Most jobs are going away
In a Nutshell
Most jobs have derived demand—the human input is incidental to the desired outcome like health or housing—making them vulnerable to automation when machines deliver better, faster, cheaper results. Only about 5.5% of current work (at median, 15% at high-water mark) depends on essential demand where human presence, provenance, liability, or affinity is intrinsically valued, with the rest following power-law dynamics that concentrate rewards among a tiny fraction of providers. Even if wealth from automation flows to consumers, leisure time will largely shift to unpaid activities, and attention scarcity plus positional competition will prevent human-centric services from absorbing displaced workers.
These notes were generated by AI and may contain inaccuracies.
The moat for human demand for labor, human input, and human presence is thinner than commonly assumed. Labor's share of national income has been declining for decades. Around 2000 there was a brief bump, but the trend has been downward for nearly three decades, hitting a high watermark in the early 1960s or late 1950s. Labor share of national income represents the portion of income going to labor versus business and capital. It averaged 63-64% for a long time but has fallen below 53% with no sign of slowing. Even if the economy grows and new jobs are created, trillions of dollars flow to the ownership or capital class rather than workers. This metric is the most important indicator of a post-labor economy that has been developing for decades.
Alfred Marshall introduced the concept of derived demand in 1890. Demand for an input comes from demand for the outcome it helps produce. Human input is instrumental or incidental to the desired outcome. Most labor is derived demand, meaning the human input is not the point. When wanting a house, the concern is the house itself. Builders, contractors, electricians, and painters are incidental to the outcome. The vast majority of people do not know who built their house and do not care. Labor that builds houses is entirely derived with no intrinsic demand for human labor.
Health provides another example. The desired outcome is health, which includes tests, diagnosis, treatment, and daily assistance. Nurses and physicians are incidental but optional. Human care is a means to health, while reassurance can add distinct value. Doulas and hospice care represent a thin moat where people might want human presence for birthing or death, but most cases do not require human care and many people may prefer robots.
The substitution test states that when machines provide goods or services better, faster, cheaper, and safer, substitution occurs. Capital subsumes or displaces labor. When machines win on the outcome, derived demand declines. If robots can build houses in a couple of days at half the cost because labor represents 30-50% of house costs, people will choose that option. There is no care about human input except where legally required, such as having a general contractor liable for issues.
Essential demand is a neologism where human involvement is itself valued or required. Replacing the human changes what people are buying. The Taylor Swift effect demonstrates this, where Taylor Swift is intrinsic to the demand for concerts, albums, and merchandise. Removing Taylor Swift makes the product non-fungible. Essential demand is upstream of the meaning economy, relational economy, experience economy, attention economy, and statutory economy.
Essential demand has four constituent components: presence, provenance, liability, and affinity. Presence means paying for physical embodied presence of a human, such as childcare, massage therapy, or trip sitters. Telerepresence also counts, as with online therapy where the therapist's presence is still required despite physical distance.
Provenance means a human made it. This applies to YouTube videos, blogs, tweets, handmade goods, artisanal products, and research where human effort adds value. The researcher hired for the book on post-labor economics provides human provenance through editing, research, validation, and formatting, even though AI assisted with initial research and writing.
Liability involves accountability, ownership, and responsibility. A person answers for it through accountable human agency. This applies to judges, CEOs, doctors with medical licenses, lawyers, and elected officials where someone can be fired, sued, or jailed if mistakes occur.
Affinity means wanting a particular person, involving identity and attachment to a specific name and face. This includes Taylor Swift, Robert Downey Jr., and other celebrities, or wanting specific content creators. The same offering can contain multiple forms of essential demand, though usually one dimension is emphasized.
Presence involves wanting a pleasant human moment where the human is fungible or interchangeable. This includes massage therapists, musicians at local bars, baristas, and trip sitters. Some cases involve personal relationships with specific individuals, but there remains some level of interchangeability. Human premium still faces price comparison or compaction as more people compete for these roles when automation displaces workers from other jobs.
People intrinsically value goods and services with human involvement more highly. Despite products being made on assembly lines by lights-out factories, consumers seek handmade gifts on Etsy and treasure items made by human effort. AI art and music created with human input have different provenance levels than 100% artisanal work. The researcher hired to work by hand on sources and citations adds value precisely because the work is done manually by a trained librarian.
Human authority comes with human accountability. People want humans in positions where they can be fired or jailed if problems arise. This applies to presidents, mayors, judges, CEOs, lawyers, and doctors. Lawyers and doctors may persist due to licensing, insurance, and legal accountability requirements. However, doctors may eventually become net negative to medical care similar to how humans are now net negative in chess compared to computers operating alone. The centaur method of chess shows that human-AI hybrid teams are inferior to AI alone.
Some medical processes already involve approval theater where doctors rubber-stamp prescriptions without direct patient interaction. Regulatory requirements may maintain human authority positions even when AI could perform better. Supporting work can be automated while human authority persists for accountability purposes, such as with presidents and Supreme Court justices who can be impeached or imprisoned.
