AI Doesn't Only Threaten White-Collar Jobs

A recurring debate surrounds artificial intelligence: AI is mostly a threat to intellectual jobs, to white-collar workers. Manual jobs, the argument goes, are safe. This view is appealing. It is also deeply inaccurate, and dangerous, because it reassures exactly the wrong people.

A flawed historical analogy

The reasoning behind this idea draws on past industrial revolutions: machines replaced muscles, not brains. Steam, electricity, and industrial robotics mechanized physical labor, while cognitive work remained out of reach.

But this heuristic is outdated. Artificial intelligence is not an industrial machine. It is a system that processes information and detects patterns, which is precisely the raw material of intellectual work. Comparing AI to a CNC lathe is conflating two entirely different technological orders of magnitude.

The real question is not "is it physical or intellectual?" It is "does a reproducible methodology exist?"

The real criterion: structured or ambiguous?

A job is automatable as soon as it relies on a methodology, a set of rules, implicit or explicit, that allow the expected output to be produced. It does not matter whether the work is physical or intellectual. What matters is the presence of a reproducible process.

Under this criterion, the real barriers to automation are not tied to whether a task is manual or cognitive, but to three distinct categories:

The three real barriers to automation:

• Physical environmental constraints (water, electricity, unstructured spaces): robotics is not yet robust or cheap enough.
• The absence of an explicit protocol: jobs where expertise is tacit, contextual, and relational.
• Breakthrough creativity and emotional intelligence: not execution-level creativity (AI already does that), but the kind that mobilizes genuine feeling, empathy, and human intuition.

White-collar workers are on the front line, not in safety

Paradoxically, many intellectual jobs are more vulnerable in the short term than some manual ones. The reason is simple: their work is already digital. No robotics are needed to automate it. A junior lawyer processes documents. A financial analyst looks for patterns in data. A radiologist detects anomalies in images. All of these processes are, at their core, operations on structured information.

Deployment costs are near zero. A well-trained model can be put into production in a matter of weeks. No robotic arm required.

The bricklayer still has a head start on the lawyer.

Creativity and emotion: the last human stronghold, for now

AI already generates images, composes music, and writes text. It simulates creativity with unsettling efficiency. But there is a fundamental difference between producing something that looks like creativity and creating from lived experience, from genuine pain or joy.

The therapist who senses that a patient is holding something back. The musician who reads the energy of a room. The designer who intuitively grasps what a client cannot put into words. The negotiator who reads silences as much as speech. These skills rest on a form of emotional and sensory intelligence that AI does not possess, not yet.

The nuance is critical: impossible right now does not mean impossible forever. This is not permanent protection, it is a head start. Jobs built on emotion and feeling are the last ones in line. But they are still in line.

AI + robotics convergence: the last physical lock

For manual jobs that seem safe today, the barrier is not conceptual, it is economic and technological. Industrial robots have existed for decades. What is still missing is the combination: a robotic system agile enough, cheap enough, and guided by an AI capable of handling the physical unpredictability of an unstructured environment.

That lock is breaking. Advances in humanoid robotics, combined with the planning capabilities of large language models, are paving the way for automation of structured manual work, supply chains, warehouses, partially controlled job sites, within a ten-to-twenty year horizon.

Comparative table: who is exposed, and when?

Conclusion: take off the blinders

Thinking AI is a problem for "knowledge workers" and not for "manual workers" is reading from the wrong map. The real dividing line lies elsewhere: between what is reproducible and what is not, yet.

Creative and emotional tasks are today's last solid human stronghold. Not because they are mysterious or magical, but because they are grounded in lived experience, sensory awareness, and a kind of presence that AI does not have. For now.

The right question to ask about any job remains simple: does my work follow a reproducible methodology? If yes, automation is only a matter of time and cost. If not, if your value lies in ambiguity, emotion, or disruption, you have a lead. But it is only a lead.

---

Related reading: on how human contact is becoming a luxury good in this same economy, see The Great Reversal.