The Coming Unrest
Automation, Mass Unemployment, and the Breakdown of Stability
(Revised Version Jan 18, 2026)
1) From Automation of Automation to Mass Unemployment
Artificial intelligence is causing automation to accelerate faster than ever. With AI as a propellant, the work of building automation increasingly becomes automatable itself. As mechanizing tasks becomes cheaper and easier to implement, automation propagates faster as it expands. As a result, the current wave of AI expansion is likely to produce mass disruption in the labor market.
In earlier automation waves, the limiting factor was not whether a task could be automated in principle. It was the time and coordination required to integrate tools into real operations—workflow redesign, edge-case handling, monitoring, maintenance, and training. Those costs created a lag between capability and displacement. AI compresses that lag by making the integration and iteration layer cheaper and faster.
The mechanism is straightforward. AI reduces the time and coordination required to convert work into systems, so adoption often arrives as incremental workflow insertions rather than dramatic replacement events. As the work of building automation becomes automatable itself, those insertions begin to compound: each successful integration leaves behind reusable components—scripts, templates, agent patterns, monitoring, connectors—that make the next integration cheaper and faster. This produces superlinear acceleration in adoption and an exponential-like growth dynamic in automation capacity.
Firms then translate throughput gains into staffing outcomes by holding headcount flat, consolidating responsibilities, and suppressing replacement hiring.
At the labor-market level, the first signature is a narrowing flow of openings—roles that never appear, backfills that never happen. Over time, that can accumulate into structurally elevated unemployment even if headline employment looks stable at first.
Why AI Adoption Moves Fast
Four features make adoption fast and quietly consequential: composability, friction collapse, digital legibility, and ergonomic packaging.
First: composability turns speedups into a throughput regime. AI links steps that once required separate people, separate tools, and multiple handoffs. When language work becomes modular inside a single workflow, small gains compound: less waiting, less context translation, fewer format repairs. As the integration “glue” becomes faster to build and easier to keep current, more work becomes systems—and systems don’t ask for headcount.
Second: friction collapse changes the economics of knowledge work. When AI turns many tasks into cheap first passes followed by rapid iteration, coordination-heavy work can run as a tight loop: draft, check, revise, repeat. As iteration becomes inexpensive, firms standardize processes that previously relied on slow, bespoke human effort. The immediate effect is less labor time per deliverable—and a shift in what is “worth automating.” As the cost curve moves, the automation frontier expands: more work becomes “good enough” to integrate into production, including tasks that used to be marginal because they required extensive drafting and rework.
Third: digital legibility accelerates adoption because much of white-collar work already exists in machine-addressable form. Inputs and outputs are digital, standardized, and trackable, so AI can enter workflows without new sensors or physical infrastructure. It operates directly on the representations firms already use to coordinate work. This makes adoption incremental: AI is inserted as a thin layer inside existing tools and routines, handling first-pass language work and simple coordination tasks. These insertions look like speedups, but in aggregate they reduce the labor required per unit of output.
Fourth: ergonomic packaging widens the adoption surface area. When agentic capability is delivered through a familiar, task-first interface with permissioned access to folders and connectors—rather than a CLI—it becomes usable by nontechnical roles without engineering mediation.[18] Recent desktop Cowork releases show how quickly this packaging can move agentic workflows from a niche tool to a default layer in ordinary operations. This increases the pace of delegation: more tasks become offloadable, more workflows become repeatable, and compounding effects can arrive before they are narrated as displacement.
How Capability Becomes Headcount Reduction Inside Firms
Automation becomes socially and politically dangerous when capability converts into staffing outcomes. That conversion does not require a dramatic replacement moment. In practice, contraction often occurs through non-backfilling and attrition absorption: someone leaves, the role is not replaced, work is redistributed, AI covers portions of what used to justify a full position, and the remainder is merged into adjacent jobs. Headcount falls without a single clear “replacement” event.
Once this pattern stabilizes, it produces a second dynamic: the expectation ratchet. Productivity gains become the new baseline. Output targets rise, response times shorten, scope expands, and teams are expected to produce more with the same number of people. Over time, that becomes “normal.” The consequence is not merely that AI assists workers; it changes what counts as a standard workload and reduces the number of roles required to sustain a given level of throughput.
