Key Takeaways
The threat is not zero work, but too little work for everyone.
A gradual squeeze. In the 1890s, cities worried that horse manure and carcasses would overwhelm their streets. Cars eventually displaced horses, and the crisis disappeared. The story is often used to reassure people that technology creates solutions. Economist Wassily Leontief saw a darker lesson: machines had pushed an essential worker out of the economy. Susskind’s argument is not that every job will suddenly vanish. It is that technology will steadily shrink the range of tasks people perform, until remaining work cannot employ everyone who wants it.
Look beyond job counts. Farming and manufacturing still produce substantial output, but each needs far fewer workers than before. The danger is not simply unemployment. Pay, security, status, and the quality of available work can deteriorate even while employment figures look healthy.
The horse story is useful precisely because it is not a clean forecast. It reveals how a technology can solve one problem while creating a less visible transition for those displaced. The distinction between no work and not enough decent work also improves public debate: headline employment rates can conceal falling participation, insecure jobs, or declining wages. The analogy has limits, however. Humans can vote, learn, and organize in ways horses could not, so the outcome is shaped by institutions as well as invention.
Track substitution and new demand separately before judging automation's net effect.
Two forces at once. The substituting force removes people from tasks machines can perform. The complementing force raises demand for people elsewhere. Historically, complementarity worked through three channels: technology made workers more productive, economic growth expanded demand, and new industries created different kinds of work.
The ATM puzzle. As ATMs spread across the United States, bank branches needed fewer tellers each. Yet teller employment rose about 20 percent between the late 1980s and 2010. ATMs lowered branch costs and freed tellers to help customers with more complex needs, while banks opened more branches to serve growing demand. This example shows why counting displaced tasks alone can mislead. It does not prove that new demand always wins; the balance depends on what people still do better than machines.
The two-force model replaces a simple jobs-destroyed-versus-jobs-created argument with a causal account. It asks where productivity gains go, how prices and demand respond, and whether new tasks still require workers. That is a stronger question for policymakers than asking whether a technology is good or bad. The ATM example also warns against treating historical averages as guarantees: a technology that creates complementary work in banking does not establish that future tools will do so in every industry.
Judge automation task by task, because job titles conceal mixed work.
Jobs are bundles. The Autor-Levy-Murnane hypothesis, often shortened to the ALM hypothesis, explains how technology can hollow out middle-paying occupations. It focuses on tasks, not job titles. Earlier computer systems excelled at routine tasks, meaning activities people could describe as explicit steps. Tasks relying on tacit knowledge, which people use without easily explaining it, seemed harder to automate. Many clerical and production jobs contained routine tasks, while some high-paid professional and low-paid service roles relied more on judgment, interaction, or manual dexterity.
Partial automation matters. A 2017 McKinsey study found fewer than 5 percent of occupations could be fully automated with existing technology, but more than 60 percent contained tasks that were at least 30 percent automatable. A profession can survive while much of its work changes.
Task-level analysis helps explain why an occupation can remain on paper while its entry-level work, staffing, and bargaining power change substantially. It also prevents misleading claims that a whole profession is either safe or doomed. The routine-task account is valuable but not a permanent map: machine learning can handle activities once classified as tacit or non-routine. The more durable lesson is methodological. Forecasts should describe which activities change, for whom, and under what conditions, rather than attach a single automation probability to an entire occupation.
Machines need not think like people to outperform people.
A different route to capability. Early artificial intelligence researchers often tried to imitate human reasoning. The pragmatist revolution names the shift toward building machines that perform a task effectively by any method, whether or not that method resembles ours. Deep Blue defeated chess champion Garry Kasparov by searching vast numbers of possible moves, not by copying his intuition. AlphaGo learned from expert games and repeated play; AlphaGo Zero later learned using only the game’s rules and self-play.
Beware the AI fallacy. This is the mistaken belief that a machine can perform a human task only by reproducing human thought. A system can identify a cancerous lesion or generate a design without human-like judgment or creativity. Its inner process may be alien, yet its practical ability still matters.
This distinction separates two questions that are often blurred: whether a machine has human-like understanding, and whether it can perform a valuable function. The first matters for philosophy and consciousness; the second matters for labor markets and accountability. A human comparison can still be useful when the task involves trust, explanation, or responsibility. But demanding human-like cognition as proof of capability sets an unreliable test. Institutions need standards for outcomes and oversight, not just comforting stories about how a system works.
Map machine capability separately from whether employers will adopt it.
