What Statistics Canada’s 2026 AI-Use Series Shows

What does Statistics Canada’s first 2026 series show?

Statistics Canada’s first March 2026 worker-use snapshot shows meaningful reach, but limited task breadth for most users. The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026 reports that 35.9% of workers aged 15 to 69 used generative AI in their main job or business during the previous 12 months. Separately, among those users, 63.5% said it currently touched some but not most tasks. Another 24.9% reported almost no tasks, while 11.6% reported most or almost all. The first figure describes all workers; task-breadth percentages describe users only. They answer different questions and cover different time frames.

The release therefore establishes workplace reach, not widespread coverage of work. It does not show that AI independently completed reported tasks, how much time they took, whether outputs were accurate, or how much checking remained. A worker-use estimate also cannot establish employer-wide deployment or job losses. Statistics Canada says later collection is planned for trend monitoring, so this first March result is a baseline, not a trend. For a worker deciding what to do, the next question is which recurring task is changing in their own workflow, rather than what a national percentage predicts about their job.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026

What exactly did the March 2026 survey ask?

The March 2026 Labour Force Survey supplement asked respondents aged 15 to 69 who lived in the provinces about artificial intelligence in their main job or business. Its questions separate three things that can sound like one statistic: whether a worker had used any listed AI or automation technology in the previous 12 months, how often generative AI is used currently, and what proportion of tasks currently involve it. The release, “Use of generative artificial intelligence tools among Canadian workers, March 2026,” reports the first wave of a planned regular series. The survey universe also has defined exclusions: people living on reserves, full-time members of the regular Armed Forces, and people living in institutions were not included. The first usage question is deliberately broader than generative AI. It asks whether, during the last 12 months, the person used any of several technologies as part of their main job or business: generative AI, natural language processing, voice recognition software, machine learning, or another technology. The answer “none of the above” is also available. Statistics Canada reports that 41.6% used at least one technology in this basket; within it, generative AI was the most commonly reported, at 35.9%. Those figures have the same past-year frame, but they answer different questions. The broader basket matters because a worker can report a non-generative technology without reporting generative AI; the two percentages should not be treated as competing estimates of the same event: 41.6% is not the generative-AI rate, and 35.9% does not mean use was routine. For generative AI, the instrument then asks users two separate questions prefaced by “Currently.” One asks how often they use it at work, offering daily, a few times a week, a few times a month, or a few times a year. The other asks for what proportion of their tasks they use it, with choices of almost all, most, some, or almost no tasks. The questionnaire’s “Labour Market Indicators – March 2026” makes this sequence explicit. Frequency and breadth are not interchangeable: a tool could be used often for one narrow duty, or less often across several kinds of work. Neither response records exact hours, a task-by-task inventory, output quality, checking time, or whether the tool performs a task without human involvement. That wording sets the proper boundary for the series. The past-year item measures reported reach; the current items add rough measures of frequency and task coverage among users. Together they create a useful baseline for later comparison, but they cannot establish how much work AI performs, whether it improves results, or whether a workplace has adopted it across a team. When reading the percentages that follow, keep each denominator and reference period attached.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026; Labour Market Indicators – March 2026 questionnaire

What is the headline use figure actually counting?

Statistics Canada’s 35.9% estimate counts workers aged 15 to 69 who said they had used generative artificial intelligence as part of their main job or business at least once during the previous 12 months. “Use of generative artificial intelligence tools among Canadian workers, March 2026” presents it as past-year use among all workers. It is not a measure of how often each person uses a tool, how many tasks involve it, or how much of an employer’s workforce has been formally equipped or instructed to use it. Nor does a yes answer establish use from week to week. An occasional trial and repeated use could both produce a yes answer. The estimate therefore establishes reach over a year, while leaving regularity and depth for other questions.

The release also reports that 41.6% used at least one AI or automation technology in their main job or business over that period. That wider basket includes generative AI alongside natural language processing, voice recognition software, machine learning and other technologies. It should not be read as a second estimate of generative-AI use: the 35.9% figure is the specific generative-AI result, and the larger figure covers several kinds of technology. These are reports of use, not an audit of workplace systems or employer policy. Availability to one worker does not establish organization-wide adoption.

The questionnaire makes the time distinction explicit. In Labour Market Indicators – March 2026, Statistics Canada asks first whether respondents used listed technologies in the last 12 months. A separate item asks users how often they currently use generative AI at work, with answers ranging from daily to a few times a year. Another asks what proportion of tasks they currently use it for, with qualitative choices from almost no tasks to almost all tasks. Past-year use can include a trial no longer part of current work; frequency concerns recurrence and task proportion concerns breadth. Neither question measures the quality of output, time saved, amount of checking, or whether the tool can complete a task without human involvement. The terms “some” and “most” are survey categories, not audited shares of a worker’s hours or a count of duties.

