For most of human history, work was visible.
A farmer plowed a field. A blacksmith shaped iron. A dockworker lifted cargo. A seamstress stitched fabric. Even early industrial labour, however brutal, retained a certain material clarity. Work involved observable physical effort transforming the physical world.
Then something changed.
Over the course of the twentieth century, advanced economies slowly detached the idea of work from visible material production. Increasingly, work became administrative, symbolic, managerial, communicative, and digital. Millions of people now spend their working lives moving information rather than objects.
And now artificial intelligence may be destabilizing the very definition of work itself.
What is work?
Much of the current discussion around AI focuses on job loss. Will accountants disappear? Lawyers? Graphic designers? Coders? Teachers? Journalists? The anxiety is understandable, but it may also be too narrow. The deeper transformation is not simply that AI threatens particular occupations. It is that AI challenges long-standing assumptions about what society considers legitimate human labour in the first place.
Modern societies often speak about work as though its meaning were timeless and self-evident. It is not.
Aristocratic societies historically viewed much manual labour with contempt. In many premodern cultures, true status came precisely from not needing to work physically. Land ownership, inheritance, military command, clerical authority, or political office conferred prestige. Labour belonged lower in the hierarchy.
Industrial capitalism transformed this moral landscape. Work became not merely economically necessary but morally valorized. Productivity, discipline, punctuality, and measurable output became defining virtues of citizenship itself. The factory clock helped shape the modern understanding of responsibility and worth.
But even industrial capitalism contained a hidden instability.
As economies became wealthier, more work gradually shifted away from direct physical production toward administration and abstraction. By the late twentieth century, vast numbers of people were earning middle-class incomes through activities that would have appeared bewildering to earlier generations: branding, consulting, coordination, compliance, strategic communications, content moderation, user experience design, stakeholder engagement, data analysis, and countless other forms of symbolic labour.
Much of modern work increasingly involved manipulating language, images, systems, and relationships rather than material goods.
The digital economy accelerated this further. Entire careers emerged around attention itself. Influencers, streamers, podcasters, social media managers, and content creators monetize visibility, personality, and audience engagement. Millions now perform labour that is simultaneously real and strangely intangible.
And now AI enters directly into this symbolic terrain.
Work, AI and modern identity
Unlike earlier technological revolutions, AI does not primarily threaten repetitive physical labour. Industrial machinery mechanized muscle. AI increasingly mechanizes cognition, synthesis, communication, and pattern recognition — precisely the forms of labour modern professional economies expanded most aggressively.
This creates a peculiar cultural shock.
Societies had already adjusted, often uneasily, to the idea that sitting at a keyboard constituted “real work.” AI now forces a second adjustment: what happens when many forms of symbolic work can be partially or substantially reproduced by machines?
The question is not merely economic. It is existential.
Modern identity is deeply entangled with employment. People are routinely asked, “What do you do?” meaning what work defines your social role, competence, and legitimacy. Entire welfare systems, tax systems, pension systems, educational systems, and moral narratives are organized around paid employment as the central organizing principle of adulthood.
AI destabilizes this arrangement because it weakens the historic connection between labour time and value creation.
For centuries, even when technologies displaced particular workers, human labour remained visibly central to production. A machine might increase output, but people still operated the machine, supervised the process, repaired the equipment, coordinated logistics, and performed adjacent labour.
AI differs because it increasingly intrudes into tasks once assumed to require uniquely human intellectual mediation: drafting text, summarizing research, generating code, producing images, translating languages, and synthesizing information. Even elements of emotional communication can now be simulated at scale.
The result is not simply displacement. It is conceptual confusion.
When AI assists a lawyer, who produced the work? When AI drafts a report edited by a consultant, what exactly constitutes the human contribution? When an influencer uses AI to generate scripts, thumbnails, captions, and marketing copy, where does labour begin and end?
The boundaries become blurred.
Ironically, modern economies may have prepared the ground for this confusion themselves. Over decades, many workplaces increasingly measured output through presentations, emails, dashboards, strategic plans, reports, and metrics detached from direct material production. Entire layers of organizational life evolved around symbolic performance, managerial abstraction, and administrative signalling.
AI thrives in precisely such environments because language-heavy systems are often easier to replicate computationally than embodied human activities rooted in physical complexity and unpredictable environments.
The plumber may ultimately prove harder to replace than the junior analyst.
Status disruption
This inversion carries profound implications for class and status assumptions. For decades, advanced societies implicitly encouraged people to move “up” from physical labour toward knowledge work. AI complicates that hierarchy. Certain forms of manual, relational, and embodied labour may prove more economically resilient than many white-collar roles once considered secure.
At the same time, AI exposes another uncomfortable reality: much modern work already contained significant performative elements before AI arrived.
Many employees recognize that parts of organizational life involve the production of appearances as much as tangible outcomes. Meetings signal coordination. Reports signal diligence. Strategic frameworks signal competence. Corporate language often functions ritualistically rather than operationally.
AI does not invent performative work. It reveals it.
And this may explain some of the unusual emotional intensity surrounding AI debates. People are not simply afraid of losing income. They are confronting the possibility that activities long treated as markers of expertise and status may be more replicable than previously believed.
