Harnessing the Ripple Effects: Prospects of Making AI Work for Canadian Jobs
Author
Kyrylo Khutornyi
Editor
Daniel Ebrahimpour
“Harnessing the Ripple Effects: Prospects of Making AI Work for Canadian Jobs” explores the potential impact of AI on Canadian jobs, using the Jevons paradox to illustrate how technological efficiency can lead to unexpected outcomes, such as increased job creation in some sectors despite displacement in others. It advocates for proactive policies to address automation risks, including real-time tracking systems, portable benefits, retraining programs, and fostering collaboration between government and tech companies. By aligning AI innovation with workforce resilience, Canada can mitigate disruptions and harness AI’s transformative potential for job growth and productivity.
160-Year-Old Riddle
The 19th-century economist W. S. Jevons discovered a surprising pattern of how consumption behaves vis-a-vis efficiency amid technological advances. Jevons observed a paradox: technological advancements reducing coal consumption per output unit increased overall coal use by spurring industrial growth, contradicting traditional beliefs.
Amid anxieties of AI-induced work displacement, the Jevons paradox is eerily relevant and as captivating as a make-believe. It would save a lot of human capital (as well as nerves and money for policymakers and CEOs) if AI-driven automation, designed to enhance efficiency and reduce costs, could magically drive greater demand for AI-powered products and services, creating new industries and employment opportunities even as it erodes traditional jobs. If we extrapolate the Jevons paradox curve to AI resistance and employment growth by industry, it would imply that industries with higher employment growth should exhibit less resistance to AI, as efficiency gains typically stimulate complementary labour demand (in an ideal world). In particular, McKinsey and the World Economic Forum believe this pattern will hold in healthcare, education, agriculture, and trade, where AI reduces operational costs and adds new value to labour.
Grounding Ourselves
We won’t rain on anyone’s parade by saying that the 160-year-old paradox serves better as a theory of change framework than as a rock-solid, tangible plan for the future of work. The threats of layoffs are in the air, widely acknowledged publicly. According to a Statistics Canada paper released in September 2024, 60% of all Canadian jobs face automation risk. On the other hand, there’s only one great case of AI-driven layoffs in Canada to date. In July 2024, Intuit announced 1,800 job cuts (incl. 106 in Canada after the closure of Edmonton offices), as part of its strategy to invest in AI-powered tax preparation and financial products.
The development is consistent with the observed data, plotted along the blue curve in the figure below. It shows no clear or significant correlation between AI resistance and employment decline or boost (yet). The AI effects have yet to sink in, and that’s where the Jevons paradox comes in handy as a strategic lens to make sense of the chaotic data. The areas marked 1 through 4, where the Jevons curve intersects the observed real-life curve, highlight unpredictable deviations from an ideal world. It’s probably symptomatic of poor awareness and significant gaps in expectations.
Note. Standard Occupational Classification (SOC) industries as units of observation. AI resistance is a complement of the min-max normalized AI Occupational Exposure Index (AIOE), initially calculated by Statistics Canada. Since the normalized AIOE is scaled from 0 to 1, resistance is calculated as 1 − AIOE. This calculation provides a simple and intuitive way to interpret resistance as the inverse of exposure: higher exposure corresponds to lower resistance, and vice versa. The 7-year compound employment growth rate was obtained via the standard compound annual growth rate (CAGR) formula. The time horizon was set to 7 years, spanning from 2016 to 2023, because Statistics Canada’s AOIE study used the 2016 labour data. The relationship between the AI resistance and the compound employment growth rate was projected via the moving regression (LOESS) technique. The Jevons paradox curve is an elastic demand curve (monotonically decreasing, concave-up function).
Source. Own calculations based on Statistics Canada’s study and data; crowdsourcing.
Therefore, the impacts of AI may be more disruptive than anticipated. If AI-driven efficiency fails to generate the expected job growth and labour adaptability, where’s the disconnect? Are policies lagging? Do industries need to prepare? This uncertainty underscores the urgent need for targeted interventions to better align AI’s potential with workforce resilience. The key point here is one voiced by the recent Noble Prize winner, Daron Acemoglu: “Automation is not our enemy. Excessive automation is our enemy.” Mindful of industry-specific preferences, we must adopt an AI-as-a-complementation mentality, aiming to harmonize these deviations and gradually align areas 1–4 with the smooth curve.
