The great audit of useless work

Posted on: 10 December 2025

The numbers appear to tell two opposite stories.

On one side, collapse. American managers now supervise three times as many employees as they did in 2015. Google has eliminated 35% of its small team leaders, Intel has halved its management positions. Gartner estimates that by 2026, one in five companies will use artificial intelligence to cut half of their hierarchical levels. They call it the "Great Flattening". The papers speak of it as an employment catastrophe.

On the other side, explosion. Hiring in skilled trades has increased by 376% in the last quarter compared to the previous year. Four hundred thousand welders are missing in the United States alone. Electricians, HVAC technicians, solar panel installers, industrial maintenance workers: demand is through the roof, supply dramatically insufficient. Young people, women and veterans are turning to vocational schools at levels not seen since the Second World War.

Two separate crises requiring different solutions?

No. It's the same story seen from different angles. And understanding this completely changes how we read what's happening.

The comfortable narrative and the uncomfortable one

The dominant narrative about artificial intelligence and work follows a predictable pattern: the machine arrives, the human is replaced, technological progress produces unemployment. It's a story we've been telling for two hundred years, since the mechanical looms. Every generation fears that this time it's different, that automation will finally render us all superfluous.

But the 2025 data tells something more nuanced and, in certain respects, more interesting.

Peter Cappelli, director of the Center for Human Resources at the Wharton School, has introduced a term worth noting: "AI-washing". Many companies, he argues, are using artificial intelligence as a socially acceptable justification for cuts they would have made anyway. It's easier to say "technology forced us" than to admit strategic errors, excessive hiring during the post-pandemic boom, or simple investor pressure for better margins.

Cappelli also notes a herd effect: when companies see competitors cutting, they cut in turn. "If it looks like everyone is reducing, you think they must know something you don't." And financial markets reward this dynamic: layoff announcements often push share prices up because they signal "discipline" and "efficiency".

There's truth in this reading. But stopping here means missing the deeper structural mechanism.

The lights come on in the room

For decades we've built organisations where a significant percentage of roles existed primarily to manage communicative inefficiency between departments. Coordinators coordinating other coordinators. Managers whose main output consisted of requesting updates and redistributing them upward. Analysts producing presentations that other analysts would synthesise into other presentations.

The anthropologist David Graeber, in his 2018 book, called them "bullshit jobs". Not out of cynicism or contempt, but from clinical observation. Graeber noted that many people in these roles knew perfectly well that their work produced no real value. Some suffered deeply from it. But the system kept creating them because they served to justify hierarchies, to signal importance, to fill organisational charts that needed to appear substantial.

Artificial intelligence hasn't eliminated these roles. It's made them visible.

When an automated system can do the work of "passing information from A to B", when it can generate the weekly report, synthesise the data, coordinate calendars and respond to routine emails, it suddenly becomes clear that this work wasn't creating value. It was merely occupying space in the organisational structure.

It's as if someone switched on the lights in a room where everyone was pretending to be busy.

The distinction that matters

The phenomenon isn't uniform. There's a crucial difference between repetitive conceptual work and creative conceptual work.

Writing the hundredth quarterly report following the same template is repetitive conceptual work. Designing a strategy to enter a new market requires judgement, synthesis of incomplete information, understanding of human dynamics: that's creative conceptual work. Artificial intelligence excels at the former, struggles with the latter.

The same distinction applies to manual work. Tightening bolts on an assembly line is repetitive manual work, automatable for decades. Diagnosing why an air conditioning system isn't working properly in a 1970s building with pipes modified three times requires accumulated experience, situational reasoning, ability to adapt to unforeseen variables. That's skilled manual work, and it remains stubbornly difficult to automate.

What emerges from the data is a reconfiguration of value along these lines. Work that can be reduced to standardised procedures, whether conceptual or manual, loses economic value. Work that requires physical presence, contextual judgement, adaptation to unique situations becomes relatively more valuable.

And herein lies the historical irony.

A generation's bad advice

For thirty years we told young people that the future belonged to those who "work with their heads, not their hands". That manual trades were for those without better alternatives. That real status, real success, lay in office work, impressive titles, careers that took you ever further from the materiality of things.

