ANALYSIS | THE NEXT 20 YEARS

This feature examines possibilities for 2036–2046, not a prediction that any one future will occur. Research findings are identified separately from the scenarios and policy analysis.

If artificial intelligence can do more of the work, how will people earn a living—and who will own the wealth it creates?

That question could shape the next twenty years as profoundly as any technical breakthrough. A society able to produce more with less human labour might offer shorter working weeks, better services and greater freedom. It might also leave millions competing for fewer secure jobs while profits accumulate elsewhere.

The difference would reach far beyond the workplace. It would affect whether families can afford housing, how children learn, who receives medical attention and how much control people retain over decisions about their lives.

Between 2036 and 2046, AI could change how Canadians work, learn, receive care, form relationships and deal with government. But better technology does not automatically produce a better standard of living. Ownership, public policy, bargaining power and access will help determine whether greater productive capacity becomes greater freedom—or greater insecurity.

What we know—and what we cannot yet know

There is no established basis for declaring that AI will eliminate every job within ten or twenty years. There is also no basis for assuming that every displaced worker will move smoothly into something better.

The International Labour Organization’s 2025 assessment found that one in four workers worldwide was in an occupation with some exposure to generative AI. It judged transformation more likely than complete replacement for most jobs, given the continuing need for human input. That is an assessment of potential exposure, not a timetable for layoffs.

A Statistics Canada study released September 3, 2024, using 2016 and 2021 census data, estimated that about 60 per cent of Canadian employees could face AI-related job transformation. Roughly half of that exposed group held jobs where AI could strongly complement their work. The researchers explicitly cautioned that exposure does not mean job loss.

These studies examine different technologies and populations; their percentages are not interchangeable. Neither describes the Canadian labour market in 2046. They provide a starting point for thinking about change, with uncertainty about technical progress, adoption and the economy’s response.

Software that performs impressively in a demonstration still has to work reliably inside a business. Cost, accountability, customer preferences, security, infrastructure and organizational change all affect adoption. Automating physical work also requires suitable machines and safe operation in unpredictable environments. Progress in text generation does not establish a matching schedule for replacing electricians, caregivers or transit operators.

A job can survive while opportunities disappear

Work consists of tasks. AI might draft correspondence, summarize records or prepare a first design while a person checks accuracy, handles exceptions and takes responsibility. An employer might use that assistance to serve more customers with the existing team. Another might reduce staffing.

Consider a hypothetical law office. Lawyers and assistants still work there in 2040, but software handles much of the document preparation once assigned to junior employees. The office has not disappeared. The number of entry-level opportunities may nevertheless have fallen.

That creates an apprenticeship problem: if machines do the introductory work, how do newcomers develop the judgment required for senior responsibilities? Employers and educators may need deliberate training placements, supervised practice and new ways to assess competence.

New demand could also create work. More affordable professional assistance might allow small businesses and households to buy services they previously went without. Whether that expansion outweighs displacement—and whether the new jobs are accessible to the same people—is unresolved. National employment totals alone would not reveal every community’s losses.

If machines produce the goods, who has money to buy them?

Your employer’s productivity and your household’s purchasing power are connected, but they are not the same thing. A business can produce more per worker while the benefits go mainly to customers through lower prices, to owners through profits, or to workers through higher pay.

If automation substantially reduces labour income across the economy, households would need other sources of purchasing power, lower living costs, or both. Remaining wages, business ownership, investment returns, government transfers and publicly provided services are possible channels.

This does not mean an economy instantly collapses whenever employment falls. Investment, government spending, exports and spending by owners also support demand. But a society with abundant production and widespread inability to afford essentials would face a serious distribution problem.

Nor would cheap digital services automatically mean cheap living. AI might lower the cost of preparing architectural plans without creating serviced land, construction materials or an available building crew. A near-free online tutor would not pay the rent.

The useful question is therefore broader than how much AI adds to economic output: how much of that gain reaches ordinary households, in forms they can actually use?

Four futures that could coexist

1. Greater prosperity and more time

In the most attractive scenario, productivity rises enough to support good incomes with fewer working hours. Some employers adopt shorter weeks. Public services improve. People spend more time caring for relatives, building businesses, making art or participating in their communities.

This future would require more than capable software. Competition could help pass savings to customers; worker bargaining and shared ownership could spread gains; public investment could improve essential services. Reducing hours without reducing pay would depend on productivity and how its benefits are divided. It would not be equally affordable in every occupation.