Encroachment across abilities. Task encroachment is the continuing expansion of machines into work once done by people. Susskind groups human capabilities into manual, cognitive, and affective domains. Robots now harvest crops and lay bricks; software reviews legal contracts and supports medical diagnosis; systems also analyze faces, voices, and emotional signals. The point is not that every task is already automated, but that yesterday’s limits can become tomorrow’s ordinary tools.
Capability is not adoption. Whether a machine is used depends on its cost compared with wages, the tasks common in a region, regulations, and public acceptance. Cheap labor can make handcrafting more economical than machinery. Japan’s aging population and care-worker shortages have encouraged investment in nursing robots, while other societies have been more hesitant.
Separating technical possibility from real-world adoption avoids two opposite errors: treating every demonstration as an imminent labor-market revolution, and assuming that today’s barriers are permanent. Adoption is a social and economic choice, not a direct readout of engineering progress. The distinction also raises a distributional question: who captures the savings when an employer automates, and who bears the transition costs? A society can welcome safer or more accessible services while still requiring employers to account for affected workers.
Prepare for three kinds of displacement: skills, identity, and geography.
Work can be out of reach. Frictional technological unemployment describes a situation where jobs still exist, but some displaced people cannot access them. Susskind identifies three obstacles: skills mismatch, when workers lack required training; identity mismatch, when available work conflicts with how people understand their role or status; and place mismatch, when jobs are far away and moving is difficult. Former factory workers, for example, may struggle to retrain, reject lower-status service jobs, or lack the means to relocate.
Count the hidden exits. The unemployment rate excludes people who stop looking for work. Participation rates, wages, security, and job quality therefore matter too. When many workers compete for a narrow pool of accessible jobs, pay and conditions can fall, even if the official number of jobs does not.
The three-mismatch framework makes labor-market transitions more human and more concrete. Training alone cannot solve a problem rooted in place, social identity, or the availability of stable work. It also suggests that policy should be evaluated by whether people can realistically use it, not merely whether a program exists. Identity deserves particular care: rejecting a job can reflect social expectations, but it can also be a rational response to low pay, poor conditions, or the loss of a valued occupation.
Do not assume a bigger economy automatically creates human jobs.
Growth can bypass workers. Structural technological unemployment means there are too few jobs for people overall, not merely a mismatch between workers and vacancies. Historically, economic growth complemented labor because people were usually best placed to perform new or expanding tasks. Susskind calls this the superiority assumption. As machines take on more tasks, that assumption weakens: demand for more goods does not necessarily mean demand for more human labor.
Output and employment can separate. British agriculture produced over five times as much as in 1861 while its share of the workforce fell from 26.9 percent to 1.2 percent. In a study of US labor markets from 1990 to 2007, one additional industrial robot per thousand workers was associated with about 5.6 fewer jobs and wages 0.5 percent lower across the economy. Growth alone does not guarantee broadly shared gains.
The argument challenges a powerful intuition in economics: more production should create more work because people spend their new income. That feedback can be real, but it depends on who performs the extra tasks. The central uncertainty is not whether technology raises output, but whether labor remains a necessary input as output expands. The robot findings are evidence about a specific technology and period, not a universal law. Their importance is that they show economy-wide losses are possible even after complementary effects are counted.
Prepare to distribute prosperity when wages no longer reach everyone.
Income follows ownership. Human capital means the skills and abilities people use to earn wages; traditional capital includes assets such as property, machinery, software, and shares. If technology reduces demand for labor while increasing returns to valuable assets, people without either kind of marketable capital may receive little income. Existing inequality offers an early warning: wealth and capital income are concentrated, and productivity gains have not consistently kept pace with workers’ pay.
Build institutions for sharing. The Big State is Susskind’s proposal for shifting the state’s role from producing prosperity to distributing it. Its possible functions include taxing high incomes, capital, and large-company profits; sharing capital through public wealth funds; and supporting workers’ pay and conditions during the transition. The appropriate combination depends on political choices, but relying on wages alone becomes less viable as labor demand shrinks.
This reframes automation as an institutional question rather than a technical puzzle. If productivity gains accrue to owners, then ownership and tax rules determine whether abundance becomes shared security or sharper stratification. Public wealth funds offer a different lever from cash transfers because they broaden claims on future returns, though they demand transparent governance and public trust. The proposal is not politically neutral: deciding which assets to tax and who qualifies for support expresses judgments about fairness, responsibility, and membership.
Teach for tasks machines cannot yet do, and keep learning throughout life.
Education must adapt. In the near term, Susskind recommends teaching skills that machines do not already perform well, while improving access to technical education. This means changing what is taught, how it is taught, and when. Personalized learning systems could tailor instruction, while online courses can reach far more students than a physical classroom. Education should also continue through adulthood as workers move between roles.
