OpenAI released ChatGPT in November 2022. In Canada, the mainstreaming of artificial intelligence blended into an existing scramble for people with computer skills. Employers had added more than 111,000 such people since the end of 2019, a 41-per-cent increase, according to Statistics Canada’s monthly payrolls survey.

It was an exciting time. All that hiring appeared to confirm then-prime minister Justin Trudeau’s idea that a modern Canada could be known for its “resourcefulness,” and not just its resources. Total payrolls had increased only about four per cent over the same period, so maybe the shift to a nation of designers of software and builders of meta-worlds was on. Bartenders and servers who lost their jobs during the COVID-19 pandemic were learning to code. The Fourth Industrial Revolution was setting up to be a non-violent one, at least for Canada, home to more than a few of AI’s leading visionaries. 

The AI fairy tale has taken an uncomfortable twist. The promise remains in the forefront. Canada now has a federal cabinet minister devoted to AI, elevating the status of its practitioners to that of the country’s foundational industries—agriculture, natural resources and fisheries—which have had their own minister for decades.

Former Bank of Canada governor Stephen Poloz thinks the technology could solve the country’s productivity problem and cause us to stop obsessing over U.S. tariff rates. “By far the biggest potential move for Canada is to fully embrace AI and the rest of the Fourth Industrial Revolution,” he said Thursday at an event hosted by Western University’s Lawrence National Centre. “Facilitate it. Speed it up in every way that we can.”   

The incredible amount of money that the biggest AI companies are spending on developing the technology makes you wonder if faster deployment is possible. The stories this summer about what Meta was willing to do to poach the best AI researchers offered additional evidence that the world’s most powerful companies were all-in. 

But here in Canada, something unusual was happening. That hiring scramble for computer talent lost momentum—and went into reverse.

The number of people working in positions that Statistics Canada categorizes as “computer systems design and related services” peaked at about 390,000 in June 2023, a few months after OpenAI released GPT-4. Two years later, nearly 14,000 of those positions were gone, according to Statistics Canada’s payroll survey. The composition of Canada’s labour market is still more digitally oriented than a few years ago, but as the AI bubble continued to inflate, digital technology employment stagnated. 

Why? 

Initially, the deceleration looked like a correction. Tech companies such as Shopify overhired during the COVID-19 pandemic when life moved online, and then fired material numbers of people when they realized they had misjudged humanity’s desire to populate a metaverse. More recently, U.S. President Donald Trump’s upending of the global trading order has left Canada’s economy teetering on the edge of a recession. Total labour demand, which Statistics Canada measures by combining employment and vacancies, declined in the second quarter, according to data released this week. 

But it’s becoming increasingly difficult to avoid the thought that employers have realized that they can improve their internal output per hour worked by licensing AI models and reducing their headcounts. “When you develop a tool, it makes [things] easier and takes you to a higher strategic level,” said former Alberta energy minister Ken Hughes, who now chairs the board of Providence Therapeutics and is vice-chair of Beacon AI Centers’ board of directors. “That’s what [AI] is doing, but it is creating a gap in existing organizations. There’s kind of a big gap at the bottom that is no longer necessary. Normally, technology doesn’t have that impact that quickly. The thing that’s different this time is that it’s happening overnight.” 

Hughes got in touch after I wrote about the alarming spike in youth unemployment. Weaker growth could explain it. Employment levels of younger workers are leading indicators of an economy’s strength and employers have all kinds of reasons to focus on the workers they already have, rather than train up new ones. But earlier this year Hughes started to think that something else was happening after he used Perplexity to generate a perfectly acceptable terms-of-reference document in a couple of minutes. “Five years ago—two years ago—I would have handed this off to a director or a mid-level person,” he said.  

At the start of the summer, The Logic’s Catherine McIntyre flagged that AI might be replacing humans as the top choice for internships. Her story was based primarily on anecdotal evidence and informed speculation like that of Hughes. Circumstantial evidence is starting to pile up, too. 

A year ago, Statistics Canada researchers Tahsin Mehdi and René Morissette published a study that identified data entry clerks, computer network technicians and computer programmers among professions that were highly vulnerable to AI-powered automation. Mehadi and Morissette also found that workers with bachelor’s degrees faced the greatest risk of redundancy, upending conventional wisdom that a university degree was a guarantee of success. That might now depend on the degree. Employers were offering to pay software engineers and designers an average of $53.80 per hour in the second quarter, according to Statistics Canada’s latest vacancy data, a seven-per-cent increase from a year earlier. 

At the same time, the average offered hourly rate for database analysts and computer systems managers fell over the same period. There were five unemployed people with a bachelor’s degree or higher for every vacancy, significantly more than any other educational group. It’s also  the most since 2015, according to Statistics Canada analyst Jonathan Bridekirk. “This is high,” he said. “It’s a tight labour market for higher educated people.”

It’s wise to avoid conclusions based on anecdotal and circumstantial evidence. But given the stakes, it might also be wise to accelerate plans about what to do if AI already is knocking out the lower rungs of organization charts. Hughes said he hears even ministerial offices are turning to AI over hiring junior staffers. “That’s how I got my start,” he said. 

Hughes eventually used that experience to get elected, and he’s now putting it to use helping to lead innovative companies. AI can’t yet replace human experience. But what happens if it blocks or delays a generation of humans from getting that experience? The inevitable disruption will be worse.  

Kevin Carmichael is The Logic’s economics columnist and editor-at-large. He has spent more than two decades covering economics, business and finance for outlets including Bloomberg News, The Globe and Mail and the Financial Post, where he also served as editor-in-chief.