Robotics & Automation
The Last Job Isn't the One You Think
The jobs that automation is coming for next aren't the repetitive, predictable ones — those are mostly gone — it's the ones that require a body in an unpredictable world.
The Idea
For decades, the standard story about automation was tidy: machines take the routine, humans keep the creative. Factory floors got robots; knowledge workers stayed safe. That framework is now badly out of date — and the revision is stranger than most people expect. The hardest thing to automate isn't thinking. It's moving through a messy physical world. This is sometimes called Moravec's paradox, after roboticist Hans Moravec, who noticed in the 1980s that the things hardest for humans — chess, calculus, logical inference — are easy for computers, while the things effortless for a toddler — picking up a dropped spoon, navigating a crowded room — remain fiendishly difficult for machines. A warehouse worker, a plumber, a care assistant: these roles demand constant, fluid adaptation to environments that change by the second. That's genuinely hard to engineer. But the paradox is eroding. A new generation of general-purpose humanoid robots — trained not on rules but on vast datasets of human motion, reinforced through trial and error in simulation — is learning to handle physical unpredictability in ways earlier robots never could. The question is no longer whether physical labour will be automated, but in what order, at what cost, and with what consequences for the roughly two billion people worldwide whose livelihoods depend primarily on their bodies rather than their credentials.
In the World
In 2023, Figure AI — a startup less than a year old — released footage of its humanoid robot completing a full shift of tasks at a BMW manufacturing plant in Spartanburg, South Carolina. The robot wasn't bolted to a spot on the assembly line performing one motion on a loop. It was navigating the floor, picking up sheet metal parts, inspecting them, and placing them into a fixture — adapting, in real time, to slight variations in position and weight. BMW had chosen Spartanburg partly because the plant already struggled to fill physically demanding roles; the robot wasn't replacing someone who wanted the job. That detail matters. The standard anxiety about automation tends to assume a zero-sum displacement: robot in, worker out. But the actual picture emerging from early deployments is more tangled. Some facilities adopt robots because they genuinely can't recruit for certain roles. Others use them to extend what human workers can do, pairing a person's judgment with a machine's endurance. And some — particularly in logistics — are replacing workers in conditions so gruelling that turnover is already near-total. Figure's robot learned its tasks not through explicit programming but through a technique called imitation learning: it watched humans perform the motions, then refined its behaviour through millions of simulated repetitions before touching a real part. The line between trained and programmed has blurred — and that blurring is exactly what makes this wave of automation categorically different from the last one.
Why It Matters
The political and economic conversation about automation tends to run about a decade behind the technology. We're still debating the disruptions of the last wave while the next one is already moving into production environments. What's worth carrying from this isn't panic — the history of labour is a history of adaptation — but a more precise map of what's actually at stake. The people most exposed to this transition aren't necessarily low-skilled; they're people whose skills happen to be physical and situational. A skilled tradesperson, a surgical nurse, a logistics coordinator who knows the warehouse floor by feel: these roles will face pressure in ways their holders aren't being told to expect. At the same time, the technology is creating genuine demand — for the people who train, maintain, and supervise these systems. That's real, but it requires acknowledging that the worker being displaced and the worker being hired are rarely the same person, or in the same place. Knowing where the pressure is actually building — not where the headlines say it is — is what lets you think clearly about your own work, your industry, and what kinds of skills will remain stubbornly human for the next decade.
A Question to Ponder
Which parts of your own work, or the work of people you rely on, depend on a body moving through an unpredictable physical world — and have you ever thought of that as a form of expertise worth naming?
Get a new one of these every morning.
Start learning with Thinkable