Affinity represents the nominal dimension where specific individuals are wanted. Disney brought back original Avengers cast members because the MCU was declining. Wanting Taylor Swift specifically because she is Taylor Swift exemplifies affinity. Content creators with specific followings represent affinity where the individual is not interchangeable.
Jobs are bundles of tasks where some tasks can be automated while others remain. One essential task does not preserve every other task. Human tasks can remain while the rest automates, resulting in compression where instrumental or incidental tasks are offloaded to machines. Roles like CTOs may persist due to liability for signing documents, affinity from being hired for specific outcomes, and presence in physical meetings, though provenance is less important since AI assistance is standard.
Analysis of all 831 BLS job categories covering 99.9% of American jobs found that 0.3% have humans as central, 3.8% are relationship-intensive, 16% are mixed offerings, nearly 50% are niche preference, and over 31% have no identifiable channel for essential demand. In aggregate, about 5.5% of work done today is essential demand-based labor at the median, with a high water mark of 15%. This suggests potential for 85% unemployment if current automation levels were achieved, though massive unemployment would trigger reallocation of money, time, and energy.
As economies grow wealthier, spending shifts toward leisure, discretionary spending, luxury experiences, and attention economy goods. Basic necessities like food, housing, clothing, and healthcare consume most income in poor societies, but wealthier societies spend more on non-necessities. Luxury spending tends to involve more human provenance or presence, such as coffee shops and restaurants. However, households may not access wealth created by automation, which flows primarily to capital owners rather than workers.
Even if everyone were effectively a billionaire, there is a limit to how many massages need to go around daily. Even if everyone received a massage every day, massage therapists would comprise around 30% of the economy. The shift toward more leisure and valued human experiences is an established trend, but the question remains how many jobs this will create.
Alex takes the view that mimetic and positional goods will drive growth in the experience economy. Luxury spending includes fancy houses and fancy cars, neither of which require human input. Even though the luxury economy might grow, a huge chunk goes toward wanting a fancier car, nicer house, or new Xbox. These are mimetic desires wanted because other people have them.
Mimetic desire creates an economic ratchet effect where every time someone else gets more, others want more, creating a ladder effect of people wanting more and more. Human appetites are functionally infinite over time as technology provides more goods and services at lower prices.
Positional goods are valued for relative status or scarce access, such as Malibu homes where there is limited real estate. Neither status nor imitation requires a human maker. This represents a significant oversight in the argument that luxury spending growth will necessarily translate to more jobs supporting that spending.
Some spending will go toward massage therapy and psychedelic retreats, but this represents only a small slice. Even at the high end, this amounts to around 15% of the economy, which is not enough to significantly grow jobs already in demand.
In a future where everyone is functionally a millionaire or billionaire due to AI, robotics, solar, and fusion, the question becomes whether essential human demand for goods and services will expand. In Star Trek's utopian vision, people wear whatever they want, eat whatever they want, go wherever they want, and never need to work.
Current data shows average daily leisure hours per person aged 15 plus in America is about 5 hours. Eliminating all work would provide about 8.5 hours. Eliminating all work and household chores would provide about 10.5 hours, roughly doubling available free time.
Even doubling the high number for leisure demand would mean 30% of today's jobs at most, but it's actually going to be closer to maybe 11%. The real bottleneck is that a wealth of information creates a poverty of attention, a concept from Herbert Simon in 1971. Even if leisure demand doubles, human attention becomes the most scarce resource.
The attention economy, experience economy, and meaning economy all follow a power law. Therapy is intrinsically one-to-one, representing the high water mark where one pays for an hour of someone's time one-on-one. A yoga class might have 20 students for one teacher, creating a 20:1 ratio. Live music at a local bar might have four musicians serving 70 people for two hours.
Local stars can still take most of the audience. Global reach means global rivals, creating intense competition in the attention economy. The creator economy is growing, but revenues are going down because the number of people joining is increasing faster than the money being put into it.
Physical massage is one of the most durable modes because it requires being physically present with human hands and is intrinsically local and one-on-one. However, as more people are fired from office and white collar jobs, many will try to enter fields like massage, increasing competition for clients.
Over the last few years, creators have grown 183% but revenue has only grown 60%. The median creator takes home about $300 a month, including on OnlyFans. The power law means that scalable affinity rewards the few, with concentration in creator payments. The top 5% are the only ones who can really make a living in the creator economy because of how tough the competition is, and it's getting worse.
Leisure can be unpaid through friends, family, hobbies, and nature. If everyone goes from 5 hours of leisure per day to 10 hours, that time will not necessarily go toward paid engagements. People might spend more time on hiking trails, biking, mountain biking, with friends, or partners. Ethical non-monogamy is huge, and activities with children, partners, friends, and hobbies are often unpaid financial transactions.
Even though essential demand includes wanting to spend more time with people cared about, this does not necessarily translate to more demand in the marketplace. In an automated future, the demand for human presence is going to fall off significantly.
Keep David Shapiro in your library
Save the videos and channels worth coming back to, and find them again in one place.
