Senior fellow at MLI says Ottawa's "AI for All" strategy has right intentions but must be paired with red-tape reduction, measurable results and occupation-specific worker training
Canadian workers are using generative AI on the job at less than half the rate of their American counterparts, and the federal government's new AI strategy will not close that gap on its own, according to a commentary published Monday by the Macdonald-Laurier Institute (MLI).
Writing for MLI's Inside Policy, senior fellow Jon Hartley and research assistant Maddux Shipton report that, in their survey, 21.2 per cent of Canadian workers used generative AI at work as of mid-2025, compared with 45.6 per cent in the United States.
"Canada faces a generative AI wake-up call: adoption lags our peers, while the economy is increasingly defined by sluggish growth, underutilized talent and low productivity," they wrote.
The authors challenge Prime Minister Mark Carney's claim that Canada is among four countries "at the forefront of AI." Canada has outstanding researchers, they acknowledge, but leadership in research is not the same as broad use of AI across the economy.
Subsidies alone won't change how firms work
Hartley and Shipton describe the federal "AI for All" strategy as well-intentioned but only one step toward solving the problem. Funding to help small and medium-sized businesses adopt AI cannot remove regulatory barriers, they argue, and should be paired with a red-tape reduction strategy.
"A subsidy may prompt a firm to try a tool; lasting gains require businesses to redesign workflows, train workers, and compete for customers," they wrote. They call on Ottawa to publish sector-level adoption and productivity data and to end support that simply pays for purchases firms would have made anyway.
Citing a KPMG–University of Melbourne study, the authors report that Canada ranked 44th of 47 countries on AI training and literacy and 42nd of 47 on trust in AI systems. Fewer than a quarter of Canadians reported having received AI training, according to the study as described in the commentary.
The authors credit the strategy's national AI literacy initiative but recommend building on it with occupation-specific instruction delivered through colleges and employers. That training should teach workers when to verify an AI answer, how to protect confidential information and which tasks require human judgment, they wrote. They also suggest that participating employers report actual workflow changes, and that firms using AI in consequential decisions explain when it is involved and give people a way to challenge errors.
Building, and earning trust
On infrastructure, the authors argue that regulatory friction, slower permitting and higher building costs have made Canada less attractive to industry than the United States. They point to Meta's planned first Canadian data centre in Alberta, which they say represents more than $13 billion in proposed investment, and recommend clear approval criteria, disclosure of effects on electricity bills and water use, and requiring developers to pay the infrastructure costs their projects create.
They also fault AI firms' own messaging, saying founders too often present the technology as either a historic triumph or a threat to humanity. "Trust has to be earned through experience, transparency and accountability, rather than grand claims," they wrote.
"The federal strategy correctly identifies much of the challenge," the authors concluded. "The test now is whether governments across Canada will remove barriers, address legitimate local costs and measure the gains."
Hartley is a senior fellow at MLI, an assistant professor of economics at the University of Texas at Austin School of Civic Leadership and a policy fellow at the Hoover Institution. Shipton is his research assistant and a fourth-year student at the University of Alberta.