For a worker deciding what to do next, the headline is a reason to inspect a real workflow, not to infer that their occupation has broadly adopted AI. Choose one recurring task and establish whether a permitted tool is used, what it produces, and what review or correction the person still performs. The national estimate cannot supply that individual detail. It also cannot tell whether an employer has changed staffing or responsibilities because of AI. Read the 35.9% as meaningful reported reach; look to the separate frequency and task-breadth measures for different questions, and to direct workplace evidence for an individual decision.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026; Labour Market Indicators – March 2026 questionnaire

How broad is use among the workers who use it?

Among workers who reported using generative AI for their main job or business during the previous 12 months, 63.5% said they currently used it for some but not most tasks, 24.9% for almost no tasks, 8.0% for most tasks and 3.6% for almost all tasks. The last two categories total 11.6%. In Statistics Canada’s March 2026 release, then, the modal answer is limited or moderate reach; reported use across most or nearly all tasks is less common. These are shares of past-year users answering a current task-proportion question, not shares of all Canadian workers or of all work performed. That denominator changes the meaning of the figures. The 63.5% does not mean that nearly two-thirds of workers use AI on some tasks: it means nearly two-thirds of the users selected that category. Nor does it mean that 63.5% of each user’s tasks involve AI. The questionnaire offers broad labels, not a list of duties to count. A person with a narrow set of duties and a person with many different duties can both answer “some tasks,” even if the tool enters a different number or kind of activities in each case. “Almost no tasks” also matters. Someone may have used a tool during the past year but currently apply it to almost none of their work. Past-year use does not establish that the practice continued or became routine. In the other direction, “most tasks” does not describe how often the worker uses the tool, how much of each task it handles, or whether the output is usable without revision. The categories indicate perceived breadth, not an observed automation rate. The distinction is easier to see by reading the questions in Statistics Canada’s “Labour Market Indicators – March 2026” questionnaire. It first asks whether respondents used generative AI as part of their main job or business in the last 12 months. For eligible users, it separately asks how often they currently use it and what proportion of tasks currently involve it. These are different dimensions: a tool could be used daily on a narrow recurring duty, or less often across several kinds of work. Frequency cannot fill in task breadth, and breadth cannot tell us frequency. Neither answer, by itself, shows whether the tool completes a task independently. Imagine a worker who uses an approved tool each morning to summarize one recurring document. That could be frequent use while still covering only a small part of a varied job. Another worker might apply a tool occasionally to drafts, research notes and routine data descriptions, yet still classify the share of all duties as “some.” These are illustrations, not survey cases. A person may still need to check claims, handle exceptions and take responsibility for the final work. The questions do not measure review time or output quality. For a career decision, the useful unit is therefore a named task, not the national label. Write down one recurring duty, what information goes in, what the tool returns, what must be checked ed, and who is accountable for the result. If a permitted tool is involved, compare several real instances with the current process before deciding whether the task has changed meaningfully. The survey’s 63.5% describes a common self-reported category among users; it does not tell any particular worker which duties are exposed, or whether their role is shrinking. That requires examining the workflow.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026; Labour Market Indicators – March 2026 questionnaire

Does occupational exposure map neatly onto broader use?

No. Among workers who had used generative AI at work in the previous 12 months, moderate use was the largest category in every occupational exposure group. Broad use, meaning use for most or almost all tasks, was reported by 15.4% of users in high-exposure, low-complementarity occupations, compared with 9.2% in high-exposure, high-complementarity occupations and 9.9% in low-exposure occupations. The result complicates a simple assumption that greater occupational exposure automatically means wider use across a worker’s tasks.

Chart 3, “Extent of generative artificial intelligence (AI) use across work tasks among workers aged 15 to 69 by potential occupational exposure to and complementarity with AI, March 2026,” reports three use bands for each group. In high-exposure, high-complementarity occupations, 67.6% of users reported moderate use and 23.2% minimal use; in high-exposure, low-complementarity occupations, the corresponding shares were 63.6% and 21.1%. Among users in low-exposure occupations, 50.7% reported moderate use and 39.3% minimal use. Broad use was a minority in all groups. Percentages may not add to 100 because of rounding.

The denominator matters: these figures compare task breadth only among people who said they had used generative AI at work during the prior year. They do not compare every worker in each occupational group or show how often a tool was used. “Most tasks” also does not mean most work hours, most valuable duties, or reliable completion without review. The categories are self-reported descriptions of task coverage, not an inventory checked against job records or observed workflows. The chart cannot tell us output quality or correction effort.

