A Bank of Canada staff paper published in June 2026 found that 70 percent of surveyed business leaders use AI in a typical work week, while only 8 percent of their firms reported significant use of AI in producing goods or delivering services. The gap between a manager's own habits and a company's operations is the main reason claims that AI is transforming business need a second look at what was counted.
Data preparation is one place where AI projects depend on older software. One example is Microsoft SQL Server Integration Services (SSIS): SSIS is a data integration tool by Microsoft that moves and transforms data between systems before analysis.
Key takeaways
- In the Bank of Canada's December 2025 Business Leaders' Pulse, 70 percent of respondents used AI at work in a typical week, mostly for under five hours.
- For operations, 8 percent of firms reported significant use, 21 percent moderate use, and 29 percent infrequent or experimental use.
- Among firms with no plans to adopt, the most common reason was that AI is not useful to the business, ahead of missing skills and software incompatibility.
- Over three years, 18 percent of firms expected AI to reduce their headcount and 9 percent expected it to add staff.
- The survey has 314 responses and is not built to represent all Canadian firms, so it describes a direction and not a national rate.
What the Bank of Canada asked, and of whom
The staff analytical paper "Survey Evidence on Firm AI Adoption and its Implications" by Chanya Chawla and Crystal Arnburg adds special AI questions to the bank's December 2025 Business Leaders' Pulse. The sample was 314 responses. The Pulse recruits firms by regional and industry quotas tied to shares of economic output, and the authors say it is designed to inform the assessment of aggregate conditions and not to produce population-representative estimates of firm behavior. Firms with fewer than 20 employees are over-represented relative to their contribution to output, and they made up 41 percent of this sample. The bank's AI definition was broad: text generation with large language models such as Microsoft Copilot, data or image processing with machine learning, and visual content creation.
The authors note that their paper reflects the views of its authors and not the bank's Governing Council.
Personal use versus use in production
Two questions produced two very different pictures. Asked how often they personally use AI in a typical working week, 70 percent of leaders said at least some time, and most said less than five hours. Asked how far the firm uses AI to produce goods or deliver services, the answers spread out as shown below.
| Level of use in operations | Share of firms |
|---|---|
| Significant use | 8 percent |
| Moderate use | 21 percent |
| Very infrequent, experimental, or pilot use | 29 percent |
| Not in use, plan to start within 12 months | 11 percent |
The remaining firms either did not use AI and did not plan to, or did not know. The paper summarizes that more than half of firms either have no use and no plans, or use AI infrequently. Manufacturing shows the pattern clearly: about one quarter of manufacturers reported moderate or significant use, while over a third used AI infrequently or experimentally. Micro firms had a larger share with no current or planned use, about two fifths against about one fifth of firms with 20 or more employees, but the paper says that difference came mainly from fewer experimental users and not from fewer serious ones.
What firms use it for, and why others stay out
Text generation with large language models was the most cited current use. Roughly a quarter of firms also reported visual content creation and data processing with machine learning. Robotics and autonomous vehicles were among the least common. Looking three years ahead, an additional quarter of firms expected to use machine learning for data processing, and the expected rise in robotics came mostly from manufacturers.
Among firms that neither use nor plan to use AI, a majority said the technology was not useful to their operations. Missing skilled staff and incompatibility with existing hardware or software came next. For a business owner, that answer is a recorded and common position among firms in the sample.
What firms expect it to do to spending and jobs
Expectations were modest in the short run and larger further out. Over the next 12 months, 31 percent of firms expected no AI effect on capital spending, compared with 16 percent over three years. Over three years, about 20 percent expected a minor positive effect on capital spending and 10 percent a major one, against 3 percent and 1 percent expecting minor or major negative effects.
For employment, about 11 percent of firms expected declines and about 8 percent increases over the next 12 months. Over three years the figures were 18 percent and 9 percent. Among small, medium, and large firms, the share expecting declines roughly doubled from 10 percent to 21 percent between the two horizons. These are expectations from managers, and the paper treats them as such. It does not report realized job losses.
Why published adoption rates disagree
The paper itself lists the spread. Statistics Canada's Canadian Survey on Business Conditions found about 15 percent of firms using AI for production purposes. KPMG Canada's generative AI survey reported adoption above 90 percent. A study of firms in the United States, United Kingdom, Germany, and Australia, cited in the paper, found around 70 percent using at least one AI technology, and McKinsey's 2025 global survey reached nearly 88 percent of organizations. The authors attribute the range to differences in survey timing, question wording, sample makeup, and what counts as adoption. A rate quoted without its question is close to meaningless, and two rates can both be accurate while differing several-fold.
A practical reading for a business weighing a tool: ask whether a figure counts personal experimentation, pilots, or work embedded in a production process. The Bank of Canada's own conclusion is that adoption among Canadian firms remains at an early stage, concentrated in experimentation and task-level uses.
FAQ
Does the 70 percent figure mean 70 percent of Canadian businesses use AI?
No. It is the share of surveyed decision-makers who personally use AI at work in a typical week. Use inside the firm's production process was much lower.
Is the Bank of Canada survey a national statistic?
Not in the sense of a census-style estimate. The paper says the Business Leaders' Pulse is designed for assessing aggregate economic conditions and is not meant to give population-representative estimates of firm behavior.
Why do some firms say AI is not useful to them?
The survey does not explain the reasoning behind that answer. It only records that it was the most common reason among firms with no plans to adopt, followed by lack of skilled staff and incompatibility with existing systems.
Do firms expect AI to cut jobs?
Over three years, 18 percent expect to hire fewer people because of AI and 9 percent expect to hire more, with the rest expecting no effect. The paper describes the net expectation as modestly negative.
Related reading
- which AI tool types Canadian businesses report using
- what field experiments measured on AI and costs
- FTC cases about AI sales claims
- free AI libraries and their licenses
More in AI Tools and Industry.
Last verified: October 2026
Sources
- Chawla C., Arnburg C., Survey Evidence on Firm AI Adoption and its Implications, Staff Analytical Paper 2026-22, Bank of Canada, June 2026