That’s where Jevons paradox meets Conway’s law—another witty epigram from IT asserting that a technology’s design and efficiency mirror its environment’s norms of organization and communication. And public policy steers the wheel here.
Four Steps to Futureproof
Canada can work on four evolving policy measures to address the potential moral panic of layoff risks.
The paramount challenge is to understand who the at-risk people are. Therefore, the first and cheapest measure is to aggregate and make sense of Big Data, engineering a specialized early warning system (EWS) or adjusting the existing ones for AI layoff risks. Inspired by the 2008 Global Financial Crisis, EWSs have become a popular framework for macroeconomic vulnerability detection, and they are now employed across the Canadian institutional landscape (incl. the Bank of Canada and Ontario Securities Commission). Real-time tracking of employment trends, skill demands, and industry shifts would enable policymakers to respond timely with tailored retraining programs and economic adjustments, rather than reacting to crises after they unfold.
Second, cash-plus programs could blend financial support with skill-building opportunities for displaced workers. Partnering with big tech companies, many of which already advocate for Universal Basic Income (UBI) philosophies, Canada can combine relief with pathways for upskilling. Since many of the at-risk industries are what in IT jargon is called a “brownfield environment” (sectors with rigid change management and/or little to no technological integration), chances are such programs will unleash a new generation of “reluctant entrepreneurs”—rank-and-file employees who have been innovating under constraints while adapting to emerging scarcity.
Canada is well-positioned to expand this approach, with C$50 million already dedicated to the Sectoral Workforce Solutions Program to address AI-driven layoffs. However, the current fiscal climate has yet to improve. Despite fiscal consolidation, there’s a pressing need for a more focused budget, especially concerning the country’s productivity decline. Without solutions to this challenge, any proactive labour policies may lead to no avail.
Third, portable benefits and softening labour taxes are essential to respond to the rise of new gig and contract work that might expand due to AI-driven labour flexibility. These actions were advocated in a recent IMF paper, where researchers, simulating an automation shock, found out that active labour market policies and unemployment insurance can significantly alleviate the initial surge. The good news is that there’s an evolving jurisdictional legacy. Portable benefits have become an agenda item in Ontario’s official public policy discourse, resulting in a specialized advisory panel. According to IMF experts, preventing the implementation of an “AI tax” is also vital, as it puts pressure on innovation. Such a tax would also crowd out the AI market and make it even less competitive by damaging the early-stage AI firms, contributing roughly 10-20% to the market now. If living standards fall further because of productivity decline, it’s way more efficient to reduce labour taxes to slump the costs of employment for enterprises.
The issue with portable benefits is dissatisfaction with forgone benefits among at-risk workers. Analyzing the case of portable benefits and Uber in Canada, labour lawyers pointed out that cost savings for the company are massive. Uber’s “Flexible Work+” proposal amounted to C$40 million, significantly less than the C$80 million figure for a traditional pension and insurance scheme. The financing gap, with the company’s push for labour law reforms to roll out the program, raised concerns about corporate interference in the integrity of employment rights. In this case, the government needs to be on board with gig economy activists and experts to ensure a platform with fair and equitable benefit distribution.
Fourth, negotiations on AI whistleblowers and risk ownership with big tech should start now. Mutual understanding of AI whistleblowers between the government and suppliers is essential to contain bias in algorithms, unintentional surveillance, and excessive automation. To balance workplace safety with R&D autonomy, one possible scenario would be a data exchange protocol on critical developments. In this case, AI firms should be invited to participate in government-led AI safety research. The most prominent example is from the neighbouring U.S., where OpenAI and Anthropic partnered with the U.S. National Institute of Standards and Technology—the brain behind the federal tech regulations. At the same time, the OpenAI whistleblowers’ letter signifies an ever-growing need to outline risk ownership in these very regulations.
On the public-private cooperation note, policy analysts suggested paying attention to the European Digital Markets Act (EDMA) to mitigate the fusion between corporate and political power. In particular, the following components are of interest for Canada: interoperability, anti-self-preferencing rules, transparent lobbying, and data-sharing mandates (with penalties of up to 10% of revenue).
Combined, these four measures will address the mismatch between AI’s promise and its current labour market impact, changing the cross-sectoral imbalances in expectations and layoffs. We will be able to enact timely risk management so there’s no need for crisis management later.