The Uneven Rise
of Workplace AI
The first release in a new regular series on generative AI use by Canadian workers, drawn from a supplement to the March 2026 Labour Force Survey. Adoption is real. But it splits the same way every time, along the line of what kind of job you have.
The short version, six numbers deep
- 35.9% of Canadian workers used a generative AI tool at work in the 12 months to March 2026. A wider 41.6% used some form of AI or automation technology, so generative tools are nearly all of what "AI at work" currently means in Canada.
- Industry spread is roughly fourfold. Professional, scientific and technical services sits at 65.6%; accommodation and food services at 16.3%. The all-industry average is 35.9%.
- Occupational exposure predicts use better than anything else. 53.8% in high-exposure, high-complementarity roles and 45.9% in high-exposure, low-complementarity roles, against 14.2% in low-exposure occupations.
- The gender gap only appears where AI exposure is high. 14.1% of men versus 14.4% of women in low-exposure work — effectively no gap — widening to 52.9% versus 41.2% in high-exposure, low-complementarity occupations.
- Most non-users aren't afraid of it, they just don't see the point. Of the 64.1% who didn't use generative AI, 56.0% said it doesn't apply to their work and 21.6% had no interest. Security and privacy concerns came in at 9.8%.
- Users are mostly dabblers. 63.5% used it for some tasks but not most; only 11.6% used it across most or nearly all of their tasks. Among users, 38.3% reach for it a few times a week and 31.4% daily.
Everyone's heard of it.
Half see where it fits their job.
Awareness of generative AI is close to universal among Canadian workers. What's not universal is the sense that it has anything to do with the work in front of you.
In March 2026, 93.4% of workers said they were aware of generative AI tools. Not knowing about it is now the outlier position, only 6.6% hadn't heard of it. But awareness is the easy part. Among those who'd heard of it, just over half felt they understood how it could apply to their own job, and a further 37.4% knew the tools existed but genuinely didn't think they were relevant to what they do. That's the real fault line here: not ignorance, but applicability.
Among workers aware of generative AI (93.4% of all workers)
Share of aware workers by familiarity with applying the tools to their own work
One in three, doing the work differently
41.6% of workers used some form of AI or automation technology on the job in the past year. Generative AI accounts for nearly all of it: 35.9% of all workers, or just over one in three. The gap between the two bars below is small and worth sitting with for a second, it means generative tools aren't a niche corner of "AI at work," they're most of what that phrase currently means in Canada.
Share of workers using AI technology at work, past 12 months
Comparing any AI or automation technology against generative AI specifically
Knowledge work adopted it first. By a lot.
Professional, scientific and technical services sit at 65.6%, roughly four times the rate in accommodation and food services (16.3%). The dashed line marks the all-industry average of 35.9%, most industries sit well on one side of it or the other, there isn't much of a "typical" industry here. The pattern is legible: industries built around documents, analysis and client communication took to the tools fastest. Industries built around physical tasks, shifts and equipment mostly haven't yet.
Generative AI use at work, by industry
Share of workers using generative AI tools in the past 12 months
The occupations AI actually touches
StatCan sorts occupations using a complementarity-adjusted AI occupational exposure index (C-AIOE), built on work by Felten, Raj and Seamans, extended by Pizzinelli and colleagues, and applied to Canada by Mehdi, Morissette and Frenette. It splits jobs into three buckets: high exposure with high complementarity (doctors, teachers, engineers, work AI is more likely to augment than replace), high exposure with low complementarity (retail, office support, parts of accounting and software work, more exposed to task replacement), and low exposure (trades, frontline service, first response). The gap between these groups is the single strongest predictor in the whole release.
Generative AI use, by occupational exposure group
HEHC = high exposure, high complementarity · HELC = high exposure, low complementarity · LE = low exposure
The pattern holds inside every group too: core-aged workers (25 to 54) lead, younger and older workers trail. It's probably less about AI comfort and more about who's mid-career in exactly the kind of role where these tools have an obvious use case right now.
