Key takeaways

• Across the four survey periods, 22% of workers reported using generative AI in their main job or business during the previous 12 months. Natural language processing—measured separately in the survey—was the next most commonly used technology, at 11%.
• Reported generative AI use rose to 30% in July 2025 from 17% in September 2024. Nearly half of this increase (7 percentage points) occurred between the March-to-April and June-to-July 2025 collection periods.
• Three industry sectors—professional, scientific, and technical services; educational services; and finance, insurance, real estate, rental, and leasing—accounted for 25% of workers, but 49% of generative AI users.
• Workers with a bachelor’s degree or higher were five times as likely to report generative AI use as workers with a high school diploma or less: 37% compared with 7%. This difference remained even after accounting for other measured job and worker characteristics.
• Among all workers, 16% said automated software or computer procedures affected their pace of work. The proportion was 24% among workers who used AI or automation technologies. However, the study does not establish that the technology caused the difference.

Plain language summary

Statistics Canada used the new Canadian Survey on Working Conditions to examine who is using AI or automation technologies at work. The survey looked at employed people aged 15 to 69 living in the provinces. Statistics Canada collected the data over four periods from September 2024 to July 2025. The study mainly describes patterns, although the authors also used statistical models to examine whether some differences remained after accounting for other measured characteristics.

Generative AI was the most-reported technology used. Across the survey periods, 22% of workers reported using it in their main job or business during the previous year. The estimate rose to 30% in the last collection period of the study from 17% in the first. The study also notes that worker-reported use was nearly twice the 12% of businesses that reported using AI in a separate 2025 survey. The authors say this may indicate that some AI adoption is being driven by workers alongside or outside formal organizational strategies.

Use was concentrated in work that involves generating, analyzing, and communicating information. About half of workers in natural and applied sciences occupations reported using generative AI. Use was much lower in manufacturing and utilities occupations and among those in trades, transport, and equipment operator occupations. Workers with more education were also more likely to report using generative AI, even after the authors accounted for industry, occupation, and other measured characteristics.

Workplace context also mattered. Reported use was lower among employees of smaller private sector firms than among several other groups, although most firm-size differences were not statistically significant after adjustment. Workers aged 55 to 69 were less likely than those aged 25 to 54 to report using generative AI, after adjusting for other factors. Women and men both reported 22% use overall, but women were less likely to report use after the authors accounted for differences in education, occupation, industry, and other characteristics.

The results describe whether workers had used a technology at least once during the previous 12 months. They do not show how often workers used the technology, which tasks it supported, whether employers approved its use, or whether it improved productivity. The study also does not measure job creation, job displacement, wages, or the quality of AI-generated work.

Definitions

Generative AI: Technology trained on large datasets that can produce new content—such as text, code, images, or speech—and support tasks such as answering questions and solving problems.

Worker-reported use: A worker’s report that they used a technology in their main job or business at least once during the previous 12 months. Frequency and intensity of use were not measured.

Adjusted association: A relationship that remains after a statistical model accounts for other measured characteristics. It does not prove that one characteristic caused another.

Work-pace dependence: A worker’s report that automated software or computer procedures affected the speed or pace of their work.

Why the results matter

For the public sector

Worker-reported generative AI use averaged 22%, nearly twice the proportion of businesses reporting AI use in a separate survey cited by the paper. The authors say this may reflect worker-driven use alongside or outside formal organizational strategies. Public-sector organizations can use worker-level data to inform workplace guidance, training, and technology governance. However, this study does not identify which approaches are effective.

For employers and employment services

The results suggest that procurement records and official implementation plans may not capture all workplace AI use. Employers may need to understand how workers are already using accessible tools and where guidance, support, or safeguards are missing. Employment services should also avoid drawing conclusions about an occupation’s future prospects based on its current level of AI use: the study measures recent use, not whether an occupation will grow, decline, or change.

For post-secondary institutions and students

Use of generative AI was highest in occupations that typically require a bachelor’s degree and among workers with higher educational attainment. This may suggest a need for practical AI literacy across knowledge-intensive fields, not only in computer science. However, students should not read the findings as proof that education leads to AI use or that using AI improves job outcomes because the study did not test either relationship.

For educators, career development professionals, and job seekers

The raw rates of generative AI use were 10% among workers aged 15 to 24, 27% among those aged 25 to 54, and 15% among those aged 55 to 69. The lower rate among younger workers disappeared after adjustment, while the difference for older workers remained.This suggests that differences in AI use by age may also reflect other factors, such as occupation, education, and job tasks. This indicates that career guidance can consider these factors alongside career stage rather than treating age alone as an explanation for AI use.

For people working with underrepresented and underserved groups

Women and men reported the same overall rate of generative AI use, but women were less likely to report use after accounting for other measured characteristics. The survey excludes people living in the territories, on Indigenous reserves, and in other settlements, and the article does not present results by disability, racialized or Indigenous identity, immigration status, or income. These gaps limit what the study can say about equitable access to workplace AI.

For people in the LMI ecosystem

The paper shows why AI measures need clear labels. Worker use, business adoption, occupational exposure, skill, productivity, and employment effects are different concepts. Regular worker-level data on frequency, tasks, training, and outcomes would make it easier to see whether the current adoption gaps narrow, persist, or translate into differences in job quality and opportunity.

For most Canadians

Use of generative AI is growing, but it is not evenly distributed across the labour market. A worker’s industry, occupation, education, age, and workplace context all relate to reported use. The results do not mean that workers outside high-use groups will avoid AI-related change. Nor do the results show that workers using AI will gain productivity or job security.

Notes and important disclaimers

• The Canadian Survey on Working Conditions represents employed people aged 15 to 69 living in the provinces. It excludes people living on Indigenous reserves and in other settlements, full-time members of the Canadian Armed Forces, and unpaid family workers.
• Statistics Canada collected the survey data during four periods from September 2024 to July 2025. The estimates describe reported use in the previous 12 months, but do not measure frequency or intensity of use.
• The technology categories are self-reported. Some may overlap or rely on similar underlying technologies.
• Most findings are descriptive. Adjusted associations from the statistical models do not establish cause and effect.
• The study does not measure productivity gains, job creation or displacement, wage effects, employer authorization, data protection, or the accuracy or quality of AI outputs.