How will this impact social assistance systems?
All of this has enormous implications for social assistance systems built around assumptions about work that emerged in the industrial era.
Programs like Ontario Works rest on a foundational premise: if people are poor, there is usually work available somewhere and the role of the welfare system is to push, incent, monitor, or discipline recipients into labour market participation. Eligibility is therefore linked not simply to need, but to behavioural expectations around employability, job search, training, compliance, and labour force attachment.
But what happens if AI gradually destabilizes the labour market itself?
No one fully knows.
It is entirely possible that Canada simply produces more low-paid service work — what many call “McJobs” — to absorb displaced labour. Historically, technological revolutions have often created new forms of employment even as they destroyed older ones. The economy may continue generating large numbers of precarious service positions in food delivery, warehousing, personal support, retail assistance, hospitality, elder care, security, logistics, and various forms of contingent human labour that remain difficult to automate fully.
If that occurs, governments may simply preserve the existing welfare architecture. Ontario Works recipients may still be required to demonstrate active job search, attend employability programs, complete mandatory training modules, or accept low-wage precarious employment regardless of how economically marginal or unstable such work becomes.
In fact, one can easily imagine governments intensifying work requirements precisely because AI increases labour insecurity. Political systems often react to economic anxiety by doubling down on older moral narratives around work ethic, deservingness, and dependency.
But another possibility exists.
Revealing the fiction of welfare systems
AI may slowly expose a contradiction already embedded within welfare systems themselves.
Workfare models assume that insufficient employment primarily reflects deficiencies in individual attachment to work. Yet if labour markets become increasingly unstable, fragmented, casualized, algorithmically managed, or partially displaced by AI systems, it becomes harder to maintain the fiction that all insecurity stems primarily from personal failure or insufficient effort.
At some point, governments may confront uncomfortable questions.
What exactly are welfare recipients being prepared for?
What constitutes meaningful labour market participation in an economy increasingly mediated by algorithms and automation? If AI systems can perform large portions of entry-level administrative work, junior coding, customer support, content production, or information synthesis, what becomes of traditional pathways into the labour market itself?
Historically, many middle-class careers relied upon apprenticeships within routine cognitive work. Junior lawyers reviewed documents. Junior analysts prepared summaries. Junior reporters assembled background information. Junior office workers learned organizational systems through repetitive administrative tasks.
AI increasingly threatens these very entry ramps.
This creates the possibility of a bifurcated labour market. A relatively secure upper tier of highly skilled workers may use AI as productivity amplification while a growing lower tier cycles through unstable service labour, gig work, surveillance-heavy employment, intermittent contracts, and low-autonomy care or logistics work.
If that future emerges, welfare systems may become even more punitive because governments will fear large populations permanently detached from stable labour markets. One can imagine increasingly elaborate systems of monitoring, behavioural conditionality, digital compliance, mandatory retraining, algorithmic employability scoring, or endless cycles of activation programming designed less to produce stable employment than to preserve the moral legitimacy of the work-based social order itself.
In other words, work requirements may persist not because they work particularly well, but because governments remain psychologically and politically dependent upon the idea that paid labour must remain the primary condition for social legitimacy.
But if there’s no work…?
This is where AI may ultimately force a deeper reckoning.
Industrial societies organized welfare systems around labour scarcity. The assumption was that work fundamentally existed and that the central challenge was motivating, training, or disciplining people into participation.
But AI raises the possibility — still uncertain, but increasingly plausible — that societies may eventually confront not labour scarcity but labour redundancy in certain sectors.
If that happens, social assistance systems built primarily around work enforcement may begin to look increasingly archaic.
Governments would then face profoundly difficult choices.
Do they maintain low benefits and punitive work requirements even as stable employment erodes? Do they create endless forms of simulated employability activity to preserve older moral narratives around work? Do they expand public employment artificially? Do they move toward guaranteed income models? Do they redefine contribution itself to include caregiving, volunteering, community participation, or forms of social usefulness not fully captured by traditional labour markets?
We do not know.
But it is increasingly difficult to imagine that welfare systems designed around twentieth-century assumptions about work will remain unchanged in an AI-shaped economy.
Re-evaluating the undervalued
And yet there is another irony.
The forms of labour likely to remain most indispensable may be precisely those long undervalued within industrial capitalism itself: caregiving, companionship, emotional intelligence, maintenance, physical dexterity, trust-building, teaching, mentoring, and human presence.
A nurse aide calming a frightened dementia patient performs something profoundly difficult to automate. A child-care worker managing emotional unpredictability performs labour far more complex than many spreadsheets can capture. A tradesperson improvising solutions inside an aging building encounters physical reality in ways algorithms still struggle to navigate.
The future may therefore contain a strange reversal. Societies that long privileged abstract symbolic work over embodied human labour may rediscover the enduring economic and social importance of deeply human forms of work.
But whether those forms of labour will finally be properly valued — economically or morally — remains an open question.
For centuries, societies organized themselves around the assumption that human labour was the indispensable engine of economic value. AI does not abolish that assumption entirely. But it weakens it enough to force a profound reconsideration of what work is, what work means, and why societies value certain forms of human activity over others.
That reconsideration may ultimately matter more than the technology itself.