NBER study of 5,179 customer-support agents using an AI assistant.
Option 3 · The transition
AI IS COMING.
WHO BENEFITS
IS A POLICY CHOICE.
YESAI takes the best case. NOAI takes the worst. FUTUREAI looks at the middle: AI keeps advancing, creates real productivity and may reduce demand for some kinds of human labour. Alberta can decide how it responds even if it cannot decide what happens to AI globally.
The benefit is already real
Small teams can do work that used to require much larger ones.
That is not hype. Controlled studies already show meaningful productivity gains from generative AI in specific kinds of work. It can make experienced knowledge easier to use, shorten routine work, and let people attempt projects they could not previously afford.
MIT/Science experiment on professional writing tasks; output quality also rose 18%.
Some AI-assisted production is work a small firm or individual could never have afforded to commission conventionally.
A three-person startup doing work that once required fifteen people does not have to fire twelve people. Those twelve jobs may simply never be created. That is economically different from a layoff, but it still changes how much new income flows through wages.
Do not mix three different effects
AI changes labour in three ways.
Enabled production
AI makes a project possible that the person or business could never have afforded before. New output is created. Treating all of that as “lost jobs” would be nonsense.
Avoided labour
A company grows, but does not hire as many people as the same growth historically required. No one is laid off, yet fewer wages are created.
Actual displacement
Existing paid tasks or positions are replaced. This is the familiar layoff story, but it is only one part of the labour transition.
A serious Human Dividend policy has to distinguish these categories. The objective is not to tax every useful AI tool or punish a small company for becoming more capable.
The progression to plan for
The risk grows in stages, not in one overnight apocalypse.
People + AI produce more
Workers finish some tasks faster, small businesses gain capabilities, and new products become affordable to create.
Growth needs fewer new hires
More output is created without proportional payroll growth. Entry-level roles and routine knowledge work can face pressure even before mass layoffs appear.
Existing jobs are removed
If systems become reliable enough across more tasks, companies may cut positions as well as avoid new hiring. The scale and timing remain uncertain.
The wage-distribution system strains
If production keeps rising while human labour income falls sharply, the problem becomes bigger than unemployment: who has the purchasing power to buy what the automated economy produces?
What the evidence says
Large exposure. Unknown final job loss.
The International Labour Organization estimates that one in four workers worldwide are in occupations with some generative-AI exposure. It also says transformation, not replacement, is the most likely outcome for most exposed jobs.
The IMF estimates almost 40% of global employment is exposed to AI, rising to about 60% in advanced economies. Its analysis includes both workers whose productivity may rise and workers whose labour demand or wages may fall.
Neither number is an unemployment forecast. “25% exposed” does not mean 25% unemployed, and “40% exposed” does not mean 40% of jobs disappear.
A useful case study
Musk sees the destination. What pays for it?
Elon Musk has repeatedly said that AI and robotics could make work optional and lead to a “universal high income.” In April 2026 he described government checks as the answer to large-scale AI-driven unemployment and argued that rapidly expanding AI and robotic output could support them.
That recognizes the purchasing-power problem. It does not fully answer the transition problem. If automation reduces wages and employment substantially, governments can also lose part of the labour-income and consumption base they currently tax.
The unanswered question: if AI and robotics create the productivity that makes universal high income necessary, should part of the profits and productive value created by that infrastructure permanently help fund the public system that distributes it?
Why building here can still make sense
Alberta cannot veto global AI. It can decide the terms of hosting it.
Rejecting one Alberta campus can stop that project in Alberta. It does not stop Meta, Microsoft, Google, OpenAI, China, Europe, the United States or everyone else from continuing to develop AI somewhere else. Alberta can still refuse a badly sited or badly structured project. The strategic question is what we gain when a good one is built here.
Capture local value
Construction, skilled trades, municipal tax base, local procurement, grid and generation investment, fibre, training and supplier activity can remain in Alberta when the agreements are real and measurable.
Keep strategic value in Canada
A Canadian Compute Dividend can turn foreign-owned hyperscale infrastructure into usable Canadian capacity instead of making Canada only the landlord and energy supplier.
Attach the costs to the project
Water mitigation, emissions controls, noise limits, grid upgrades, monitoring and eventual cleanup should follow the project instead of becoming someone else’s future bill.
Responsible local infrastructure is not valuable because every data centre is good. It is valuable when Canada captures enough of the economic and strategic upside to justify the land, power, water, emissions and community burden.
Mitigation before crisis
Build the economic safety system while the old one still works.
Responsible infrastructure
Industrial siting, low-water design, emissions mitigation, full-spectrum noise verification, grid cost causation, public monitoring and real reclamation security.
Canadian Compute Dividend
Use a working target around 5% compute-equivalent value for foreign-controlled hyperscale campuses, delivered through access, credits or funding for equivalent Canadian sovereign compute when direct access is impractical.
AI Prosperity Fund
Require a defined contribution from very large AI infrastructure projects into a professionally managed public investment fund. Build the ownership stake while employment, wages and public revenues are still strong, rather than trying to invent one after major displacement occurs.
Measure the right thing
Do not count only layoffs. Track labour share, hiring intensity and the value of human work no longer purchased. Also separate enabled production so the system does not punish genuinely new economic activity.
Keep it invested
Do not treat the contribution as ordinary annual revenue. Invest it as a long-lived public asset so future purchasing power can come partly from capital returns instead of depending entirely on wages and future taxation.
Choose the eventual distribution later
If severe displacement arrives, the accumulated value could support transition income, shorter work weeks, negative income tax, citizen dividends or other systems. We do not need to guess the final mechanism today to start building capacity now.
The FUTUREAI bargain
Do not stop productivity. Make humans shareholders in it.
AI may create enormous new value. The public-policy failure would be allowing the mechanism that creates that value to weaken the wage system while doing nothing to preserve the purchasing power wages used to distribute. A permanent public ownership stake is one practical way to connect future machine productivity back to the people living in that economy.
Build good infrastructure here when it passes the environmental, economic and community tests. Require Canadian strategic benefit. Prepare for labour disruption before it becomes a fiscal emergency.
The goal is not to stop productivity. It is to make sure humans continue sharing in the value productivity creates.
Sources and evidence notes
- Brynjolfsson, Li & Raymond / NBER — Generative AI at Work, workplace study of 5,179 customer-support agents; roughly 14% average productivity gain.
- Noy & Zhang, Science — Experimental evidence on the productivity effects of generative AI, selected professional writing tasks completed about 40% faster with 18% higher evaluated quality.
- International Labour Organization — Generative AI and Jobs: A 2025 Update, one in four workers in occupations with some GenAI exposure; transformation more likely than redundancy for most exposed jobs.
- International Monetary Fund — Gen-AI: Artificial Intelligence and the Future of Work, almost 40% global employment exposure and about 60% in advanced economies, with both complementarity and displacement channels.
- Stanford HAI — 2026 AI Index for the global scale of AI development, adoption, model competition and data-centre infrastructure.
- Forbes — Elon Musk on “Universal HIGH INCOME” as a response to AI-driven unemployment, including his proposed government-check / money-supply mechanism.
- Fortune — Musk on work becoming optional and money eventually becoming less relevant through AI and robotics.
Evidence discipline: the later stages on this page are scenarios to prepare for, not forecasts that a particular unemployment rate or economic collapse will occur.