This does not deny that AI can complement labor and expand some kinds of work. In some sectors, adoption will raise demand for certain skills, and new job categories will appear. The claim is narrower: even with pockets of job creation, the net flow of openings can still shrink if headcount compression outpaces new-role creation—and if the new roles are fewer, harder to access, or concentrated in particular firms and regions.
Mass unemployment can arrive through the labor market’s flow before it shows up in headline employment levels. A labor market is, in part, a system for continually creating and refilling roles. It needs expansion and replacement hiring. It also needs entry points and recovery hiring. If firms learn that they can maintain or increase throughput while suppressing replacement hiring, the flow of new positions narrows even when revenue is stable. A macro-visible signature of this shift is not mass firings; it is missing jobs—roles that never open, backfills that never materialize, opportunities that quietly vanish. That is the leading edge of a deeper shift that can accumulate into mass unemployment: sustained, structurally elevated joblessness and underemployment (including labor-force exit), not a short recessionary spike.
The Missing-Jobs Signature
At the labor-market level, destabilization begins as a ratio problem: whether there are enough openings for the number of people who need to move—new entrants, displaced workers, and anyone trying to exit deteriorating roles. When openings shrink faster than people can adapt, the system becomes brittle even if many remain employed.
This is why the first wave appears as missing jobs. Job postings thin out. Entry-level roles narrow. Hiring cycles lengthen. Firms redistribute work internally instead of creating new positions. Unemployment may not spike at first, but the labor market stops providing the same number of entry points and second chances.[1][2]
The visibility problem is structural. Many changes occur inside stable job titles. Teams keep the same roles on paper while the task mix shifts: more AI-assisted drafting, more templating, more automated triage, fewer routine production hours. Headline employment can look steady even as the labor content of jobs shrinks and the opportunity structure deteriorates.
Measurement distortion compounds this further. Shadow AI spreads unevenly and is undercounted. Productivity changes are attributed to “better tooling” or “working smarter,” while the underlying driver—a new capability embedded inside workflow—remains partially invisible. The combined effect is an upward shift in output per worker, weakening the historical link between growth and proportional job creation.[3][4]
Taken together, these mechanisms produce a labor-demand shock that rarely announces itself as a single event. Adoption is broad and distributed. It arrives as incremental workflow insertions that compound over time and shift where human effort is required. Firms translate those gains into staffing outcomes by not backfilling roles and consolidating work. Baseline expectations rise. Even as employment snapshots look stable, the openings pipeline can narrow. Missing jobs become the early signature: fewer openings, fewer entry points, and slower churn.
A “quiet” employment shock can be mistaken for a temporary freeze or cyclical cooling. Silent reduction lowers headcount through non-backfilling rather than announcements. Counterfactual job loss accumulates when roles that would have existed in prior cycles simply never appear.[5][6] Output can rise, or remain stable, without proportional hiring, because AI shifts more of the workload into tools and processes rather than payroll.
So where does mass unemployment enter—concretely, rather than rhetorically?
It enters when the missing-jobs phase stops being transitional and becomes durable: a persistent narrowing of openings that outlasts a business cycle, combines with compounding task substitution, and keeps lowering the cost of converting work into systems—so the market’s usual ability to absorb displacement through churn fails. In that world, unemployment is not merely a recession spike. It is structural residue: the labor-demand curve has moved.
A labor market can look “mostly employed” for a time while mobility and second chances degrade underneath. But the direction of travel is what matters. When firms can raise throughput while suppressing hiring, and when those gains compound across workflows and vendors, the system stops reliably converting growth into jobs. The early phase is a door quietly narrowing. The late phase is a door closed.
Feedback: Insecurity and Demand
Insecurity also feeds back into demand. When households treat the future as precarious, they cut discretionary spending and delay commitments. Demand softens, firms tighten, and automation becomes the easiest way to hold output while trimming payroll. The result is a reinforcing loop: weaker demand accelerates substitution, further shrinking openings.
Once openings thin, outside options weaken—and a labor-market shock becomes a social and political shock.
2) Domestic Disruption and Unrest
The employment shock matters because it does not stay inside the labor market. When automation-driven contraction moves from missing jobs to mass unemployment, the core failure is not only lost income but the loss of credible futures. People can endure hardship when they believe there is a path back. They destabilize when the path disappears.
