Image generated by ChatGPT 
A relative of mine, a young man, recently completed his MSc in AI at a top EU university. One would have expected that the world would be at his feet and that he would have a wide array of diverse job offers from which to choose.
It was not so.
He (and his graduating cohort) had carefully developed articulate and attractive CVs and sent them off to companies ranging from big-brand corporates to NGOs to smaller entrepreneur-driven projects. The response was underwhelming, if there was any response at all.
The bottleneck in job applications these days is LinkedIn, which has become the primary (and sometimes the only) place for companies to recruit. Given the ease with which AI tools can be set to task, job seekers can instantly CV to every single company in the world with an appropriate job opening. The result is that there is now a vexing problem. Every recruitment executive is receiving hundreds (and sometimes thousands) of applications for every job opening.
Recruiters are now being forced to turn to AI to filter applications. The fine drama of sitting across a desk from a job applicant is thinning, nuance is being lost in the automation, and hiring is being submerged under algorithmic dictates.
To add insult to injury (actually, to add injury to injury) AI has started to colonise increasing job categories, especially at entry-level, a matter to which I will return presently.
My relative went through some anxious months. Six years of continuous study in a technologically critical and booming field, all for nought. Life’s decisions being questioned, depression looming on the horizon.
He did finally land a terrific job. But it was not through LinkedIn. It was through a friend of a friend who gave him a number, which he called, and asked for an interview, which was granted. It was old-style hiring (except that the first few interviews were on Zoom, only the final interview was face-to-face). Not all of his cohort have been so lucky. Some are still looking.
Every week brings a new paper or study or post or headline about “the future of work”. As new technologies (especially AI and robotics) move from research to development to production to commercialisation, every researcher, analyst or commentator from academia to TikTok tries to forecast the impact on jobs and employment.
Why? Because everyone has a dog in that race. But facts are hard to extract, data is thin and the landscape shifts weekly. Worse, sensational headlines dominate the public square and skew public opinion.
So it was with great interest that I came upon a very recent podcast from the folks at Your Undivided Attention that featured two academics who are steeped in these matters and who have access to the latest data. These are Ethan Mollick, associate professor at the Wharton School of the University of Pennsylvania, where he studies innovation, entrepreneurship, and the future of work, and Molly Kinder, a senior fellow at the Brookings Institution, where she is leading a multi-year project exploring how generative AI is transforming work. She recently led research with the Yale Budget Lab examining AI's real-time impact on the labour market.
The podcast is essentially a collective deep breath for anyone who has been doom scrolling their way into a career crisis. The host Daniel Barcay sets the mood early: AI has injected a kind of ambient panic into ordinary life planning — what to study, which skills to build, whether to buy a house, even whether it’s sensible to assume there’ll be a recognisable rung on the career ladder in five years. We’re stuck between two unhelpful extremes: the “AI will unlock endless productivity” camp and the “job apocalypse is already here” camp. So the show’s mission is refreshingly practical: what does the labour market actually show now – what is the real data?
Molly Kinder’s Yale Budget Lab work is the backbone of the reality check. She and colleagues went hunting for economy-wide signs of disruption since ChatGPT’s launch three years ago. The headline, to many people’s surprise, is stability more than chaos. When you zoom out across the whole economy and sort jobs by AI exposure, you don’t see the occupational mix dramatically shifting away from roles highly exposed to AI. That doesn’t mean nobody has been hurt — some people in “creative” industries, software developers and customer service roles may already feel pressure — but her methodology is deliberately designed to catch a house fire, not a small kitchen flare-up.
The important nuance is the early-career caveat. Kinder flags more churn among the youngest workers and nods to an earlier study by Stanford, titled “Canaries in the Coal Mine”, which suggests early warnings for people 25 and under in AI-exposed roles. She’s careful not to claim a clean causal line from AI to all-youth-employment pain, pointing to other plausible culprits: interest rates, tariffs, a shaky macro outlook, and the post-pandemic whiplash of over-hiring in tech. Her position isn’t “AI isn’t doing this”; it’s “AI is probably part of the mix, but don’t pretend it’s the whole story yet.”
Mollick comes at the same landscape from the shop floor rather than the satellite view. He says macro data does not show massive displacement — not because AI isn’t powerful, but because adoption and organisational redesign take time. He argues that AI now overlaps meaningfully with highly educated, creative, well-paid work, which means change is a “minimum” outcome. But he worries about the speed at which these systems are getting better. We may be OK now, but forced transformation is, at some point, a certainty. There is just not enough data to pinpoint when.
They converge on a crucial idea: AI exposure isn’t the same as usage. Adoption is wildly uneven across sectors, a phenomenon for which he has coined a widely used phrase – the “jagged frontier”. In some jobs, AI is frictionless; in others, privacy fears, regulation, and institutional caution slow everything down. That lag helps explain why the labour-market scoreboard still looks calmer than the headlines.
The episode’s most useful tension is moral rather than technical: incentives. With weak union density in many highly exposed sectors, workers may have limited power to negotiate gain-sharing with the companies that will use AI and increase their profits. Mollick adds a hopeful counterweight: this isn’t destiny. Companies can choose different models — and he wants more deliberate efforts to design “humans + AI” systems that outperform AI alone.
Their advice to new entrants is simple and bracing: aim for bundled, human relationship-rich work; avoid narrow single-task roles; lean into being both human and fluent with AI. The vibe is neither panic nor complacency. Her message is: we’re not in the apocalypse today - which is precisely why this is the moment to shape what comes next. DM