A Canadian worker now produces roughly seventy cents of output per hour for every dollar produced by an American worker.
By Romel Dhalla
Nearly every discussion about artificial intelligence in business ends up in the same place. Which jobs will disappear? Which professions are safe? How many people will be replaced? Those questions make good headlines. They are also the wrong questions. The right ones are about productivity, and about who owns the technology that delivers it.
Canada needs to get both right, because our starting position is weak. The Bank of Canada has called this country’s productivity performance an emergency, and the numbers support the language. On the Bank’s own figures, a Canadian worker now produces roughly seventy cents of output per hour for every dollar produced by an American worker. The OECD projects that Canada will record the lowest growth in real GDP per person of any advanced economy over the coming decades. A country in our position cannot afford to get artificial intelligence wrong.
Businesses don’t exist to employ the greatest number of people possible. They exist to produce the greatest value possible with the resources they have. Every productivity leap in history worked the same way: it let people accomplish more in the same time. Artificial intelligence belongs on that list, and unlike most claims made for it, this one has been measured. In a randomized trial run by Harvard researchers inside Boston Consulting Group, consultants using AI completed 12.2 per cent more tasks, finished 25.1 per cent faster and produced work rated 40 per cent higher in quality. The weakest performers gained the most.
None of that is automatic. The same study found consultants were far more likely to get things wrong when they pushed AI past its competence. The gains go to professionals who learn the tool, challenge it and verify its work. Most professionals will be forced to adopt AI, because their competitors already have.
The lawyer who reviews more case law in an afternoon, the accountant who delivers faster at lower cost, the advisor who arrives at every meeting with deeper research: each one forces the rest of the market to respond. Clients don’t pay for slower. Firms will still need experienced people to exercise judgment and stand behind the work, so the labour market will adjust more slowly than the headlines suggest. But as each professional becomes more productive, firms grow without hiring at the same pace.
The consequences land first at the bottom of the ladder. Every profession has relied on junior employees to research, prepare reports and review documents while learning the business. Stanford researchers, working from American payroll records, have already documented a sixteen per cent relative decline in employment among early-career workers in the occupations most exposed to AI, while employment among experienced workers in the same fields held steady.
If AI does the apprenticeship work, firms will recruit fewer graduates, and every profession will face the same question: if nobody does the junior work, where do the senior professionals of 2040 come from? Firms that treat AI purely as a reason to stop hiring will discover, fifteen years from now, that they automated away their own succession plan. The winners will use the gains to train juniors faster on harder problems, not to eliminate them.
Government should be paying the closest attention of all, because no employer in Canada processes more information than government. Applications, permits, licences, procurement, benefits and internal administration consume enormous amounts of time and money.
Treasury Board figures show the federal public service grew from 257,000 in 2015 to a peak of nearly 368,000 in 2024, and even after two years of decline it remains about a third larger than it was a decade ago. Yet Canadians still wait weeks or months for services that involve little more than collecting information, verifying it and making a routine decision. The government’s own figures put the immigration backlog above two million applications at its peak. More people did not fix the problem. Better processes will, and AI can now support exactly those processes.
Public sector unions understand where this is heading, and their responsibility is to protect their members. Governments have a different responsibility: they must protect taxpayers by delivering better services as efficiently as possible. Those objectives will collide. The collision is manageable. The public service loses thousands of employees to retirement every year, so a government serious about modernization can hold service levels while shrinking through attrition and redeployment, replacing processes instead of people.
Ottawa has already accepted the principle: Budget 2025 targets a public service of roughly 330,000 by 2028-29, some 40,000 below the 2024 peak, with attrition named as a primary driver. What it cannot do is protect outdated administration and rising headcount at the same time and call that fairness to anyone.
There is a second question, and in the long run it matters more: who owns the technology producing these gains? Canadian businesses are building their operations around foreign AI. Customer relationships, pricing models, software development and product design will increasingly depend on intelligence developed somewhere else, and AI will end up embedded in whatever proprietary knowledge a company has built. Most business owners are focused on the productivity gains. Far fewer are asking whether they’re comfortable building one of their most valuable assets on technology they don’t own, can’t control and can’t audit.
The frustrating part is that Canada helped invent this technology. The foundational research behind modern AI came out of Canadian universities and earned Canadian scientists the field’s highest honours. Yet according to Stanford’s AI Index, private AI investment over the decade to 2024 reached roughly US$470 billion in the United States against about US$15 billion in Canada, a gap of more than thirty to one that is still widening. We produce the researchers and export the returns. We’ve done this before: decades of shipping out raw materials and buying back finished products.
The predictable objection is that frontier AI models cost billions and Canada is a small market. That misunderstands the opportunity. Canada doesn’t need to outspend Silicon Valley. It needs domestic capability where it counts: models built for Canadian industries, trained on Canadian data, running on Canadian infrastructure, governed by Canadian law. We have the ingredients: world-class expertise in finance, mining and agriculture, abundant hydroelectric power, and a climate built for data centres. Previous generations built railways, banks and energy systems here because they formed the backbone of the economy. AI belongs on that list.
Getting there takes more than speeches. It takes three things: regulatory approvals for computing infrastructure measured in months rather than years, government procurement that gives real weight to Canadian data residency and Canadian-owned AI, and patient capital that keeps our companies from selling early or leaving. None of this requires new invention. It requires decisions.
Every generation faces a handful of choices that shape its economic future for decades, and this is one of ours. Canada can build the infrastructure, protect its intellectual property and become a country that exports intelligence. Or it can keep renting one of the most important technologies of this century and accept that the wealth, the expertise and the decisions will live somewhere else. We have everything we need except the urgency. It’s time to find it.
Romel Dhalla, is President of Dhalla Advisory Corp., provides strategic corporate finance advice to companies and high net worth individuals. He was a portfolio manager and investment advisor with two major Canadian banks for 17 years. Contact him at romel@dacorp.ca. Any views or opinions represented in this article are personal and belong solely to the author and do not represent those of people, institutions or organizations that the author may or may not be associated with in professional or personal capacity, unless explicitly stated. Any views or opinions are not intended to malign any religion, ethnic group, club, organization, company, or individual.
To find out about Canada’s new National Artificial Intelligence Strategy click here.