In this context, domestic refers to the internal dynamics within any given country—its labor market, institutions, and social order—rather than cross-border spillovers. The focus is largely on advanced democracies, but the mechanisms described are general.
Defining System Disruption
“Disrupt the system” does not mean “society collapses.” It means the stabilizing routines of a modern social order stop working reliably under sustained labor-market pressure. When the economy no longer produces broadly accessible mobility, the moral story that stabilizes the system weakens: people stop believing that effort is rewarded and that institutions are impartial. As insecurity spreads, political time horizons shorten. Governments cycle through “fixes,” emergency measures, and symbolic actions; policy becomes reactive, and reversals become common.
At the same time, public systems face rising demands for relief, adjudication, and enforcement just as budgets tighten and trust erodes. Backlogs grow. Service quality declines. Governance drifts from provision toward control because control is easier to demonstrate than repair. As exits close and competition increases, politics is pulled toward sorting—who deserves protection, who is blamed, and who must be excluded. These are not separate problems. They are linked outcomes of a society attempting to stabilize itself when the labor market stops functioning as the primary allocator of dignity, security, and forward motion.
How Disruption Propagates
The key transmission variable is not merely unemployment; it is the strength of outside options—the practical alternatives that allow people to say “no”: the ability to leave a bad job, reject unfair conditions, relocate, retrain into a real opening, or start over without catastrophic loss. When outside options are strong, conflict stays containable. Workers can exit rather than escalate. Employers must compete rather than coerce. Institutions can govern through rules rather than force.
Automation-driven contraction weakens outside options in two stages. First, during the missing-jobs phase, the openings pipeline narrows: fewer roles to move into, fewer entry points, fewer second chances.[1][2] Then, as mass unemployment becomes visible, competition intensifies and the cost of refusal rises. Under those conditions, “choice” becomes formal rather than real. The labor market stops absorbing risk; it concentrates risk.
This is how a labor-market shift becomes a social shift. Constraint produces compliance. It also produces resentment, because people feel trapped. A society with weak outside options behaves differently. It is more fearful and brittle, and easier to polarize. The political system, facing mass insecurity, becomes the arena where people seek protection, compensation, and explanation.
The collapse of outside options produces security collapse. Small failures no longer stay small. A reduced-income month becomes a rent problem. A health issue becomes a debt problem. A schedule disruption becomes a childcare problem. Each pressure increases the cost of the next, and vulnerability cascades.
These vulnerabilities also stack across domains. The fear stack is the accumulation of exposures that makes households feel one mistake away from collapse. People respond rationally by tightening spending and avoiding moves that add risk. The lived experience is a narrowing corridor where small shocks can become irreversible.
Outside options collapsing also changes the terms of work. When exit is costly, employers can tighten schedules, intensify monitoring, and erode benefits with less risk of losing staff. Even without layoffs, workers feel extraction rising: more output is demanded, fewer gains are shared, and the promise that effort buys mobility stops holding. That mismatch—higher productivity expectations alongside narrowing opportunity—reads as illegitimate.
Defining Unrest
Unrest does not mean a single explosion. It begins with blocked exit: when enough people can no longer improve their situation through ordinary mobility, conflict spills into public space and hardens into zero-sum political competition. Unrest is a sustained rise in contention and institutional volatility—more strikes and protests, more flashpoint incidents that turn disruptive, and conflict that becomes normal in daily life. Some of it is visible in the street—protests, riots, localized disruptions—but much of it appears as churn and overload: sharper electoral swings, leadership instability, persistent polarization, and public systems that cannot keep up.
Unrest is thus not merely “anger.” It is a pattern of behavior in a society that absorbs shocks poorly and resolves conflict less peacefully because the underlying opportunity structure has narrowed.
Why Fast, Legible Levers Win
The same political pattern often emerges across different ideologies under disruption because the selection environment changes. Selection pressures are the incentives and constraints that cause certain strategies to spread and persist by winning attention, legitimacy, and compliance under stress. As insecurity widens, politics becomes a competition over signals: what looks like protection, what looks like punishment, and what looks like control.
