Work & Economy Futures

— where work and livelihoods are heading

This page reports what serious research actually says about jobs, automation, and income — including several widely repeated numbers that turned out to be wrong, or at least much less settled than they sounded.

What’s actually happening

Automation forecasting has a documented history of disagreement, not convergence. A 2013 Oxford study found 47% of U.S. jobs were “at risk” of computerization. A 2016 OECD study redid the same question task-by-task instead of job-by-job, across 21 countries, and found only 9% of jobs at high risk on average. A third method, in the OECD’s own 2023 Employment Outlook, found 27% of OECD jobs at high risk from AI specifically. Three credible teams, three very different numbers, and a decade of real employment data hasn’t settled the argument: a 2022 retrospective found the U.S. economy added 16 million jobs since the 2013 study, and the correlation between that study’s original risk scores and what actually happened to employment in each occupation was weak (−0.26).

The most current, narrow evidence on AI and jobs specifically: a Stanford study (published November 13, 2025) found workers aged 22–25 in the most AI-exposed occupations saw a 16% relative decline in employment since generative AI spread — showing up in reduced hiring, not in pay cuts. Anthropic’s own research (published March 5, 2026) found no detectable rise in unemployment economy-wide for AI-exposed occupations, but did separately find a roughly 14% decline in job-finding rates for the same 22–25 age group in exposed fields — a real, narrow, replicated early signal, not evidence of broad job loss. Other researchers using different measures find the opposite direction for the wider labour market, and the field’s own commentators note that commercial AI use is still too recent for any lasting economic effect to be reliably separated from noise.

Gig and platform work is consolidating, not exploding: the International Labour Organization’s own February 2026 research brief counted 653 active digital labour platforms globally as of October 2025, down from 777 in 2021. On June 12, 2026, the International Labour Conference adopted the first binding global labour treaty covering gig work (406 votes in favour, 8 against, 36 abstentions), covering pay, safety, and algorithm transparency.

Where credible people think it’s heading

MIT economist David Autor argues plainly that “the industrialized world is awash in jobs, and it’s going to stay that way,” pointing to demographic labour shortages and AI’s comparative advantage at judgment-heavy tasks. His own research found that most current U.S. employment sits in job types created after 1940 — direct evidence that job-creating technology has historically outpaced job-destroying technology over the long run. But the same research is two-sided: it also found the job-destroying effect of automation has intensified over the past four decades while the job-creating effect of new technology hasn’t kept pace.

Canada’s own official labour-market projection points toward a shortage story, not a displacement story — the opposite kind of claim from the automation-risk debate above. The federal government’s Canadian Occupational Projection System (published January 30, 2025) forecasts 8.1 million total job openings from 2024 to 2033, with 5.5 million of those driven by retirements — more than double the 2.6 million driven by economic growth. More than 100 occupations face a moderate-or-strong shortage risk, against only 17 projected toward surplus.

Basic income has been tested several times, with genuinely mixed, not one-sided, results. Finland’s national pilot (2017–2018, 2,000 people, €560/month unconditional) found essentially no employment effect but real gains in wellbeing (published May 2020). Stockton, California’s SEED pilot (2019–2021, 125 people, $500/month) showed a strong positive employment signal in its first-year results — but contemporaneous coverage of its final report reported the employment effect as not statistically significant over the full two-year window, while health and wellbeing gains held — that finding is CalMatters’ April-2023 characterization of the study, carried here as coverage: the journal article itself (Journal of Urban Health, April 2023) is a health-and-agency study, not an employment re-estimation, and the research team’s own release emphasized “no negative employment impacts.” A separate, larger study funded substantially by OpenAI’s Sam Altman (2020–2023, 1,000 people, $1,000/month) found the opposite direction on employment: labour supply actually fell, by 1.3 fewer hours worked per week on average. Ontario’s own basic-income pilot (launched April 2017, roughly 4,000 people) was cancelled by an incoming government after about a year, in July 2018, before its official evaluation could run — so no official verdict exists; only smaller, independent, after-the-fact surveys do, and they report largely positive self-assessed outcomes.

What’s honestly uncertain

Whether the AI-linked hiring slowdown for young workers is an early warning sign for the wider economy, or a narrow effect specific to a few exposed occupations, is genuinely unresolved — the researchers making the narrow claim say so themselves. A separate debate over whether economic growth itself should be pursued at all continues between “degrowth” advocates (led by anthropologist Jason Hickel) and mainstream economists like former World Bank economist Branko Milanović, a real, sustained argument that has run for close to a decade inside serious economics, not a strawman fight.

Why it matters for Toronto

The Toronto Region Board of Trade’s most recent dedicated workforce report (published March 8, 2022) found that 75% of internationally educated immigrants in the region were not working in the field they trained for. On the city’s current numbers: Statistics Canada’s Labour Force Survey for June 2026 describes the Toronto census-area unemployment rate as “little changed at 6.9%,” with Ontario holding near 7.0% — a gradual easing that continued into the summer of 2026. Wellbeing Economy Governments, an international group some cities point to as a model for measuring success beyond GDP, is a coalition of exactly five national governments (Scotland, Iceland, New Zealand, Wales, and Finland) as of a 2026-08-05 correction to this file — Canada participates only as an observer, not as a member.