Entire generations followed this advice. Universities churned out millions of graduates in disciplines designed to fuel the knowledge economy. Vocational schools were left to languish, perceived as a second-rate option. The result is a generation of workers overqualified for roles that are contracting and underqualified for roles that are expanding.

Meanwhile, those who ignored conventional wisdom, who learned to weld, to install electrical systems, to repair industrial machinery, now find themselves in a position of bargaining power not seen for decades. Not because the market has become irrational, but because real scarcity produces real value.

The numbers are eloquent. A certified welder in the United States can today earn more than many graduates with MBAs. And unlike the middle manager, they needn't worry that an algorithm will learn to do their job in the next update.

The geopolitical convergence

There's another element that makes this moment particular: convergence with apparently separate geopolitical dynamics.

Economic protectionism is returning. Tariffs, import restrictions, incentives for production reshoring, more restrictive immigration policies. Whether it's Trump's America, Europe seeking strategic autonomy, or China pursuing technological self-sufficiency, the direction is the same: less dependence on foreign sources, more domestic production.

This means factories reopening or being built from scratch. Infrastructure to modernise. Supply chains to rebuild locally. And all of this requires people who actually know how to build, install, make physical things work.

The two forces, automation of repetitive conceptual work and return to local production, aren't in contradiction. They converge toward the same result: a profound reconfiguration of which human work has economic value.

This isn't a nostalgic return to the industrial past. Modern factories are high-tech environments where a single technician supervises automated systems producing as much as hundreds of workers did fifty years ago. But that technician must exist, must be qualified, must be physically present. And there are far too few of them.

What this means for those working today

What does all this entail for those building their careers or rethinking existing ones?

The first observation is that "learn to code", the standard advice of the last fifteen years, is rapidly becoming insufficient. Routine programming is exactly the type of repetitive conceptual work that artificial intelligence does well. The programmers who thrive are those doing system architecture, who understand business problems, who know how to translate human needs into technical specifications. Creative conceptual work, not code execution.

The second observation is more counterintuitive: skills requiring physical presence, manipulation of real objects, adaptation to concrete situations are acquiring an economic premium they haven't had for generations. Not because the market has gone mad, but because they've become relatively scarce while demand explodes.

The third observation concerns the nature of "coordination". Middle management is evaporating not because coordinating people is useless, but because the type of coordination consisting of passing information between hierarchical levels has been made obsolete by technology. The coordination that survives and thrives is that requiring human judgement: managing conflicts, motivating people, making decisions under ambiguity, building relationships of trust. Fewer managers, but those who remain do qualitatively different work.

The limits of this reading

It would be easy to transform this analysis into a triumphant prediction about the "return of real work". But intellectual honesty requires signalling the limits too.

The manufacturing that's returning employs a fraction of the workforce it once did. The semiconductor factories opening in Arizona create thousands of jobs, not hundreds of thousands. Automation advances in traditional trades too: robots already lay bricks and drones inspect infrastructure.

Moreover, the transition isn't painless. Someone who spent twenty years building a career in corporate coordination cannot simply reinvent themselves as an electrician. Skills don't transfer that easily. There's a generation of professionals who risk finding themselves caught in this reconfiguration with qualifications losing value and few options to acquire new ones.

Finally, much depends on political decisions that could go in different directions. Protectionism might reverse. Artificial intelligence might make unexpected progress in physical manipulation. Demographic dynamics might change the work equation in ways we cannot foresee today.

The question that remains

But if current data is any guide, and intellectual honesty requires admitting it might be, what we're witnessing isn't the end of human work. It's the end of a certain type of work that, retrospectively, produced less value than it claimed.

The great flattening isn't a catastrophe. It's an audit. And like all audits, it reveals truths that were always there, merely hidden beneath layers of organisational complexity and social convention.

Work that creates real value, whether creative conceptual or skilled manual, isn't disappearing. It's becoming more visible, more recognised, better compensated.

Work that pretended to create value, coordination for its own sake, performative bureaucracy, organisational theatre, is losing the cover that protected it.

For some this is an existential threat. For others, a historic opportunity.

The difference probably lies in the answer to a simple but pitiless question: does what you do every day create something that didn't exist before? Or does it keep alive structures that exist primarily to justify their own existence?

Artificial intelligence doesn't answer this question for you. But it's making it ever harder to avoid.