Generative AI use by exposure group and age
Share of workers within each age group and exposure category
A gap that opens where AI matters most
In low-exposure occupations, men and women use generative AI at essentially identical rates (14.1% vs. 14.4%). Move into the high-exposure, low-complementarity group and a gap opens: 52.9% of men versus 41.2% of women, 11.7 points. It narrows again in the high-complementarity group but doesn't close. StatCan's release doesn't explain why, and neither will we, but it's worth noting the gap tracks exposure rather than running flat across the whole labour market.
Generative AI use by gender, within each exposure group
Share of men and women using generative AI at work
Management leads. The trades lag.
Three in four workers in management occupations (75.1%) used generative AI at work, the highest of any group StatCan named in this release. Natural and applied sciences follows at 67.5%. At the other end, trades, transport and equipment operators sit at 14.7% and natural resources and agriculture occupations at 17.0%. These four are the only broad occupational groups singled out in the release, not the full breakdown, but they trace the same line the exposure index draws: desk-and-decision work versus hands-and-equipment work.
Generative AI use, selected occupational groups
Highest and lowest groups named in the release, not a full occupational breakdown
The sector gap is mostly a job-mix gap
On the surface, public sector employees (41.2%) use generative AI more than private sector employees (33.4%), with self-employed workers in between (39.6%). But split by occupational exposure and most of that gap evaporates: within high-exposure, high-complementarity roles, private (54.9%) and public (54.5%) sector workers use the tools at almost the same rate, and the pattern repeats in the high-exposure, low-complementarity group (45.8% vs. 43.1%). The sector difference looks less like a culture difference and more like an artifact of which occupations each sector happens to employ more of.
By sector, overall
All occupations combined
By sector, within exposure group
Self-employed not broken out by exposure group in this release
Most users are dabblers, not power users
Using generative AI at all and using it constantly are very different states. Among people who used it in the past year, 63.5% did so for some tasks but not most, the moderate middle. Only 11.6% used it broadly, across most or nearly all of their tasks. The picture that emerges is less "AI transformed my job" and more "AI showed up in a few corners of my job," at least so far.
Extent of generative AI use among users
Share of workers who used generative AI at work, past 12 months
Here's the twist: broad usage isn't highest among the theoretically most-augmented HEHC group, it's highest in HELC, the high-exposure, low-complementarity occupations that are also the ones more exposed to task replacement. The jobs closest to being automated are also, right now, the ones where the tools get used hardest.
Moderate use, by exposure group
Some tasks, not most
Broad use, by exposure group
Most or nearly all tasks
Weekly is the norm. Daily is common too.
Among generative AI users, 38.3% reach for it a few times a week and 31.4% use it daily, together nearly seven in ten users. Daily use concentrates hardest in high-exposure roles regardless of complementarity (37.1% in HELC, 31.1% in HEHC) versus 18.2% in low-exposure occupations. By named occupation, daily use runs highest in natural and applied sciences (45.6%) and lowest in manufacturing and utilities (18.6%) and natural resources and agriculture (18.2%).
Frequency of generative AI use among users
Segments may not sum to exactly 100% due to rounding
Daily use, by exposure group
Share of users reporting daily use
Why 64.1% of workers opted out
The dominant reason for not using generative AI at work isn't fear, cost or a skills gap, it's that most non-users simply don't see it applying to their job (56.0%). One in five (21.6%) has no interest. Everything else trails well behind. Split by class of worker, two smaller patterns stand out: security, privacy and ethical concerns are more than twice as common among public sector employees (16.9%) as private sector employees (7.4%), and a lack of skills or knowledge is reported almost twice as often by self-employed workers (10.7%, use with caution) as by employees in either sector, probably because self-employed workers don't have an employer's training budget to lean on.
Main reasons for not using generative AI at work
Share of non-users citing each reason, workers aged 15 to 69 (multiple reasons could be selected)
Seven things people ask about AI at work in Canada
What share of Canadian workers use generative AI at work?
35.9% of Canadian workers used a generative AI tool at work in the 12 months to March 2026 — just over one in three. A wider 41.6% used some form of AI or automation technology, meaning generative tools account for nearly all AI use at work in Canada.
Which Canadian industries use generative AI the most?