Slow repair—building housing, fixing schools, expanding healthcare capacity, rebuilding labor-market absorption—takes time. It is costly, complex, and its benefits arrive late. Under high insecurity, late benefits lose to immediate signals. That creates an advantage for fast, legible levers: policies that are visible, emotionally satisfying, and easy to narrate.
These levers often take three forms: tighter enforcement, narrower eligibility and exclusion, and expanded surveillance or emergency powers. The details vary by regime, but the shared feature is that the lever produces an immediate, legible output: “control is being asserted.”[7][8]
Visible local AI costs—especially data-center buildouts, grid strain, and rising electricity prices—can make automation politically salient early, reinforcing the appeal of fast, legible levers.[9][10]
This is how a society can become more coercive even without a single ideological cause. It is selected for. Under sustained labor-market contraction, politics becomes less about maximizing long-run capacity and more about producing short-run legitimacy under fear. That shift increases volatility and raises the baseline probability of misfires—because fast levers are often blunt, and blunt levers create backlash.
The system enters a loop: insecurity selects for control signals; control signals degrade capacity; degraded capacity increases insecurity. This is the domestic core of the argument. Missing jobs become mass unemployment. Mass unemployment weakens outside options. Weak outside options spread insecurity. Insecurity pushes strain into politics and institutions. Under strain, the system selects for fast, legible levers. Those levers stabilize perception in the short run while increasing brittleness in the long run—setting the conditions for sustained domestic unrest.
The risk increases sharply when these dynamics operate synchronously across countries, rather than being redistributed through migration, trade, and capital flows.
3) Global Synchronization
The domestic story is already destabilizing. The global story is worse because the same technologies, incentives, and market structures tend to propagate across borders and supply chains. When AI adoption is widespread, disruption does not remain local. It synchronizes.
Synchronization matters for one reason: it removes the escape valves that usually soften shocks. In past transitions, people and capital could move to less-affected sectors, regions, or countries. When the shock is broadly shared—arriving through the same platforms, workflows, and competitive pressures—those alternatives weaken. The system becomes less able to redistribute stress. Volatility rises.
Why Synchronization Is Likely
AI tools diffuse fast because they travel as software. They are delivered through global vendors, integrated into widely shared platforms, and copied through competitive imitation. A capability that reduces costs in one market becomes a competitive requirement in others. Firms that adopt gain an advantage; firms that delay face margin pressure. That pressure forces convergence.
This dynamic extends beyond headline “AI companies.” It spreads through procurement, consulting, outsourcing, and enterprise software. Once AI becomes a default layer inside common tools—customer service, document processing, marketing, compliance, internal analytics—adoption becomes background infrastructure rather than a discrete decision. The result is simultaneous exposure: many organizations, in many places, making similar labor-substituting changes within the same time window.
Synchronization is also amplified by the fact that modern economies share production networks. Supply chains, logistics, services outsourcing, and financial conditions link local labor markets to global demand. When firms everywhere pursue the same productivity strategy, the shock propagates through trade and investment rather than remaining contained inside one country.
Economies Most Exposed
The vulnerability is not simply “poor countries” versus “rich countries.” The key question is what an economy exports and how it captures value. Broadly, exposure falls into three overlapping categories.
First are economies that depend heavily on labor-intensive export manufacturing, especially where margins are thin and competition is global. If automation allows onshoring, nearshoring, or simply producing more with fewer workers, demand for low-cost labor can fall quickly. Even partial substitution can be destabilizing when employment is concentrated in a few industries.
Second are economies that function as global service back offices: call centers, business-process outsourcing, routine accounting and compliance services, basic software and IT services, and other forms of “white-collar labor export.”[11] These are often described as services-export economies. They can be hit hard because AI substitutes precisely the language-heavy, rules-driven, high-volume tasks that make these sectors scalable.[2][12]
Third are economies with large informal or underemployed labor pools that rely on steady absorption into formal roles—whether domestic or export-linked—to keep social expectations stable. When absorption slows, pressure accumulates quickly because there are fewer institutional buffers and fewer reliable pathways into secure work.
These categories overlap. A country can be exposed through manufacturing and services at the same time. And even countries that “benefit” from cheaper AI may suffer if global demand weakens or if domestic labor markets become brittle under the same missing-jobs dynamics described in Section 1.
How Synchronization Amplifies Unrest
When shocks are unsynchronized, affected people and regions can lean on external demand, remittances, migration, or new investment from less-affected zones. When shocks synchronize, those channels weaken simultaneously.