2. A prolonged and uneven adjustment

A second possibility is repeated disruption. People change careers more often, take courses between jobs and accept temporary reductions in earnings. Some prosper; others never regain their previous income.

Retraining can help, but it cannot guarantee that an employer will have a suitable vacancy nearby. A mid-career worker with children, health limitations or a mortgage faces different choices from a recent graduate. Support would need to address income, time and geography as well as tuition.

3. Highly productive businesses, insecure households

AI could also reinforce concentrated ownership. Companies controlling useful models, computing capacity, distribution or customer relationships might capture substantial profits. Workers could face weaker bargaining power even as the economy produces more.

In that scenario, access to the best tools might become another dividing line: a well-funded organization buys effective assistance while a small firm or household makes do with less capable products. This is a possibility, not an inevitable market structure. Competition, open technologies and public alternatives could change it.

4. A larger public role

If paid employment becomes less dependable, governments could expand income guarantees, publicly provided services or collective ownership. Citizen investment funds might distribute returns from diversified assets. Worker cooperatives and pension funds could also broaden ownership without governments directly operating every business.

These arrangements would involve choices about funding, investment risk, eligibility and accountability. Public ownership does not automatically ensure good management or equal access. The four scenarios could appear in different combinations within one country—or even within one city.

Would this make every government socialist?

Large-scale displacement could strengthen the political case for redistribution. It would not mechanically determine an economic system.

A capitalist economy can retain private businesses and markets while taxing income and profits to fund benefits. Socialism generally places greater emphasis on social ownership or control of productive assets, although its traditions differ. Communism, in its theoretical sense, envisions common ownership and a classless society; historical governments using the label have taken particular political forms.

A basic-income payment, by itself, would not turn Canada into a communist country. The more revealing questions would be who owns productive assets, who receives their returns, and how decisions are made.

Political freedom is a separate dimension. Public ownership can be subject to democratic oversight or authoritarian control. Private ownership can exist alongside strong civil liberties or extensive surveillance. AI supplies capabilities; institutions determine how power can be exercised and challenged.

How would people receive an income?

A universal basic income is one option: a regular payment without a work requirement. Its practical effect would depend on the amount, the taxes financing it, and which existing benefits it supplements or replaces. A small payment cannot provide security merely because it is universal.

Other options include targeted income guarantees, wage insurance for workers accepting lower-paid employment, stronger unemployment protection and portable benefits that follow workers between jobs. Public provision of housing, health services, transport or education can reduce the income a household needs.

Ownership offers a different route. If workers and citizens hold income-producing assets through pension funds, employee shares, cooperatives or public funds, automation earnings could reach them as investment returns. But dividends fluctuate, and an ownership scheme needs genuine assets and accountable management.

In its June 2024 fiscal-policy discussion, the International Monetary Fund argued for adapting social protection and taxation to spread AI’s gains. That does not establish one universal solution. Governments would still have to balance support, investment incentives and sustainable revenue.

Money transfers also do not create scarce goods by themselves. If housing supply remains constrained, additional purchasing power may partly become higher rents. Income policy and the supply of essentials would need to be considered together.

Health care: a better assistant is not an entire system

Imagine arriving at a medical appointment with an organized history, translated instructions and a clear list of questions prepared with AI. A clinician might spend less time on documentation and more time examining and listening to patients.

Those are plausible applications, not a promise that every diagnosis will be correct or every waiting list will disappear. The World Health Organization’s January 2024 guidance identified potential uses in clinical care, administration, education and research while warning about inaccurate outputs, bias, affordability and cybersecurity.

By 2046, a crucial distinction could be whether technology expands access to qualified professionals or becomes the cheaper substitute offered mainly to people with fewer resources. Patients would need understandable explanations and a practical route to human review when something seems wrong.

Education: having an answer is different from understanding it

A responsive tutor could explain a concept several ways, translate material and provide practice at a student’s pace. Adults changing careers might study around work and family commitments.

But education also develops judgment, persistence and the ability to recognize a faulty answer. If a student delegates every difficult step, a polished assignment may conceal weak understanding. Schools could place more emphasis on discussion, supervised demonstrations and explaining how a conclusion was reached.