But training is not a universal fix. People differ in abilities, time, and resources, and some skills may be difficult or impossible for them to acquire. Even successful retraining cannot solve structural unemployment if the economy lacks enough demand for human labor. The point is to use education as a practical response to present transitions, not to promise that more schooling can indefinitely keep everyone employable.
The strongest version of lifelong learning is not a demand that individuals absorb every risk created by technological change. It is a public commitment to accessible, useful education, paired with realistic expectations about what training can accomplish. Personalized instruction may improve learning, but digital delivery by itself does not ensure motivation, completion, or opportunity. The deeper policy test is whether education expands people’s choices, rather than simply sorting them more efficiently into a shrinking set of jobs.
Regulate Big Tech as political actors, not only as monopolies.
Power exceeds prices. Competition policy asks whether dominant firms harm consumers through prices, quality, or reduced innovation. Susskind argues that technology companies also exercise political power: platforms shape what people see, how information spreads, and which decisions are made about them. A search ranking, recommendation system, or data-driven eligibility decision can affect public debate and individual opportunity even when the service is free.
Create public oversight. The proposed Political Power Oversight Authority would investigate technology systems, require transparency, and limit or break up firms when their influence over shared civic life becomes excessive. This is not a claim that technology companies should have no influence. It is a call to ensure that decisions affecting liberty, democracy, and justice are not left solely to private firms or their self-regulation.
The distinction between economic and political power is important because consumer choice does not automatically confer democratic legitimacy. A person may value a platform while remaining unable to inspect how its algorithms shape exposure, or how its data affects decisions. Oversight would need technical expertise as well as legal and ethical judgment. It also carries a risk: regulators can misuse power or freeze beneficial innovation. The challenge is not to eliminate influence, but to make its exercise visible, contestable, and accountable.
Design meaningful roles beyond paid work, including care and civic contribution.
Work is not everyone’s purpose. In the 1930s, researchers studying Marienthal, an Austrian village after its factory closed, observed apathy, lost routines, and withdrawal from social life among unemployed residents. Work can provide structure, identity, and social esteem, not just income. Yet this link is neither universal nor timeless: many people find work empty, and earlier societies often treated it as drudgery rather than fulfillment.
Make contribution visible. Susskind proposes leisure policies and a conditional basic income, or CBI: public support tied to a contribution to the community when paid employment is unavailable. Possible contributions include care, education, cultural activity, and civic participation. This approach could recognize valuable work that markets often underpay or ignore, including household and caregiving labor. Communities would need to decide what counts as contribution and how to avoid turning support into coercion.
This is the book’s most speculative proposal, and its most revealing one. If paid employment stops serving as the default proof of usefulness, societies must recognize contribution in other forms, particularly care that has long been undervalued. Conditions attached to income may strengthen reciprocity, but they can also become intrusive tests of deservingness. A credible system would need broad democratic input, room for disability and personal circumstance, and respect for forms of meaning that cannot be measured by a state checklist.
Analysis
Susskind’s central contribution is a change in the unit of analysis. Public debate asks whether jobs disappear; the book asks how machines alter tasks, access to work, and the allocation of income. That shift makes the argument stronger than a dramatic forecast about a fully automated future. The claim is not that a date can be named for mass unemployment, but that the familiar historical mechanisms that absorbed displaced workers may weaken when machines become capable of performing new tasks too.
The argument is especially persuasive when it distinguishes technical possibility from labor-market outcomes. A technology can be capable of replacing workers without being adopted, and adoption depends on costs, regulation, culture, and organizational choices. This qualification prevents technological determinism. At the same time, the long-run conclusion remains conditional: machines must continue improving, employers must find them economically attractive, and new demand must not create enough distinct human work. Historical evidence cannot settle those future conditions. It can, however, show that growth and employment can separate, and that transitions can be painful even when aggregate employment remains high.
The policy discussion follows logically from the labor-market diagnosis. If wages are the main route through which prosperity is shared, reduced labor demand makes ownership, taxation, and public transfers central political questions. The proposed state roles also expose a tension the book cannot resolve by economics alone: redistribution requires social legitimacy. A conditional income may answer concerns about reciprocity, yet it risks making basic security depend on official judgments about acceptable contribution. Similarly, oversight of technology companies must constrain private power without granting unchecked power to government.
The deepest challenge is cultural. A society organized around employment treats wages as both income and evidence of social worth. Replacing that system requires not only new fiscal mechanisms, but new ways to recognize care, civic service, creativity, and leisure. The book is therefore less a prediction of a jobless world than a warning that institutions built for the Age of Labour may fail before work itself disappears.