Migration becomes harder to treat as a pressure-release mechanism because destination labor markets also tighten. Remittances can fall when migrants face weaker outside options abroad. Trade-based recovery slows when many economies are attempting to cut labor costs in parallel. Even fiscal stabilization becomes harder when capital is risk-off and governments everywhere face rising demands at the same time.
This is not only an economic problem. It is a political one. When multiple societies experience tightening opportunity structures concurrently, the global environment becomes more reactive and less cooperative. States become more protectionist. Border politics intensifies. Scapegoating becomes more attractive. The same selection pressures described in Section 2—favoring fast, legible levers—begin to operate in multiple places at once, increasing the odds of policy cascades and feedback loops across borders.
The Global Instability Loop
Synchronization increases the probability of a self-reinforcing loop:
As labor markets tighten, insecurity rises. As insecurity rises, governments reach for visible control signals. Those signals often degrade long-run capacity and trust. Degraded capacity makes societies less resilient to the next shock. Meanwhile, global competition pushes firms to adopt the same labor-saving strategies, which tightens labor markets further.
In this environment, instability is not a one-off crisis. It becomes a condition: a world where multiple governments face the same pressures, in the same period, with fewer available escape valves. Domestic unrest is harder to contain when external relief mechanisms fail. External conflicts become more likely when internal pressures are high.[13][14] And policy mistakes propagate faster when everyone is improvising under similar constraints.
This is what global synchronization adds to the argument. It is not merely that AI can disrupt jobs everywhere. It is that it can cause many societies to lose their usual buffering mechanisms at the same time—making unrest more persistent, more widespread, and harder to de-escalate.
Under sustained stress, institutional responses tend to converge on a small set of stable endpoints—attractors.
Attractors Under Sustained Stress
Under sustained strain, political systems tend to fall into a small set of destination patterns. These are attractors: governance basins that become more likely once feedback loops dominate, legitimacy weakens, and ordinary stabilizers fail. Not every system ends the same way. Still, the range of stable configurations narrows.
One attractor is hardened authoritarian control: stability pursued through surveillance and coercive enforcement. Another is brittle populist cycling: repeated swings, escalating reversals, and chronic institutional conflict. A third is fragmented governance: authority splinters across regions and institutions, reducing coherent capacity and increasing local volatility. A fourth is managed adaptation: institutions preserve legitimacy by buffering losses and rebuilding opportunity flow, reducing the demand for coercive control.
4) The Brittle, Coercive Normal
If the trajectory described above holds, the outcome is not a single rupture. It is a new baseline: a society that continues to function, but with less slack, less trust, and more coercion built into the routine management of everyday life.
The defining feature of this baseline is brittleness. The system still produces goods and services. Institutions operate. Elections still occur. But conflict becomes harder to resolve peacefully because the underlying opportunity structure has narrowed. When fewer people believe they can exit bad conditions—bad jobs, bad regions, bad institutions—more conflicts become fights over enforcement rather than negotiations over terms.
Ambient Insecurity
In a labor market defined by missing jobs, mobility becomes rare and costly. People hold on to positions they would previously have left. They accept worse terms. They delay major life decisions. The future feels less like a set of options and more like a narrowing corridor.[15][16]
This does not require everyone to be unemployed. It requires enough people to experience constrained movement that insecurity becomes culturally ambient. Even those who remain employed behave differently when they see openings vanish around them. In that environment, narratives that explain hardship by naming enemies rather than mechanisms become easier to sell.
Governance in Crisis Mode
As insecurity spreads, demand rises for relief, adjudication, and enforcement. At the same time, institutions face declining legitimacy and tighter constraints. When the system cannot deliver broad mobility, it increasingly governs through containment.[17]
Instead of expanding capacity—housing, healthcare access, education quality, labor-market absorption—governance leans toward measures that are immediate and demonstrable. Eligibility rules tighten. Surveillance expands in the name of fraud prevention or safety. Enforcement becomes the default response to disorder because enforcement is legible. Provision is slow. Control is fast.
The political outcome is not uniformly “authoritarian” in a single ideological direction. It is a drift toward coercive management across multiple ideologies, because coercion is easier to signal under pressure than repair.
