Canadians are scared of AI. Their employers keep quietly using more of it.

Start with how Canadians feel about it. In a survey the Angus Reid Institute released on June 2, 68 per cent said the government should heavily regulate AI even if that slows the technology down. Three-quarters doubt any government can keep up with it. Last fall the same pollster found that fewer than three in ten Canadians think AI is a force for good. And next to the rest of the world, that gloom stands out. When Ipsos asked people in 30 countries last spring whether products and services using AI make them excited, Canada finished dead last. Just 31 per cent of Canadians said yes, against a 52 per cent average across the 30 countries and 80 per cent in Indonesia. No country surveyed was less enthusiastic, and Canada's own figure has slid from 37 per cent two years earlier.

So you might expect the workplace to be holding the line. It is not. Statistics Canada asks businesses every quarter whether they have used AI to make their goods or deliver their services, and the answer keeps climbing. In the spring of 2024, 6.1 per cent said yes. A year later it was 12.2. This spring it hit 19.2 per cent. The number has roughly tripled in two years while the public mood went the other way.

Tripling sounds dramatic, and it is, but read the level too. Four out of five Canadian businesses still did not use AI in the past year. This is a fast-growing minority, not a takeover. If you pictured AI as something already woven through the whole economy, the data says you are a few years early.

It is mostly a white-collar tool

The national average hides a split that is much larger than the average itself. Break the number down by industry and AI adoption runs from almost nothing to nearly half.

Information and cultural industries lead at 42.3 per cent. Finance and insurance is right behind at 40.4, and professional, scientific and technical services at 32.4. Then it falls off a cliff. Accommodation and food services sit at 12.7 per cent, construction at 9.2, and agriculture, forestry and fishing at 4.5. A finance firm is nine times more likely to use AI than a farm.

None of that is surprising once you say it plainly. The industries adopting AI fastest are the ones that already ran on screens and spreadsheets. AI mostly reads, writes, sorts and predicts, so it lands first where the work was already reading, writing, sorting and predicting. The trades and the fields are not behind because they are slow. They are behind because a chatbot does not pour concrete or calve a cow.

Firm size cuts a stranger line. The very smallest businesses, those with one to four employees, use AI at 19.9 per cent, slightly above the national average and well above firms with five to nineteen workers at 14.9. Then it climbs again with size, up to 27.8 per cent at companies with a hundred or more staff. The solo operator running everything through a chatbot and the large firm with a real budget are both ahead. It is the small-but-not-tiny shop in the middle that has not bothered.

The holdouts are not scared. They are indifferent.

Here is where the workplace data and the public mood part ways completely. When StatCan asks businesses what limits their use of AI, the runaway answer is not fear, cost or regulation. It is a shrug.

Forty per cent of businesses say AI is simply not relevant to what they do. That is by far the most common answer, three times more common than cybersecurity or privacy worries at 13.4 per cent, and nearly four times more common than cost at 10.6. Only 7 per cent named regulatory concerns, and another 7 named a lack of skilled workers. The thing the public is loudly anxious about, the thing two-thirds want Ottawa to clamp down on, barely registers as a reason businesses themselves stay away. They are not holding AI at arm's length because they are afraid of it. Most of them just do not see what it would do for them.

And the firms using it are mostly training people, not cutting them

The loudest fear about AI is jobs. The survey cannot settle that, but it points away from the nightmare version. Among businesses that do use AI, 55.6 per cent made no change at all to how they staff or train. Of the ones that did change something, the most common move by far was to train their existing employees, at 32 per cent, followed by training executives at 21.6. Far fewer went out to hire new people with AI skills, at 11.4 per cent.

Notice what is not on that list. The survey does not even offer "laid off staff" as an option, so it cannot measure job loss directly, and that is a real limit. But what businesses say they are doing is teaching the people they already have to use a new tool, which is closer to how companies absorbed the spreadsheet or the email inbox than to a wave of pink slips. For now, the visible effect of AI on Canadian payrolls is a training budget, not a severance one.

So who is right, the worried public or the calm data?

Both, about different things. The public is mostly worried about where this goes next: the deepfakes, the surveillance, the jobs that might vanish in five years. The business data describes where it is now: a tool that a minority of office-based firms picked up quickly, that most other firms find beside the point, and that so far is being handed to existing staff rather than used to replace them. A fear about the future and a measurement of the present can both be honest and still not match.

What the gap should do is lower the temperature on the present-tense version of the panic. AI is not yet washing over the whole Canadian economy. It is concentrated, it is uneven, and the businesses skipping it are bored by it more than they are frightened. The argument worth having is about the next five years, not a takeover that, in the numbers we can actually see, has not happened.

Caveats worth taking seriously

This is a survey of businesses, not households, so it says nothing direct about how individual Canadians use AI in their own lives, where adoption is almost certainly higher. "Used AI" is self-reported and broadly defined: it spans everything from a large language model to basic data analytics and a customer-service chatbot, so a "yes" can mean very different things. The figures are rounded, and the survey is one snapshot a quarter. And as noted, the staffing question offers no way to report layoffs, so it can describe how firms are adapting to AI but not whether anyone lost a job to it. The broad shape, a fast-rising but still-minority adoption rate, a heavy tilt toward white-collar industries, and holdouts who cite irrelevance over fear, held across every cut of the data we pulled.

Sources & data