One in Three Canadian Workers Now Use Generative AI. Only 4% Use It for Most Tasks.

35.9% of Canadian workers used a generative AI tool for their job in the year to March 2026. Do the arithmetic on how deeply they used it and that number falls to roughly 4%. Statistics Canada published both halves of that sentence on 30 July, and the space between them is the whole story.

The release is the first result from a new recurring Labour Force Survey supplement on AI use at work, covering workers aged 15 to 69 across the ten provinces. Because it is a first wave, it sets a baseline rather than a trend — a distinction worth holding onto, since it also means these numbers cannot be stacked against Canada's earlier working-conditions survey to manufacture a growth rate.

The "one in three" figure has a floor under it that nobody quotes

[Fact] 35.9% of workers reported using generative AI tools for work in the previous 12 months. [Fact] 41.6% used at least one AI or automation technology in their main job or business — so 5.7 percentage points of the workforce are touching automation of some kind without touching generative AI at all.

Now the part that changes the shape of the finding. Statistics Canada asked users how much of their work the tools actually cover. [Fact] 63.5% of users said "some but not most tasks." [Fact] 24.9% said "almost no tasks." Only 8.0% said most tasks, and 3.6% almost all.

[Estimate] Multiply through and a different picture appears. If 11.6% of users apply these tools to most or almost all of their tasks, then about 4.2% of all Canadian workers do (35.9% × 11.6%). Meanwhile roughly 8.9% of all workers are counted as users while reporting that the tools touch almost none of their work. That shallow group is more than twice the size of the deep-use group.

Frequency says the same thing from another angle. [Fact] Among users, 31.4% use the tools daily and 38.3% a few times a week. [Estimate] That puts weekly-or-more users at about 25.0% of the whole workforce, and daily users at 11.3%.

So: a quarter of Canadian workers reach for generative AI in a normal week. Roughly one in twenty-four leans on it for most of the job. Both numbers are true, and only the first one travels.

Adoption is highest where AI assists, slower where it substitutes

This is the finding that cuts hardest against the standard narrative, and it comes from Statistics Canada sorting occupations on two axes at once instead of one.

The agency splits jobs into high exposure with high complementarity — described in the release as doctors, nurses, teachers and engineers, whose tasks carry "more potential complementarity with AI"; high exposure with low complementarity — retail sales, office support, software development and accounting, which "may be more susceptible to task replacement by AI"; and low exposure, meaning skilled trades, service jobs and first responders.

If adoption tracked replacement risk, the low-complementarity group would lead. It does not.

[Fact] High-exposure, high-complementarity workers report 53.8% use. High-exposure, low-complementarity comes in at 45.9%. Low-exposure occupations sit at 14.2%.

The complementarity group leads the replacement-risk group by 7.9 percentage points. The spread from there down to low-exposure work is 39.6 points — five times as wide as the gap that gets all the attention.

Read across broad occupational categories and the gradient steepens. [Fact] Legislative and senior management roles report 75.1% use and natural and applied sciences 67.5%, while natural resources and agriculture sit at 17.0% and trades, transport and equipment operators at 14.7%. That is a 5.1× ratio between the top and bottom of the Canadian labour market.

Industry follows the same line. [Fact] Professional, scientific and technical services 65.6%, finance, insurance, real estate, rental and leasing 59.2%, educational services 53.0%, transportation and warehousing 21.1%, agriculture 17.5%, accommodation and food services 16.3%. Top to bottom, a 4.0× difference.

If you work in accounting, retail sales, office support or software development, Statistics Canada has placed you in the category where AI is likelier to substitute for tasks than to assist them — and your peer group is adopting more slowly than nurses, physicians and engineers. Adoption is not the same thing as safety. But it does mean the tools are being pulled in fastest by workers whose jobs this same framework expects to outlast them.

Young workers are not leading this

[Fact] Within high-exposure, high-complementarity occupations, use runs 56.9% among workers aged 25 to 54, 45.3% among those 55 and over, and 39.1% among those aged 15 to 24.