UNESCO’s guidance on generative AI in education, first published in 2023, emphasizes human agency, privacy and appropriate educational design. The future question is not simply whether every child has a chatbot. It is whether that assistance produces learning, protects the child and remains accessible regardless of family income.

Housing, transport and the physical world

In a possible 2040s city, AI improves maintenance scheduling, building design and the coordination of deliveries. More capable robotics might reduce some physical production costs. Automated transport could improve mobility where it becomes safe and affordable.

These possibilities have different technical and economic requirements. A vehicle operating successfully on one route does not prove universal readiness. Housing still requires land, infrastructure, materials and construction capacity. Software cannot simply remove every physical or political constraint.

Computing itself has a physical footprint. The International Energy Agency’s 2025 Energy and AI report examines electricity supply as both a requirement for AI and a constraint on its expansion. Future efficiency gains matter, but so do grids and the location of infrastructure.

A realistic article about digital abundance must therefore leave room for physical scarcity. Life could become easier in some respects while remaining expensive in others.

Relationships, creativity and the value of human attention

AI companions could offer conversation, reminders and help navigating daily life. For some people, especially those facing isolation or communication barriers, that assistance could be valuable. Whether it strengthens human relationships or displaces them would depend on its use and design.

A companion sold by a company also raises an ordinary commercial question: whose interests does it serve? A system that knows someone’s worries and preferences could be helpful, but it could also become an unusually persuasive sales channel. Users would need control over memory, data and the ability to leave.

Creative tools could let a small Canadian team produce work that once required a large budget. At the same time, a much greater supply of songs, images and videos could make earning attention and income harder. Easier production does not guarantee a viable career.

Human performance, local reporting, live events and trusted relationships might become more valuable in some markets. That is a possible response, not a guarantee that audiences will pay enough to sustain every displaced creator.

Government that helps—or a system nobody can appeal

By 2046, a useful public-service assistant might explain a form, identify available support and help a person navigate several departments. Reducing administrative effort could improve access.

The same technology could make an unfair decision faster and at greater scale. A benefit, hiring or housing decision should not become impossible to challenge because its explanation is hidden behind an automated process.

The practical questions are concrete: can a person see what information was used, correct an error, reach an accountable official and obtain meaningful review? How much personal information must they surrender to receive assistance?

Efficiency would be only one measure of success. A system that processes applications quickly while leaving vulnerable people unable to resolve mistakes would fail another essential test.

What this could mean for Hamilton and Canada

For a city such as Hamilton, the future would not arrive as one uniform event. Consider a hypothetical household with one person doing office administration and another doing hands-on maintenance. Their employers might adopt different technologies at different speeds, with very different consequences for training, hours and pay.

A small business could gain affordable design, bookkeeping assistance and translation while also facing competitors that use the same tools. A municipal department might reduce repetitive paperwork while needing employees who can investigate unusual cases and explain decisions to residents.

Canada’s choices would include whether domestic firms and public institutions can access effective tools, whether workers share productivity gains, and whether communities can adapt without placing the entire cost on displaced households.

This is a scenario-based Canadian lens, not a report of announced Hamilton layoffs or a verified municipal automation plan. Specific local outcomes would require separate reporting.

What to watch between now and 2046

Useful evidence will come from changes in ordinary lives. Are entry-level openings declining? Are workers earning more for each hour, or simply being asked to supervise more output? Are shorter weeks voluntary and financially viable? Do lower business costs reach customers?

Other signals matter too: household income after housing costs, access to human professionals, concentration of ownership, the effectiveness of retraining, and whether people can challenge automated decisions. Job totals should be read alongside wages, hours and security.

By 2036, we may have much clearer evidence about which tasks businesses can automate reliably and how workers adapt. By 2046, the consequences could be deeper—or adoption could remain slower and more uneven than today’s most confident forecasts suggest. Neither date is a deadline supplied by existing research.

The defining question is whether greater technological capability translates into greater human choice: time to care for family, an affordable home, useful work, trustworthy services and a meaningful voice in decisions.

AI may expand what society can produce. Who benefits from that expansion remains a decision for people, businesses and governments.

About this analysis: Published October 19, 2026. This feature draws on the dated primary research and guidance linked above. Future services and policy arrangements are illustrative scenarios, not measured outcomes, expert consensus forecasts or commitments by Canadian governments. No original interviews are represented.

Lead image: editorial illustration for The Northern Current depicting an imagined Canadian future. It is not a documentary photograph or evidence of a real event.