As these control logics expand, governance begins to feel like permanent crisis management.
Emergency authorities, exceptional measures, and punitive policies become normalized because they are repeatedly selected for in moments of strain—and strain becomes frequent, even without a single triggering event.
This “permanent emergency” mood accelerates institutional decay: rules become flexible, enforcement becomes uneven, and trust erodes further. It also makes de-escalation harder. When crisis governance becomes normal governance, actors adapt to it; they invest in it; they depend on it. The system becomes path-dependent, locked into a higher-coercion equilibrium.
Conflict Becomes Routine
In a brittle system, small shocks travel farther. A price spike, a local closure, a scandal, a protest, a weather event—each has more destabilizing potential because households and institutions have less slack. This raises the frequency of flashpoints: incidents that would once have been absorbed now produce wider disruption.
Conflict also becomes more openly zero-sum. As outside options weaken, groups compete for protection and favorable decisions—who gets relief, who gets policed, and who gets excluded. Politics becomes more about exclusion and status than shared growth. The baseline level of hostility rises because the system has reduced the set of mutually beneficial outcomes.
The core signature of this world is fewer exits and more pressure. When exits close, compliance rises and resentment deepens at the same time. People accept worse conditions because they must, and they search for explanations because the moral story that once stabilized effort and reward no longer fits. Institutions, under overload, choose what they can visibly do. Politics selects for fast, legible levers. Those levers stabilize perception in the short run while degrading resilience in the long run. The result is not a cinematic collapse. It is a harder normal: a society that can continue operating while becoming increasingly volatile, increasingly punitive, and increasingly unable to resolve conflict through opportunity rather than enforcement.
5) Conclusion
The engine is the automation of automation: AI lowers the cost of turning work into systems, so substitution and headcount compression compound. The early signature is missing jobs. The downstream risk is mass unemployment as firms normalize permanently lean operating models.
Once that shift takes hold, it does not stay inside the labor market. It collapses outside options—the practical alternatives that let people refuse bad conditions, exit failing roles, and restart without catastrophic loss. As outside options weaken, insecurity spreads beyond the unemployed to anyone whose future depends on a labor market that no longer provides credible mobility. That insecurity pushes strain into institutions: legitimacy stakes rise, time horizons shorten, and volatility increases.
Under strain, systems tend to select for fast, legible levers—visible control signals that stabilize perception quickly. The tradeoff is capacity. These levers harden boundaries, break coordination, and turn governance into a sequence of short-horizon shocks.
Unrest follows as contention becomes routine: more strikes and protests, more flashpoints, and more institutions operating at overload. The risk increases further when these dynamics synchronize globally. When multiple regions move through missing jobs, then mass unemployment, and then rising instability within the same window, the world has less slack everywhere. Trade, capital, migration, and information become transmission channels for stress rather than buffers.
The point is probability, not certainty. The timeline will vary by sector and by country. But the direction of travel is being set by incentives and capabilities that move faster than adaptation and governance typically can. Dangerous times rarely require a single dramatic moment. They arrive when societies lose room to maneuver. The question is whether institutions can build enough slack, fast enough, to keep brittleness from becoming breakdown.
#korovamode
Appendix A: Glossary
Missing jobs: The absence of expected openings and churn—the labor market’s flow narrows even when headline unemployment looks stable.
Silent reduction: Headcount decline through non-backfilling and attrition rather than announcements.
Counterfactual job loss: Jobs that would have existed in prior cycles but never appear under new production conditions.
Outside options: Credible alternatives that let people exit bad terms without catastrophic cost.
Selection pressures: Conditions that reward certain behaviors and institutions over others, regardless of intent.
Fast, legible levers: Interventions that are quick to deploy and easy to narrate as control signals.
Digital legibility: The extent to which work is already expressed in digital, standardized inputs and outputs—making it easy for AI to ingest, produce, and integrate into workflow.
Friction collapse: A sharp drop in the time, cost, and coordination needed to generate a usable first draft or action, shifting many tasks from “expensive to try” to “cheap to iterate.”
Composability: The ease with which AI capabilities can be chained with other tools, APIs, and workflows to automate multi-step work.
Automation of automation: The dynamic where AI reduces the cost of building and maintaining the integrations, tooling, and process “glue” that turns one-off assistance into repeatable systems—accelerating adoption and headcount compression.