Statistics Canada classifies occupations with the complementarity-adjusted AI occupational exposure index, or C-AIOE. As described in “Use of generative artificial intelligence tools among Canadian workers, March 2026,” its exposure score represents the potential for AI to influence, transform, or perform an occupation’s tasks; its complementarity score represents the potential for AI to enhance or augment human labour. The release groups occupations using median scores. These are occupation-level categories, not individual assessments of a particular worker’s duties. Statistics Canada explicitly says the index indicates potential task transformation, not a direct prediction of job loss; actual effects depend on economic, institutional, and technological factors.

The pattern is compatible with different task mixes, workplace rules, or AI assisting some duties while leaving other work human-led. The chart does not test which explanation accounts for the differences. It cannot establish that broader use replaces labor, increases output, or changes hiring; those outcomes require evidence about employer decisions and work results.

For a worker, an exposure label can be a prompt to inspect recurring duties, not a forecast. Identify one task that actually appears in your week; note whether an approved tool is used, what it produces, what checking or correction remains, and who is accountable for the result. The cross-tab gives a reason to avoid treating occupational exposure and reported task breadth as interchangeable. Your own workflow, local opportunities, pay needs, and constraints still determine what response makes sense.

Sources: Chart 3 — Extent of generative artificial intelligence (AI) use across work tasks among workers aged 15 to 69 by potential occupational exposure to and complementarity with AI, March 2026; The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026

Where is workplace use concentrated—and what does that comparison add?

Reported use was uneven across both industries and broad occupational groups in the March 2026 Labour Force Survey supplement. Across industries, 65.6% of workers in professional, scientific and technical services reported using generative AI at work in the previous 12 months, compared with 59.2% in finance, insurance, real estate, rental and leasing, and 53.0% in educational services. At the lower end, reported use was 16.3% in accommodation and food services, 17.5% in agriculture, and 21.1% in transportation and warehousing. By broad occupation, the release found the highest rates among legislative and senior management workers (75.1%) and natural and applied sciences workers (67.5%); the lowest were among trades, transport and equipment operators (14.7%) and natural resource, agriculture and related workers (17.0%). These figures describe workers reporting at least one use during a year, not how often they used a tool or what share of their tasks it covered. (Statistics Canada, “Use of generative artificial intelligence tools among Canadian workers, March 2026.”)

The comparison adds a map of where reported use was more or less common; it does not sort jobs from safe to threatened. A broad industry groups people doing many kinds of work, and an occupational label describes the kind of work and its important duties rather than an identical daily task list. Within an industry or occupation, workers may handle different information, customer needs, physical processes, or accountability. The survey reports these as descriptive differences. It does not establish that an industry's tasks caused adoption, that a tool produced better work, or that an employer reduced staffing. Nor should the two sets of categories be treated as interchangeable: industry describes the employer's business, while occupation describes the respondent's work.

That distinction matters when translating a national pattern into a personal question. A role in a higher-use category may include recurring drafting, information synthesis, or analysis, but the rate cannot tell whether a particular worker performs those tasks, whether an approved tool is available, or how much checking the output requires. Conversely, a lower-use category can still contain a specific information task worth examining. A worker can use the distribution as a prompt: identify one recurring task involving text or information, then check whether it is actually part of their own week and whether AI enters its workflow. The individual task and the surrounding review, judgment, and responsibility are more useful evidence for a next step than the occupation-level average.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026

Why does non-use not automatically mean a skills gap?

Among Canadian workers aged 15 to 69 who had not used generative AI for their main job or business in the previous 12 months, 56.0% said the tools had no applicability to their work. By comparison, 5.8% cited a lack of skills or knowledge and 5.0% said company or organizational policy limited use. The March 2026 Statistics Canada release therefore does not support treating non-use as a simple training deficit. It records workers’ stated reasons; it does not independently test whether a tool could help with any particular person’s tasks. The denominator matters: these figures describe non-users, who represented 64.1% of workers, rather than all workers. The release also reports other reasons: 21.6% said they had no interest, while 9.8% cited security, privacy, environmental or ethical concerns. These are selected reasons, not a set of mutually exclusive shares to add into a single explanation. The broad pattern is that people may see no suitable task, choose not to use a tool, or face constraints that training alone would not resolve. A low reported skills barrier does not prove that nobody needs instruction; it does make “everyone who is not using AI needs a course” an unsupported conclusion. A person’s stated reason is a starting point for inquiry, not a complete diagnosis of their work or learning needs. Applicability is a judgment about fit with current work. Some duties depend on physical presence, specialized equipment, or accountable decisions; even digital tasks may involve sensitive information or costly errors. These are illustrations, not occupation-level findings from Statistics Canada. The questionnaire lists other barriers, including workflow integration and access, but does not test a proposed use. Policy and privacy limits deserve particular care. The release reports that 5.0% of non-users cited organizational policy; the survey does not say that every such rule is a permanent ban or that workers should work around it. The questionnaire separately offers security, privacy, or ethical concerns as a reason. For confidential information, first check current rules and approved tools. A prohibition, lack of approved access, or unclear responsibility for checking outputs can make non-use the appropriate choice while the worker seeks guidance. A proportionate next step is to name one recurring task and ask four questions: Is there an approved tool for it? What input can be shared? What output would be useful, and who checks it? Is a real skill gap preventing a permitted trial? If the answer points to a specific gap, short, task-focused learning or guided practice may help more than a broad credential; if the task is not applicable or use is disallowed, training may not change the immediate decision. The March survey cannot choose between those paths for an individual. It shows why the decision should begin with task fit, access, and rules, then address skills where they are actually the constraint.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026; Labour Market Indicators – March 2026 questionnaire