Review Summary
A World Without Work explores the impact of automation on employment, arguing that technological progress will lead to widespread job displacement. While some reviewers praise Susskind's balanced approach and thought-provoking ideas, others criticize his solutions as unrealistic. The book discusses potential responses to technological unemployment, including education reform, taxation, and a conditional basic income. Many readers found the book informative and well-researched, though some felt it lacked practical solutions. Overall, reviewers appreciated Susskind's analysis of the challenges posed by AI and automation in the future job market.
People Also Read
Glossary
Age of Labour
Technology broadly benefits workersSusskind’s name for the period in which technological progress has, overall, expanded demand for human labor enough to offset the tasks it replaces. The term does not imply that every worker benefits equally or that change is gentle. It describes a historical balance in which complementary effects have generally outweighed substitution.
ALM hypothesis
Automation follows task characteristicsThe Autor-Levy-Murnane hypothesis explains how computerization can reshape employment by automating routine tasks, meaning tasks that can be described through explicit steps. Work requiring tacit knowledge or less easily specified activity was initially expected to resist automation. The framework helped explain why middle-paying, routine-heavy jobs declined while employment grew at both ends of the wage distribution.
Pragmatist revolution
Build machines by task performanceThe shift in artificial intelligence from primarily copying human thought to building systems that perform tasks effectively by other means. Deep Blue’s chess success illustrates the approach: it searched many possible moves rather than reproducing a human champion’s reasoning. The machine’s practical performance, not its resemblance to human intelligence, becomes the central measure.
Task encroachment
Machines expand into human tasksThe continuing process by which machines take on activities once performed by people. It can affect manual, cognitive, and affective work, and does not require machines to imitate the human abilities used in those tasks. The concept focuses on the gradual expansion of machine capability rather than a sudden disappearance of whole occupations.
Frictional technological unemployment
Available jobs remain inaccessibleA form of technological unemployment in which work still exists, but displaced workers cannot reach it. Barriers may involve missing skills, an identity conflict with lower-status or different work, or the geographic distance between workers and jobs. These frictions can persist even when the economy has vacancies and can also push people out of the labor force.
Structural technological unemployment
Too few jobs for workersA condition in which the economy does not generate enough demand for human labor to employ everyone seeking work. Susskind argues that automation could both displace workers and weaken the historical forces that created new demand for them. This differs from frictional unemployment, where jobs exist but particular workers cannot access them.
Superiority assumption
Humans remain best for new tasksThe assumption that people will remain better placed than machines to perform the tasks created by economic growth, changing consumer demand, or new production methods. Susskind argues that this has often been a reasonable historical expectation, but becomes less reliable as machine capabilities expand. If machines perform newly valuable tasks, growth need not create human employment.
Big State
State redistributes technological prosperitySusskind’s proposed state response to an economy where wages no longer distribute prosperity reliably. The state would help share income, broaden ownership of productive capital, and support workers during the transition. The proposal is not central planning of production. Its primary aim is to distribute prosperity and preserve social stability.
Conditional basic income
Income support tied to contributionA proposed public payment for people who cannot rely on paid employment, with eligibility or continued receipt linked to a contribution to the community. Such contributions might include care, education, cultural activity, or civic participation. Unlike an unconditional universal basic income, the CBI explicitly addresses questions of membership and reciprocity, though deciding fair conditions would be politically difficult.
Political Power Oversight Authority
Regulator for technology’s civic powerSusskind’s proposed independent institution for scrutinizing technology companies’ influence over liberty, democracy, and social justice. It would investigate systems, require transparency about their operation, and have powers to constrain harmful conduct. Its purpose differs from competition regulation, which focuses mainly on market power and consumer welfare.
FAQ
What's A World Without Work about?
- Explores technological unemployment: The book examines how advancements in technology and automation are leading to a future with significantly less work available for humans.
- Historical context provided: Daniel Susskind uses past instances of automation anxiety to illustrate how fears of job loss have historically been misplaced, but warns that current technological advancements may lead to real unemployment.
- Focus on societal response: Susskind emphasizes the need for society to rethink how we share economic prosperity in a world where traditional employment may not be sufficient for everyone.
Why should I read A World Without Work?
- Timely and relevant topic: As automation and AI continue to evolve, understanding their implications on employment is crucial for everyone, from policymakers to workers.
- In-depth analysis: The book provides a thorough examination of the historical, economic, and social factors surrounding work and technology, making it a comprehensive resource.
- Proposes solutions: Susskind not only identifies the problems but also suggests potential responses, including the role of a "Big State" in addressing inequality and unemployment.
