Your Copilot Just Handed You Your P45 (or Pink Slip)?
"We should look at AI much more as a co-pilot than something which is necessary going to replace someone's job. AI is a tool that can help almost anybody do their jobs better, faster, quicker” UK Prime Minister, Rishi Sunak
We read a similar point of view in the FT recently, where Microsoft’s Sarah Bird (who ironically works on its Copilot product), agreed that GenAI “is absolutely going to change the way people work [by] removing the drudgery, removing the tasks that you didn’t like doing, anyway — allowing everybody to focus on the part where they’re adding their unique differentiation, adding their special element to it, rather than the part that was just repeated and is something that AI can learn”.
We are not sure that the quote from former UK Prime Minster Rishi Sunak or indeed that from Sarah Bird are going to age well. And nor will that other sound-bite that “AI won’t replace people—but people who use AI will replace people who don’t”. The first claims that AI is nothing more than an intelligent assistant and therefore poses no threat to employment. The second insists that workers will not lose their jobs to algorithms but to humans who have mastered those algorithms. Both statements contain an intuitive appeal; neither survives intact when viewed through the lens of growing labor‑market evidence.
Perhaps the quote should be ‘AI won’t replace all people … but two people using AI will replace ten people who don’t … until AI doesn’t need those two anymore’. Although we have been discussing the dangers of substitution for the last couple of years, what makes it interesting for a topical Substack post, is that we are now beginning to see some early indicators that perhaps suggest the wheels are beginning to – if not come off exactly, then at least look decidedly wobbly.
In a recent New York Times article Has the Decline of Knowledge Work Begun, Noam Scheiber suggested that “over the past few years, [white collar workers] have seen a steeper rise in unemployment than other groups, and slower wage growth”. An whilst AI is certainly not the only factor roiling the US jobs market at the moment (the less said about that the better!), it may well be a contributing factor that will only grow in significance. As companies continually seek to make themselves leaner and more efficient, AI is offering a route to greater profitability.
“We are seeing a meaningful transition in the way work is done in the white-collar world […] I tell people a wave is coming” Carl Tannenbaum, Chief Economist of Northern Trust
And the NYT is not alone in starting to (finally) raise a flag. Dario Amodei — CEO of Anthropic, stated in a May 2025 interview with Axios that AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years, and that AI companies and government need to stop "sugar-coating" what's coming.
Right now, he says, (reflecting Sarah Bird’s point) AI models are being used mainly for augmentation, thereby freeing people up to do high-level tasks while the AI does the rote work BUT that AI use in companies will tip more and more toward automation — substituting for humans. "It's going to happen in a small amount of time — as little as a couple of years or less."
In an interesting aside in the same article, apparently Axios asks managers to explain why AI can’t / won't be doing a specific job before green-lighting human hiring. Similarly, in an HR Drive article Tech hiring managers say layoffs are coming — and workers who can be replaced by AI will be first to go. (April 2025), Senior Editor Kathryn Moody writes that, like Axios, Shopify and Fiverr also require teams to prove AI can’t do the job before hiring a human to fill a role. Few want to admit this publicly, but every CEO is or will soon be doing this.
This displacement risk scales with the percentage of daily tasks already doable by GPT-4-level models.[i] Author and Educator Josh Bersin argues that hiring freezes and wage deceleration for degree-holders amount to a white-collar recession; and cites ADP data showing jobs requiring advanced degrees are now the slowest-growing category; and Economic Policy Institute figures that lower-wage work is seeing 3× faster wage growth than professional roles.[ii]
Bersin’s assertion of a white-collar recession is supported by findings in a World Economic Forum piece[iii] that draws on its Future of Jobs Report 2025. It claims that 40 % of employers expect to shrink headcount where AI can automate tasks; and cites Bloomberg analysis that AI could already automate >50 % of tasks done by market-research analysts and sales reps. It also reports a Gen Z survey showing almost half believe AI has devalued their degrees.
For most of the past three decades, the income premium attached to tertiary education was both narrative and statistical fact. Economist David Autor famously described a “college‑for‑all” escalator that delivered higher wages and lower unemployment to graduates almost irrespective of discipline. That escalator is slowing. Credentials remain valuable, but they no longer guarantee a foothold when the first duties entrusted to a newcomer are exactly those most readily codified into software.
So where are we on the augmentation vs automation (enabling vs substitution) continuum? In his excellent book, A World Without Work, Oxford economist Daniel Susskind frames the differences between frictional and structural technological unemployment, drawing a sharp line between these two stages of technology-driven job loss.