Read that again. The youngest workers come last, 17.8 percentage points behind the middle group, inside the very occupational class where these tools are most useful.

The release does not test why, and there are ordinary explanations available: junior staff often have less discretion over their own workflow, less access to paid tools, and closer review of what they submit. The point is narrower than any explanation. "Young people adopt it first" is an assumption, and in Canadian data it runs the other way.

Two further splits deserve naming. [Fact] Within the same high-complementarity class, men report 57.0% use against 50.9% for women — a 6.1-point gap that persists with occupational exposure held roughly constant, so it is not merely a matter of which jobs men and women hold. [Fact] And public sector employees, at 41.2%, out-adopt private sector employees at 33.4% by 7.8 points, with the self-employed between them at 39.6%. The familiar assumption that public institutions lag on new tools does not survive this table.

The barrier is relevance, not skill

Retraining is the reflex policy answer. The non-user data suggests it is aimed at the wrong obstacle.

Among the 64.1% of workers who did not use generative AI, [Fact] 56.0% said it was not applicable to their work. Only 5.8% cited a lack of skills or knowledge. 21.6% reported no interest, 9.8% raised security, privacy or ethical concerns, and 5.0% said company or organizational policies limited their use.

[Estimate] Scale those against the full workforce and something odd surfaces. 64.1% × 56.0% works out to 35.9% of all Canadian workers saying generative AI does not apply to their job — the same figure, to one decimal place, as the share who used it. The workforce splits almost exactly in half between people using these tools and people who have concluded they are irrelevant. Only about 3.7% of all workers report a skills barrier, which is roughly a tenth the size of the relevance gap that training programmes are usually sold to close.

Here is the honest limitation, and it matters more than the symmetry. "Not applicable to my work" is a self-assessment, not a measurement. A worker who has never seen a demonstration aimed at their own role will reasonably say it does not apply. That answer is strong information about perception and weak information about technical feasibility — and perception is the thing that moves fastest once an employer makes a decision.

What a first wave is good for

The most useful property of this release is that it will repeat. One month cannot show direction, and Statistics Canada is explicit that the estimates are drawn from a sample and carry sampling variability. The exposure-and-complementarity groupings are assignments made from task descriptions, not observations of what happened to anyone. And these figures cannot be compared with Canada's earlier survey of working conditions, which put different questions to a different sample — anyone quoting a Canadian AI-adoption "doubling" is stitching together two instruments that were not built to join.

What it does give you is a position on the map. [Claim] If you are in a high-exposure, low-complementarity role, the gap worth closing is not the 7.9 points behind nurses and engineers; it is the 29.2 points behind senior management, because the people deciding how AI enters your workplace are already using it at 75.1% against your 45.9%. If you are in a low-exposure trade or service job, 14.2% adoption is not a reason to dismiss the question. It is time you have and others do not.

And if you already use these tools, the depth numbers are the ones to measure yourself against. Most Canadian users are in the shallow end — a few times a week, some but not most tasks. Being deliberate about which of your tasks the tools genuinely improve, rather than counting yourself a user because you tried one, is the distinction this survey can actually see.

Sources

• Statistics Canada, Use of generative artificial intelligence tools among Canadian workers, March 2026, The Daily, 30 July 2026. Labour Force Survey supplement, reference period March 2026; workers aged 15 to 69 in the ten provinces, excluding persons living on reserves, full-time members of the Canadian Armed Forces, and the institutionalized population. Described in the release as the first results from a regular series on AI use by Canadian workers.

All percentages attributed to Statistics Canada are read directly from the 30 July 2026 release. Figures marked [Estimate] are this site's own arithmetic on those published shares — the product of a reported user share and a reported within-user distribution — and inherit the sampling error of both, so they should be read as approximate rather than exact. The exposure-and-complementarity occupational grouping is Statistics Canada's own classification, not ours.

AI-assisted analysis: this article was drafted with AI assistance from the primary source release and reviewed before publication.