What are the key takeaways of A World Without Work?
- Technological unemployment is real: Susskind argues that unlike past fears, the threat of technological unemployment is imminent due to advancements in AI and automation.
- Shift in demand for work: The book highlights that as machines take over more tasks, the demand for human labor will diminish, leading to a potential future with insufficient work.
- Need for new economic models: Susskind advocates for rethinking how society distributes wealth and prosperity, suggesting that traditional labor markets may no longer suffice.
What is the "superiority assumption" in A World Without Work?
- Definition of the assumption: The superiority assumption is the belief that human beings will always be better suited to perform tasks than machines, which has historically been true but may not hold in the future.
- Implications for the future: As machines become more capable, this assumption may lead to a significant decline in the demand for human labor, as machines will increasingly take over tasks once thought to require human skills.
- Challenge to traditional views: Susskind argues that this assumption needs to be reevaluated, as it could result in a world where machines dominate the labor market.
What is "task encroachment" in the context of A World Without Work?
- Definition of task encroachment: Task encroachment refers to the gradual process by which machines take over tasks that were traditionally performed by humans, leading to a shrinking set of activities for people.
- Impact on employment: As machines encroach on more tasks, the range of jobs available to humans will diminish, potentially leading to structural technological unemployment.
- Historical examples: Susskind illustrates this concept with examples from agriculture and manufacturing, where increased automation has already reduced the number of jobs available.
What is the Conditional Basic Income (CBI) proposed in A World Without Work?
- Definition of CBI: The CBI is a form of income support that provides regular payments to individuals, but with conditions attached. Recipients are required to contribute to their communities in meaningful ways.
- Encouraging community contribution: Susskind argues that the CBI can help maintain social solidarity by ensuring that everyone contributes to the collective good, even if they are not engaged in paid work.
- Flexibility in implementation: The specifics of what contributions are required can vary by community, allowing for tailored approaches that reflect local values and needs.
What role does the Big State play in Susskind's vision in A World Without Work?
- Taxation and redistribution: The Big State is envisioned as a mechanism for taxing wealth generated by automation and redistributing it to those affected by job loss.
- Support for non-economic contributions: The Big State would facilitate the CBI, encouraging individuals to engage in community service or other meaningful activities.
- Regulation of Big Tech: Susskind argues that the Big State must regulate the growing power of technology companies to prevent abuses of political influence.
What are the implications of technological unemployment discussed in A World Without Work?
- Shift in job availability: Susskind highlights that as machines take over more tasks, the number of available jobs will decrease, leading to widespread unemployment.
- Economic and social consequences: The book discusses how technological unemployment could exacerbate existing inequalities and create social unrest.
- Need for new policies: Susskind argues that traditional labor market policies will be insufficient in addressing the challenges posed by technological unemployment.
How does Susskind address the issue of meaning in a world with less work in A World Without Work?
- Importance of purpose: Susskind argues that work has traditionally provided individuals with a sense of purpose and identity.
- Community contributions: The CBI is proposed as a way to encourage individuals to engage in community service or other meaningful activities.
- Rethinking values: Susskind suggests that society must reevaluate what is considered valuable and meaningful, moving beyond traditional economic measures.
What are the best quotes from A World Without Work and what do they mean?
- "Will there be enough work for everyone to do in the twenty-first century?": This quote encapsulates the central question of the book, highlighting the urgency of addressing the future of work in light of technological advancements.
- "The future depends on ourselves, and we do not depend on any historical necessity.": This quote emphasizes the importance of human agency in shaping the future, suggesting that while technology will advance, society has the power to influence its impact.
- "Technological unemployment is simply a more extreme version of that story.": This quote connects the historical context of economic inequality with the emerging threat of technological unemployment, underscoring the need for proactive solutions.
How does A World Without Work relate to current discussions about AI and automation?
- Relevance to ongoing debates: The book provides a framework for understanding the implications of AI and automation on employment, making it a timely contribution to current discussions.
- Historical perspective: By examining past instances of automation anxiety, Susskind contextualizes today's fears and challenges.
- Call for action: The book urges readers to consider not just the technological advancements themselves, but also the societal and economic structures that will need to adapt in response to these changes.
How does Susskind envision the future of work in A World Without Work?
- Less paid work: Susskind predicts a future where traditional paid work is significantly reduced due to automation and AI.
- New forms of contribution: He envisions a landscape where individuals engage in various non-economic activities that contribute to their communities.
- Evolving social structures: The future will necessitate a rethinking of social structures and policies to support individuals in finding meaning and purpose outside of traditional employment.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.