In the first, frictional technological unemployment, work still exists for people but mismatches keep them from taking it. Skills fall out of date faster than they can be replenished; new roles emerge in different cities or sectors; and, just as crucial, many roles clash with workers’ sense of identity. The result is a labor market full of vacancies alongside swelling queues at the job center—an allocation problem, not an absolute shortage of work. A situation that may slowly right itself over time for many if not all affected.
But if automation continues to encroach, Susskind warns, economies cross a more daunting frontier: structural technological unemployment. Here the machines are not just displacing particular roles and workers—they are eroding total demand for human labor. Training schemes and relocation grants no longer help because the jobs themselves have vanished or been priced beyond human reach. Society must then decide how to share productivity gains when wages cease to be the primary distribution mechanism.
Locate the emerging data on the frictional to structural spectrum and a pattern emerges: the ground beneath white-collar work is less solid, fissures are emerging. Susskind’s two-step model matters because it tells policymakers and firms which levers still work. While we are in the frictional phase, investment in skills and mobility can still pay off. Once the problem turns structural, bolder measures becomes unavoidable for everyone involved.
In his NYT article Scheiber doesn’t declare knowledge work “doomed,” but he argues that recent lay-offs, weak wage growth and an unusually steep rise in graduate unemployment together amount to the first persuasive hints that the white-collar labor market may be entering the frictional phase of technological upheaval – and could yet tip into something more structural if AI continues its rapid advance. And we haven’t heard anybody argue that AI has reached its apogee!
The asymmetry between augmentation (for now) at the top and displacement at the bottom calls for a revised policy toolkit. What ‘sensors in the ground’ do we need to get an early warning that the job market’s tectonic plates are on the move? We might consider some of the following …
Vacancy drift by task: Don’t watch job titles—watch the tasks inside them. If employers stop advertising for work like drafting, summarizing, or scheduling, it’s a sign the human demand is drying up.
Wage stagnation at the bottom: If entry-level wages stay flat even, automation—not shortage—is driving the hiring gap.
Career path decay: Fewer promotions from junior to mid-level roles is a leading indicator that the ladder is breaking, not just wobbling.
AI substitution inside roles: When 40%+ of someone’s job is quietly handled by automation, you’re no longer just augmenting—you’re shrinking the footprint of human work.
Graduate destination drift: When more new grads go freelance, defer work, or land in mismatched jobs, the entry-points to white-collar careers are dissolving.
Add to that the language shift you start hearing in cohort surveys … “There’s nowhere to grow,” “I just can’t get a start,” or “I trained to do X but now I am having to do Y”
We don’t need more dashboards. We need better seismographs (to maintain the metaphor)—tuned not just to the number of jobs lost, but to where and how the bottom rungs are vanishing. Because by the time those disappear completely, it’s not just junior roles we’ll have lost. It’s the entire system for how humans build expertise.
A couple of years ago we developed a ‘strawman’ structure to reflect how AI might shape the world of knowledge work in the future. Borrowing heavily from Aldous Huxley’s Brave New World, we created six social strata – including the newly advantaged, the newly disadvantaged and the dispossessed (i.e. move out of the knowledge work arena). You can find more about this in an earlier post.
The model we put forward is not static, and people can move between strata - moving up or dropping back down. The trigger for these movements up or down is developing the skills one needs to stay relevant and, indeed, thrive in the AI workplace. It’s about acquiring the set of skills to identify which we believe are most important to dial up because they are the ones that allow us to work most productively alongside generative AI.
The seven Power Skills we have often written about encapsulate this new skills agenda focused on enhancing individuals’ rich and nuanced appreciation of the big contextual picture, their ability to factor into their critical thinking a deep understanding of human complexities, their innate creativity; their unique ability to read the future; their quintessential human storytelling skills, and their ability to be inspiring leaders.
Ultimately, the choice in front of every leader, policymaker and professional is stark. If we keep clinging to the old “copilot” comfort blanket, the ground may well shift beneath us, in ways that we are not going to like. But if we lean into the evidence and invest in human skills that machines cannot yet codify, we can turn a potential moment of peril into one of renewal. The future of knowledge work is not predestined – we just need to meet the automation wave with foresight, candor and imagination.
For more information about this, please visit our new and improved website www.polymathmind.ai. You can also read about The Art of Collaborating With AI in our recently published book.
[i] Is AI Coming for Your Job? Here’s How to Tell. Investopedia (June 2025)
[ii] The White-Collar Recession Is Real. Josh Bersin (April 2025)
[iii] How AI is reshaping the career ladder. World Economic Forum (April 2025)