Part Two of a Series on the Coming Labour Disruption
No Safe Ground: The Secondary Economic Consequences of AI-Driven Displacement
The Comforting Error
The most common mistake people make when thinking about artificial intelligence and work is to assume that if their own role cannot be directly automated, they are therefore relatively safe.
The electrician assumes this because wiring a building requires physical presence, fine motor coordination, and adaptation to messy real-world conditions. The nurse assumes it because care is relational and embodied. The plumber assumes it because pipes do not repair themselves through a chatbot. The teacher assumes it because classrooms are social environments, not merely information transfer systems. Even many tradespeople and public-sector workers who have followed the AI story with some seriousness still tend to think of it as a problem for programmers, accountants, copywriters, analysts, and perhaps eventually lawyers.
This is an understandable mistake. It is also a profound one.
The first-order effects of AI displacement are indeed concentrated in occupations that involve information processing, document generation, pattern recognition, routine communication, and structured analysis. Anthropic’s early labour market evidence found AI use concentrated most heavily in software development and technical writing tasks, with the highest exposure among computer and mathematical occupations and office support roles (Anthropic, 2026). Goldman Sachs has estimated that generative AI could expose the equivalent of 300 million full-time jobs globally to automation or significant task substitution (Goldman Sachs, 2026). The IMF has warned that 40% of jobs worldwide and 60% in advanced economies will be affected by AI, whether through augmentation or direct replacement (Georgieva, 2024).
It is at this point that many workers outside those domains relax. The mistake lies in imagining that the consequences end where the automation ends.
The Secondary Shock
An economy is not a collection of isolated jobs. It is an interdependent system of demand, income, tax flows, and local exchange. When millions of workers lose income, the effects do not remain confined to those workers. The demand they represented for other people’s goods and services disappears with them.
This mechanism is simple enough to describe and devastating enough to matter. A white-collar worker loses a job in marketing, law, finance, software, or administration because the role is automated or radically compressed by AI. That worker then spends less at restaurants, delays home repairs, cancels subscriptions, postpones dental work, buys fewer retail goods, contributes less to the local economy, and may stop discretionary spending almost entirely. The businesses and workers who depended on that spending—many of whom cannot themselves be directly automated—experience reduced demand. Some cut staff. Some close. The contraction spreads.
Research on labour market spillovers has long shown that job losses in tradable or high-income sectors produce downstream employment losses in local services. The mechanism is not new. What is new is the speed, scale, and cross-sector simultaneity with which AI may trigger it. Brookings has noted that AI is increasingly transforming middle-class jobs, particularly those in professional and administrative domains, and that these roles are deeply entangled with consumption patterns that sustain broad parts of the wider economy (Brookings Institution, 2026). When the middle-class wage base contracts, the service economy built around it becomes unstable.
This is why the distinction between direct and indirect exposure is less reassuring than it first appears. A role can be impossible to automate and still be impossible to sustain economically in a community whose customers no longer have the income to support it.
No Insulation in a Demand Economy
The modern service economy is far more dependent on discretionary middle-class spending than most people realize. Local restaurants, repair services, childcare providers, gyms, salons, retailers, trades, therapists, tutors, and a wide array of independent and small-business professionals depend not merely on population, but on income-bearing population with money left after essentials are paid.
That fact matters enormously in an AI transition. The sectors most directly exposed to AI—professional services, administration, information work, support roles, routine analysis—are disproportionately middle-class sectors. These are precisely the people whose incomes circulate through local economies in stabilizing ways. Their spending supports not only consumer-facing businesses but the broader web of municipal tax revenues, charitable donations, school fundraising, housing demand, and civic life that makes communities livable.
Yahoo Finance reported in April 2026 on projections that AI could destroy 25 million U.S. jobs, citing warnings of unemployment rates reaching 20% to 30% in severe scenarios (Yahoo Finance, 2026). Even if such estimates prove directionally overstated, they highlight the central issue: once job loss reaches sufficient scale among wage earners with substantial spending power, secondary damage becomes inevitable. The tradesperson who cannot be replaced by AI still depends on clients who can pay. The human services worker still depends on tax revenues and public budgets. The restaurant server still depends on households willing to eat out. Physical presence does not create economic immunity.
Cognizant’s 2026 analysis argues that AI is creating a tax rebalancing problem precisely because labour income has historically functioned as the foundation of public and private economic circulation (Cognizant, 2026). Remove the labour income and the circulation weakens. Remove enough of it, and the system no longer behaves like a functioning consumer economy.
The Local Community Effect
The secondary effects of displacement are likely to be felt first not at the national level but in local communities. This is where the abstraction becomes most concrete.
A city or town can withstand isolated layoffs. It cannot easily withstand a structural reduction in the incomes of thousands of workers whose spending sustains local business ecosystems. Reduced retail demand affects commercial rent viability. Falling restaurant traffic affects suppliers, landlords, and employees. Lower home maintenance spending affects trades and materials providers. Declining municipal revenue affects public services. A labour market shock that begins in “knowledge work” becomes, in a short period, a community-wide contraction.
This dynamic is especially dangerous in smaller centres and regional communities where economic diversification is limited and where a relatively small number of salary-bearing roles support much of the surrounding demand structure. The worker displaced from an insurance office, school administration role, regional bank, engineering firm, government office, or logistics coordination centre is not merely one income gone. It is a node of local economic circulation removed.
