Every major technology wave produces a version of this fear. The spinning jenny was going to eliminate weavers. Electricity was going to end factory work. Computers were going to make office workers redundant. Each wave disrupted specific roles — sometimes violently — and created many more that no one had anticipated. AI is genuinely different in the breadth and speed of its reach. But the economic dynamics that governed every previous technological transition are likely to govern this one too. The frameworks that help explain why aren't reassuring platitudes. They're empirically grounded patterns with real predictive track records. Here are four.
The Lump of Labour Fallacy
The assumption underlying most job-loss anxiety is that there is a fixed quantity of work in the economy. If machines do some of it, there's less left for humans. Economists call this the Lump of Labour Fallacy — and its empirical track record is consistently wrong.
When ATMs were introduced across American banking in the 1970s, economists predicted a corresponding fall in bank teller employment. Teller employment grew. ATMs lowered the cost of running a bank branch, so banks opened more of them. The work tellers did shifted from cash handling — which the machine could do — to relationship management, problem resolution, and product guidance, which it couldn't. The total quantity of teller work expanded even as its nature changed.
The same pattern appears across the historical record. Agricultural mechanisation did not end agricultural employment for decades; it redirected it toward higher-skill tasks. Industrial automation shifted workers from production lines to quality control, logistics, and process management. In every case, rising productivity raised purchasing power, which raised demand for new goods and services, which created new categories of work that hadn't previously existed. The quantity of economic work is not fixed. It expands.
Jevons Paradox
In 1865, the economist William Stanley Jevons made an observation about coal that seemed paradoxical: as steam engines became more efficient — using less coal per unit of work — total British coal consumption went up, not down. Lower cost per unit meant more people could afford steam power. More engines ran. More coal was consumed. Efficiency had increased total demand, not reduced it.
The pattern recurs reliably. When personal computers made document creation cheaper, the total volume of documents created didn't plateau — it exploded, and so did demand for people who could write well. When digital photography eliminated film costs, the volume of photographs taken increased by orders of magnitude — and so did the commercial market for photographers who could do what the camera couldn't: direct, compose, and interpret. When search engines made information retrieval nearly free, demand for people who could synthesise and apply information grew, not shrank.
AI making knowledge work cheaper is likely to expand the total market for knowledge work — because it will lower the cost of things that were previously too expensive for many people to commission, and because it will create new possibilities that weren't previously feasible. The elasticity of demand for things people find genuinely valuable tends to outrun efficiency gains.
Comparative Advantage
David Ricardo's theory of comparative advantage — developed in the early nineteenth century to explain the logic of international trade — has an underappreciated application to human-machine collaboration. The core insight: even when one party can do everything better than another, both are still better off specialising in what each does relatively best and trading the surplus.
Applied here: even if AI performs most knowledge tasks faster and more reliably than any individual human, it doesn't follow that humans become economically redundant. What follows is that humans should concentrate on the tasks where their relative advantage is largest — and AI should handle the rest. This is not a comfortable frame for the near term, because it implies genuine transition and real disruption for people whose current roles map most closely onto what AI can now do. But at the level of the broader economy, it suggests that AI and human workers function as complements rather than substitutes — each making the other more productive by concentrating on what each does best.
The history of labour-saving technology consistently supports this reading. Roles don't disappear so much as they decompose: the parts machines can replicate are automated; the parts that require human judgment, relationship, and accountability become more valuable, not less.
Human Privilege
There's a fourth dynamic that doesn't have a canonical economic name but that is structural rather than sentimental. Call it human privilege: the category of things humans do in economic life that AI cannot substitute for — not because of current capability limits that will eventually close, but because of the nature of what those things are.
Accountability is one. When a surgeon operates, a lawyer advises, a financial adviser recommends, or an executive decides, part of the value of that action is constituted by the fact that a licensed, accountable human being stands behind it. Trust in high-stakes domains runs through people, not systems. AI can assist in all of these roles — and already does — but the accountability structure that makes the action meaningful requires a human to own it.
Embodied presence is another. For many of the roles that matter most — teaching, care, therapy, leadership, negotiation, community — physical human presence isn't one delivery mechanism among several. It is what the service fundamentally is. You cannot automate the presence of a person in a room, a conversation, a relationship.
And perhaps most structurally: humans are uniquely positioned to understand the experience of other humans. Empathy, cultural context, lived experience, moral authority — these are not soft competencies that AI will eventually replicate as models improve. They are the basis of an entire category of economic value that compounds over time precisely because it is irreducibly human. The roles most shaped by these qualities are likely to be the most resilient in the labour markets of the next decade, because they are the hardest to substitute and the most deeply valued.
The question isn't whether work will change.
It's whether you're positioned for what comes next.
None of this means the transition will be painless. It won't be. Specific roles will shrink or disappear, and the people in them will need to move. The Lump of Labour Fallacy tells us the total amount of work won't collapse — but it doesn't tell any individual that their current role is safe.
Waymark exists precisely for this moment. Not to reassure you that everything will be fine, but to help you map what you've actually built, understand where it's genuinely valued in a changing market, and build a clear path to what comes next. Most people carry more transferable value than they've been told. Finding it — and naming it — is the first step toward something better.