What does the earlier Canadian survey add—and what can’t it prove?

The earlier Canadian Survey on Working Conditions (CSWC) adds useful history: reported workplace generative-AI use was rising before the Labour Force Survey (LFS) supplement began measuring task breadth. Statistics Canada’s June 2026 release, “Workplace artificial intelligence use: A profile of sociodemographic and job characteristics, September 2024 to July 2025,” reports an average of 22% across four collection periods from September 2024 through July 2025. The share reporting use in the previous 12 months rose from 17% in September to 30% in July. These figures support a conclusion about growing reported reach, not how often people used AI, how many tasks it touched, or what the tools did to work quality or employment.

The question design is decisive. The CSWC asked whether a worker had used AI or automation technologies at work during the prior year; Statistics Canada says it captured whether use occurred, rather than intensity or frequency. Its 22% average is not a measure of regular use or task coverage. The March 2026 LFS supplement asked about generative-AI use in the main job or business over the previous 12 months, then separately asked users about current frequency and task proportion. The March release reports 35.9% for past-year use and introduces current task breadth. The two percentages may look like a rise from 22% to 35.9%, but they come from different surveys and designs. Differences in wording, timing, and context mean subtraction would not isolate change in use. The CSWC averaged four collection periods; March 2026 is the LFS supplement’s first wave. Statistics Canada planned further collection for September 2026 and regular repeats for trend monitoring. Until comparable waves accumulate, March is a baseline for its measures, not a trend line.

A separate Statistics Canada study, “Canadian employment trends in the era of generative artificial intelligence: Early evidence,” adds a boundary, not a forecast for this worker-use series. It examines employment trends and AI-related occupational groupings, distinct from self-reported use and task breadth. Neither the CSWC nor the March supplement measures an individual’s probability of job loss. A worker’s reported use also does not establish employer-wide adoption, autonomous task completion, or displacement. The narrower reading is that the earlier survey found growing past-year reach, while the newer supplement adds a first view of current frequency and task coverage among users. A worker can inspect a recurring task in their own workflow while waiting for method-consistent observations before treating a national difference as a trend.

Sources: The Daily — Workplace artificial intelligence use: A profile of sociodemographic and job characteristics, September 2024 to July 2025; The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026; Canadian employment trends in the era of generative artificial intelligence: Early evidence

What should a worker do with this first baseline?

The first baseline justifies attention to changing tasks, not a sweeping career decision. Choose one recurring task that you are permitted to handle with an approved tool. Write down what goes in, what the tool returns, what you must correct or verify, how long that checking takes, and who remains accountable for the result. Compare several real instances with the current method. This small comparison can reveal whether the tool removes a step, speeds up a draft while leaving review intact, or adds checking work. Statistics Canada’s March 2026 release measures workers’ reported use and task breadth; it does not measure these workflow details, output quality, review time, or independent task completion. Those distinctions matter more to your next move than a national use rate alone. Check workplace policy before entering work material, especially when it is sensitive or the tool is not approved. If the task is useful and the gap is specific, a short practice project or focused learning step may be enough to test an upgrade in your current role. If the task mix appears to be changing more broadly, compare adjacent roles or a larger change against actual local openings, your salary floor, location, available study time, health, and family responsibilities. Do not assume that non-use proves a skills deficit or that reported use predicts displacement. The free task checker offers change-pressure signals, not validated odds of job loss; the paid career roadmap compares scenarios and constraints without guaranteeing employment or income. Start with the task you can observe. Let evidence from that workflow, alongside your real options and limits, determine whether any larger step is warranted.

Sources: The Daily — Use of generative artificial intelligence tools among Canadian workers, March 2026

Questions readers ask