The result is that workers whose occupations depend on place—those who quite reasonably believe that AI cannot physically show up to do their jobs—discover that the place itself has become economically unstable. Their labour remains necessary in theory but unaffordable in practice.
Public Systems Feel It Too
The contagion does not stop with private demand. Public systems are deeply exposed to the same contraction.
When wage income declines, so do income tax receipts, payroll contributions, and consumption tax revenues. At the same time, demand for social supports rises. Unemployment benefits expand. Mental health pressures increase. Housing insecurity grows. Training and transition programs become more necessary at exactly the moment governments have less fiscal room to provide them. This was the central argument of Part Three, but it begins here: in the secondary effects of labour displacement spreading through public finance.
Workers in education, social services, healthcare administration, community programming, and local government may imagine themselves sheltered because they are paid by institutions rather than markets. But those institutions are themselves funded by tax bases that depend on employed people. The state is not external to the labour market. It is downstream from it.
This matters for understanding why there is no stable perimeter around the AI shock. The teacher may not be directly replaced by AI. The school budget may still be. The social worker may still be needed. The program may still be cut. The nurse may remain essential. The public system employing the nurse may still freeze hiring, compress wages, or reduce services because the tax base beneath it is eroding.
The False Refuge of the Trades
Among all the occupations people point to as durable against AI, the skilled trades are the most plausible. This is not wrong. It is simply incomplete.
Electricians, plumbers, carpenters, welders, mechanics, HVAC technicians, and others in the trades perform embodied, site-specific, non-routine work in unpredictable physical environments. These characteristics do make direct automation more difficult. But difficulty of automation is not the same as insulation from economic disruption.
Trades rely on three things: clients with income, businesses with capital budgets, and public infrastructure spending. All three come under pressure in a large-scale displacement event. Households defer renovations. Businesses delay expansions. Governments cut or postpone capital projects under fiscal strain. New housing starts soften as purchasing power falls and credit conditions tighten. The trades survive better than many white-collar occupations in the first phase of AI disruption. But surviving the first phase is not the same as standing outside the system altogether.
In other words, the trades are safer than many jobs. They are not safe from the economy in which those jobs exist.
The Psychological Trap
Part of why this secondary risk remains underappreciated is psychological. People think concretely about threats. If they cannot imagine an AI physically doing their work, they conclude the threat does not meaningfully apply to them. This is a version of optimism bias and scope neglect: individuals underestimate systemic risks when those risks do not resemble a direct one-to-one substitution of self by machine.
There is also a moral comfort in believing that physically necessary work will be rewarded and preserved in a way abstract white-collar work may not be. It feels just. It feels grounded. It feels like reality reasserting itself against the excesses of digital capitalism. There is some truth in that instinct. But it should not be confused with economic immunity.
The problem is not merely whether AI can do your job. The problem is whether enough other people can still afford to pay for it, whether the institution employing you can still fund it, and whether the broader economic web around your work remains intact. Once the question is framed properly, the sense of safety becomes much harder to sustain.
The Conclusion the Evidence Supports
The coming AI disruption should not be thought of as a narrow technological substitution event affecting a limited set of directly automatable occupations. It should be understood as a demand shock, a tax shock, and a community shock that begins in directly exposed white-collar work and propagates outward into the wider economy.
There is no meaningful safe ground in an economy where the spending power of millions of workers disappears at once. Some occupations will hold up longer than others. Some sectors will prove more durable in the early stages. But durability is not immunity, and the absence of direct automability is not a shield against secondary collapse.
This raises a more uncomfortable implication still. If the roles that remain most human, most physical, and most relational are not enough by themselves to secure economic stability, then survival in the coming transition depends on something deeper than occupation. It depends on a person’s ability to think across contexts, adapt under pressure, read changing systems clearly, and create value under novel conditions. These are the capabilities that remain relevant when no role, sector, or institution can be assumed stable.
The tragedy is that these capabilities are not being developed at anything like the scale the moment requires. The institutions that claimed to produce them have, for decades, issued credentials more reliably than they have developed real context-independent capacity. That failure belongs to the next stage of this argument. But the conclusion here is already clear enough: in an economy shaped by AI-driven displacement, no one remains unaffected simply because their own hands still have work to do.
References
Anthropic. (2026, February 25). Labor market impacts of AI: A new measure and early evidence. https://www.anthropic.com/research/labor-market-impacts
Brookings Institution. (2026, March 1). Artificial intelligence is transforming middle-class jobs. Can it also help the poor? https://www.brookings.edu/articles/ai-transforming-middle-class-jobs-can-it-help-the-poor/
Cognizant. (2026, April 8). AI job displacement: When capital can think, who pays? https://www.cognizant.com/us/en/insights/insights-blog/ai-impact-on-jobs-and-tax-rebalancing
Georgieva, K. (2024, January 14). AI will affect 40% of jobs and likely worsen inequality. International Monetary Fund. https://www.imf.org/en/Blogs/Articles/2024/01/14/ai-will-affect-40-percent-of-jobs-and-likely-worsen-inequality
Goldman Sachs. (2026, March 17). How will AI affect the U.S. labor market? https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-us-labor-market
Yahoo Finance. (2026, April 20). AI could destroy 25 million U.S. jobs. https://finance.yahoo.com/economy/policy/articles/ai-could-destroy-25-million-141527754.html