones fade, but being an economy that provides those new skills is critical. And, as Sarah’s book reminds us, we should choose which jobs we want the robots to do not just what they can do.

Stepping stones. A new report from the Centre for the New Midlands evaluates the Live and Work model that treats housing and employment in tandem. In 2015, St Basils opened Apprentice House in Sandwell: 32 rooms at £42 a week, priced so young apprentices could pay rent from their wages, which often aren’t near high enough for the private rented sector. Ten years on, the success speaks for itself. Every £1 invested generated £8 in social and economic value, with 236 young people housed and supported into work. Nearly a quarter of leavers moved into the private rented sector, a destination previously out of reach. In the West Midlands, where youth unemployment has hit 19 per cent and 18-25-year-olds now spend half their net pay on rent, the model’s logic is hard to argue with. The question is why there are still only 166 rooms of it.

Talking to the void. America’s youth mental health crisis has quietly acquired an unofficial first responder, according to this study: the AI chatbot. The authors found that almost one-in-five US teenagers and young adults asked AI for mental health support last year, up from one-in-eight a year earlier. The teens aren’t alone in this – new data suggests that companionship and therapy remains the top use case for AI. Young women were more likely to turn to AI for help than young men, and among those seeking advice Black youth were over five times more likely than White youth to ask for help regularly. Over 90 per cent of these young people said the chatbot’s advice was helpful. As the authors point out, this overwhelmingly positive response could reflect AI’s well-documented tendency towards sycophancy rather than its ability to give sound guidance to vulnerable teens. Worryingly, nearly two-thirds of these young users reported not telling anyone they had turned to AI for mental health support, leaving the bots to provide their service unchecked.

Something for the weekend | FIFAnomics

Next week the World Cup kicks off. I’ll spare you my profound knowledge of the tactics and formations that could lead England to glory and instead focus on the economics behind the beautiful game. They say money increasingly dominates sport, but it’s not necessarily a route to World Cup glory. Brazil has won more World Cups (five) than any other country, despite a relatively modest GDP per capita of just $12,300 in 2026. Germany takes the big money, big trophy cabinet double with four titles and a far rosier $65,300 GDP per head. FIFA, predictably, talks a big talk about the pleasures of hosting, promising a combined $30.5bn GDP boost to this year’s three host nations, with Mexico set to be the main relative winner. The governing body itself is set to rake in an extra $9bn this summer, with broadcast rights the big money spinner, followed by hospitality and ticket sales. But it’s not all gravy: Qatar splashed a staggering $220bn on the 2022 tournament to build up its infrastructure, tourism and hospitality sectors, more than its entire annual GDP.

Chart of the week

Hot off the back of our worker power briefing note last week, we thought we’d give a fresh update to an RF classic. This chart tracks unionisation rates across 12 five-year birth cohorts of people in employment. The oldest cohort with sufficient membership data covers those born in 1951-1955, observed from age 46 to 64; the youngest, born in 2000-2005, is tracked from age 16 to 20. Membership rises sharply as cohorts move through their twenties, plateaus around age 30, then falls away from age 55. This is a story of structural decline, with no successive cohorts reaching the peaks of those born in the 1950s. But the latest data hints at least at a slowing in that decline: the 1981-85 cohort has ticked up in recent years, while the two cohorts just after (born 1986-90 and 1991-95) have tracked it closely rather than sinking even further. And although it’s too soon to say if this will last, the youngest workers born around the turn of the millennium are almost as likely to be unionised as their predecessors. Whether this continues will be important not just for unions, but for wider workplace power imbalances in the age of automation and AI.