Professional, scientific and technical services leads at 65.6%, followed by finance, insurance, real estate, rental and leasing at 59.2% and educational services at 53.0%. Accommodation and food services is lowest at 16.3%, with agriculture at 17.5% and transportation and warehousing at 21.1%. The all-industry average is 35.9%.
What is the C-AIOE index?
The complementarity-adjusted AI occupational exposure index scores each occupation on two axes: exposure, meaning how much AI can influence or perform its tasks, and complementarity, meaning how likely AI is to augment rather than replace the work. Statistics Canada groups occupations at the median of each axis into high exposure/high complementarity (doctors, teachers, engineers), high exposure/low complementarity (retail, office support, parts of accounting and software work) and low exposure (trades, frontline service, first response). The index was developed by Felten, Raj and Seamans, extended by Pizzinelli and colleagues, and applied to Canada by Mehdi, Morissette and Frenette.
Is there a gender gap in workplace AI use in Canada?
Only where AI exposure is high. In low-exposure occupations men and women use generative AI at essentially identical rates, 14.1% versus 14.4%. In high-exposure, low-complementarity occupations the gap is 11.7 points, 52.9% of men against 41.2% of women. In high-exposure, high-complementarity occupations it narrows to 57.0% versus 50.9% but does not close.
Why do most Canadian workers not use generative AI at work?
The dominant reason is applicability, not fear or cost. Among the 64.1% of workers who did not use generative AI, 56.0% said it does not apply to their current work and 21.6% said they have no interest. Security, privacy, environmental or ethical concerns were cited by 9.8%, lack of skills or knowledge by 5.8%, and company or organizational policy by 5.0%.
How often do Canadian workers who use generative AI reach for it?
Among users, 38.3% use generative AI a few times a week and 31.4% use it daily — nearly seven in ten at weekly-or-better cadence. A further 21.7% use it a few times a month and 8.7% a few times a year. Daily use is highest in high-exposure, low-complementarity occupations at 37.1% and lowest in low-exposure occupations at 18.2%.
Do Canadian public sector workers use AI more than private sector workers?
Only on the surface. Public sector employees use generative AI at 41.2% against 33.4% for private sector employees, with self-employed workers at 39.6%. But within the same occupational exposure group the gap all but disappears: 54.9% private versus 54.5% public in high exposure/high complementarity roles, and 45.8% versus 43.1% in high exposure/low complementarity roles. The sector difference is mostly a job-mix difference.
What this is
The first release from TechStat, a new Statistics Canada measurement program announced in Budget 2025 covering AI adoption, use and impact for businesses, individuals and employees. This release draws on supplementary questions added to the Labour Force Survey for March 2026, covering workers aged 15 to 69 living in the provinces (territories, Indigenous reserves, institutional populations and full-time regular Armed Forces members are excluded). A further collection is planned for September 2026, with regular repeats after that.
The exposure index
Occupations are classified using the complementarity-adjusted AI occupational exposure (C-AIOE) index, originally developed by Felten, Raj and Seamans (2021), extended by Pizzinelli et al. (2023), and applied to the Canadian context by Mehdi and Morissette (2024) and Mehdi and Frenette (2024, 2026). It scores each occupation on exposure (how much AI can influence or perform its tasks) and complementarity (how much AI is likely to augment rather than replace the work), then groups occupations using the median score on each axis.
Caveats
- Estimates are based on a sample and subject to sampling variability; figures flagged "E" should be used with caution.
- The analysis in the original release focuses on differences significant at the 95% confidence level.
- Rounding means some grouped percentages in this piece don't sum to exactly 100%.
- Selected occupational and industry figures reflect only the categories StatCan highlighted in the release text, not the complete underlying tables beyond what's charted above.
Full release: Use of generative artificial intelligence tools among Canadian workers, March 2026, Statistics Canada, The Daily, July 30, 2026.
License: the underlying figures are the Government of Canada's. Northern Signal's compilation, charting and arrangement of this data is offered under CC BY 4.0.
The next TechStat release lands in September
Northern Signal charts Canadian AI data like this every week — policy, funding, adoption — in one free email. New Atlas dossiers are announced to readers first.
Get the weekly briefing