Jevons’ Paradox: The pattern where efficiency gains lower effective cost and can increase total use; applied here as the claim that cheaper “cognition” may expand demand enough to offset job loss.
Luddite fallacy: The claim that technological job loss is self-correcting because new jobs will reliably appear to replace old ones; treated here as a contingent historical pattern rather than a guarantee.
Shadow AI: Informal, uneven AI use inside organizations that is undercounted in official metrics but still changes productivity and staffing needs.
Expectation ratchet: The dynamic where productivity gains become the new baseline, raising output expectations without proportional staffing increases.
Non-backfilling: A contraction mechanism where vacated roles are not replaced; work is redistributed and partially automated.
Attrition absorption: The practice of absorbing departures by spreading responsibilities across remaining staff, often with AI covering parts of the workload.
Ratio problem: The mismatch between the number of people who need to move (new entrants, displaced workers, and those exiting deteriorating roles) and the number of openings available.
Services-export economies: Economies that depend heavily on exporting routinized white-collar services (e.g., call centers, BPO, back-office work).
Security collapse: When thin buffers cause small shocks to cascade across rent, debt, health, and household logistics.
Fear stack: The accumulation of stacked exposures that makes households feel one mistake away from collapse.
Attractors: Common governance endpoints that become more likely under sustained stress (e.g., coercive control, cycling instability).
Global synchronization: When multiple countries experience labor-market pressure at the same time, weakening escape valves and policy slack.
Appendix B: FAQ
Does this require mass unemployment to be dangerous?
No. The early phase is often missing jobs plus degraded outside options, driven by compounding automation and headcount compression. That is enough to raise insecurity, tighten risk, and destabilize politics even while many remain employed.
Isn’t this just another automation wave?
Past automation waves often had three stabilizers: slow diffusion, narrow substitution, and large new labor-complementary build-outs. This wave weakens all three. AI spreads as software, not capital equipment. It targets routinized cognitive work across many sectors at once. And because AI reduces the cost of integrating and extending automation, adoption can compound rather than plateau. The result is not merely task replacement, but a faster conversion of work into repeatable process with less need for net new hiring.
Isn’t retraining the solution?
Training helps individuals. It cannot stabilize the system if destination openings do not exist at the required scale, or if automation keeps narrowing the flow of new roles.
If AI boosts productivity, why wouldn’t wages rise and create new jobs?
Productivity can rise without broad wage growth when bargaining power erodes and gains are captured as margin, price competition, or scale rather than compensation. In a missing-jobs regime, the threat of exit weakens: fewer openings mean weaker outside options, which compresses wage pressure even as output per worker rises. New jobs can still appear, but the system can fail to convert aggregate productivity into widely distributed purchasing power and labor demand—especially when the fastest gains occur in roles that previously served as large employment buffers.
Doesn’t Jevons’ Paradox imply more AI means more jobs?
Not necessarily. Jevons’ Paradox can be real: efficiency lowers effective cost and can expand total demand. The question is whether demand expands fast enough, and in domains that require large amounts of human labor, to offset contraction. In this essay’s mechanism, automation increasingly lowers the cost of automating itself, so substitution can propagate across many workflows at once. Firms can capture the gains as cycle time, margin, and scale without proportional hiring. Demand expansion is most stabilizing when growth remains strongly labor-complementary. If expansion is itself increasingly automatable, more output does not imply broad job growth.
Isn’t this just the Luddite fallacy?
The “always creates new jobs” reassurance is often treated as a law. It is not. Historically, new industries and tasks have absorbed displaced workers over time, but that outcome depends on diffusion speed, substitution breadth, demand elasticity, and how gains are distributed. This essay does not claim AI creates no new work. It claims that automation of automation can accelerate adoption and headcount compression across many roles faster than new openings appear—especially in entry-level and routine cognitive layers. In that regime, job creation can still occur, but not at the scale or tempo required to preserve stable labor-market flow and bargaining conditions.
What would we expect to see first, if this is true?
The earliest signs are often not layoffs. They are a narrowing of labor-market flow: fewer entry-level roles, fewer “backfill” requisitions, and slower re-opening of positions after departures. Organizations quietly absorb attrition, raise output expectations, and standardize work so it becomes more automatable. Hiring pipelines distort: more “open” postings with delayed or selective fills, longer searches for comparable roles, and increased credential inflation. Macro indicators can lag because employment levels can look stable while churn and opportunity shrink.
Why does “outside options” matter so much?
Outside options determine whether people can exit bad terms. When exit is costly, discipline rises, bargaining weakens, and perceived extraction increases.
Why would this select for coercive politics rather than reform?
Because slow repair is hard and uncertain. Under stress, strategies with immediate, visible effects outcompete reforms whose benefits are delayed, diffuse, or hard to attribute—even when the short-term tactics erode long-run capacity.
What are plausible limits or boundary cases?
Some work remains resistant because it depends on physical presence, high-trust relationships, liability concentration, or unstructured environments where errors are costly and hard to contain. Some sectors may expand through new demand, regulation, or demographic pressure. The claim is not that all work disappears, but that the system can still enter a missing-jobs regime if the most automatable layers include large shares of entry-level and routine cognitive employment, and if the rate of substitution outruns the creation of new, labor-complementary roles.
What does “global synchronization” add?
If shocks arrive together, escape valves weaken simultaneously. Trade, migration, and information become transmission channels for stress rather than relief.
References
[1] U.S. Bureau of Labor Statistics (BLS). “Job Openings and Labor Turnover Summary (JOLTS).” News release page, accessed January 14, 2026.
[2] International Monetary Fund (IMF). “New Skills and AI Are Reshaping the Future of Work.” IMF Blog, January 14, 2026.
[3] Microsoft. “AI at Work Is Here. Now Comes the Hard Part.” Work Trend Index 2024 (WorkLab), May 8, 2024.
[4] Stanford Institute for Human-Centered Artificial Intelligence (HAI). AI Index Report 2025. Stanford University, 2025.
[5] Shapiro, David. “The Silent Disruption: Unpacking AI’s Real Toll on the U.S. Labor Market in 2025.” Substack, 2025.
[6] Hosking, Darin Lawson. “Q4 2025 Silent Layoff Update.” The Moral Algorithm, December 7, 2025.
[7] Magistro, Beatrice; Borwein, Sophie; Alvarez, R. Michael; Bonikowski, Bart; Loewen, Peter John. “The Coming AI Backlash: How the Anger Economy Will Supercharge Populism.” Foreign Affairs, October 13, 2025.
[8] Vlandas, Tim; Halikiopoulou, Daphne. “Welfare State Policies and Far Right Party Support: Moderating ‘Insecurity Effects’ Among Different Social Groups.” West European Politics 45, no. 1 (2022): 24–49. DOI: 10.1080/01402382.2021.1886498.
[9] Bloomberg. “Odd Lots: The Politics of AI Are About to Explode.” Podcast episode, November 19, 2025.
[10] International Energy Agency (IEA). Energy and AI. IEA, April 10, 2025.
[11] Reuters. “Meet the AI chatbots replacing India’s call-center workers.” October 15, 2025.
[12] International Labour Organization (ILO). Generative AI & Jobs: Global Index of Occupational Exposure (refined index). May 20, 2025.
[13] Boulanin, Vincent; Palayer, Jules; Ovink, Charles. Addressing the Risks that Civilian AI Poses to International Peace and Security: The Role of Responsible Innovation. Stockholm International Peace Research Institute (SIPRI), November 2025. DOI: 10.55163/OXHH2798.
[14] Palayer, Jules; Bruun, Laura. “Artificial Intelligence and International Peace and Security.” In SIPRI Yearbook 2025: Armaments, Disarmament and International Security, Chapter 12. Stockholm International Peace Research Institute (SIPRI), 2025.
[15] Kalleberg, Arne L. “Job Insecurity and Well-Being in Rich Democracies.” The Economic and Social Review 49, no. 3 (2018): 241–258.
[16] Standing, Guy. “The Precariat: Today’s Transformative Class?” Great Transition Initiative, October 2018.
[17] World Economic Forum. The Global Risks Report 2024. World Economic Forum, 2024.
[18] Anthropic. “Cowork: Claude Code for the rest of your work.” Claude Blog, January 12, 2026.
The Coming Unrest DOI: https://doi.org/10.5281/zenodo.18252628