Adopted without training.
Small Canadian firms took up AI at nearly the rate large ones did. They trained for it at roughly a third of the rate.
CometX · September 2026 · Free to reuse under CC BY 4.0
The number that matters
19.9% vs 24.0%
Firms with one to four employees used AI at 19.9% — close to the 27.8% of firms with 100 or more. But only 24.0% of the small adopters trained an employee to use it, against 68.1% of the large ones.[1]
Adoption is no longer the story
In the second quarter of 2026, 19.2% of Canadian businesses reported using AI to produce goods or deliver services in the preceding twelve months. That is triple the 6.1% recorded when the question was first asked in the second quarter of 2024,[2] and up from 12.2% in 2025.[1] Two in five businesses (40.0%) still say AI is not relevant to what they do.[1]
The direction was signalled a year earlier: in the third quarter of 2025 a further 14.5% of businesses said they expected to adopt AI within twelve months, with information and cultural industries most likely to say so.[3]
The interesting finding is not the headline. It is that the gap between the smallest and largest firms in whether they use AI is remarkably narrow — and the gap in how they use it is very wide.
| By employee count | 1 to 4 | 100 or more |
|---|---|---|
| Used AI in the last 12 months | 19.9% | 27.8% |
| Say AI is not relevant to the business | 41.4% | 21.3% |
| Trained existing employees on AI | 24.0% | 68.1% |
| Trained existing executives on AI | 15.6% | 51.7% |
| Hired for AI-related skills | 8.2% | 32.8% |
| Used external consultants or vendors | 10.7% | 30.2% |
Source: Canadian Survey on Business Conditions, second quarter of 2026. Training and hiring rows are among businesses that reported using AI.[1]
A firm with one to four employees is roughly seven-tenths as likely as a large firm to be using AI — but only about a third as likely to have trained anyone to use it, and a quarter as likely to have hired the skill or brought in help.[1]
Why that matters more than the adoption number
Statistics Canada's own summary of firm-level research is direct on this point: AI adoption is associated with productivity gains, but the relationship depends on complementary capabilities inside the firm rather than on the technology alone.[4] Adoption without the surrounding capability is the condition under which tools get bought and quietly stop being used.
This matches what we found in our own review of enterprise AI: the blockers are rarely the model. They are data quality, workflow rigidity and the absence of any agreed way to measure whether the thing worked. The End of the AI Pilot sets out that evidence at the enterprise scale. The 2026 business-conditions data suggests the same gap exists at four employees as at four hundred — with far less capacity to close it.
What they are actually using it for
Among businesses using AI, the most common applications in 2026 were data analytics (36.6%), text analytics (34.5%) and virtual agents or chat bots (28.2%). Two categories moved sharply in a single year: decision-making systems rose from 5.7% to 13.7%, and deep learning from 6.6% to 13.1%.[1]
Among the smallest firms the pattern is more modest and more practical — data analytics (32.5%) and text analytics (29.4%) lead.[1] This is not a picture of small businesses automating themselves away. It is a picture of them using AI for reading and summarising, which is exactly the use most sensitive to whether anyone has been taught to check the output.
Two divides worth naming
Geography. Urban businesses reported 21.0% AI use against 9.9% for rural businesses. Both have roughly tripled since 2024, but the absolute gap has widened from 3.4 points to 11.1.[1] A proportional gain that leaves the absolute distance larger is worth stating plainly.
Age of firm. Businesses more than 20 years old were the least likely to use AI (15.0%), against 25.0% of firms three to ten years old.[1] The established firm, not the small one, is the laggard here.
What stops them
Asked what limits AI use, businesses named cybersecurity or privacy concerns most often (13.4%), followed by cost (10.6%).[1] Both vary by size in a way that cuts against the intuitive story: cybersecurity concern was reported by 30.0% of firms with 100 or more employees against 11.6% of those with one to four.[1]
Read carefully, that is not evidence that small firms face fewer risks. It is evidence that they are less likely to have someone whose job is to notice them.
What this report does not establish
- These are self-reports. Businesses were asked whether they used AI, not tested on it. Definitions of "using AI" almost certainly vary between a four-person shop and a bank.[1]
- No causal claim. Nothing here shows that training causes better AI outcomes. The productivity link comes from separate firm-level research with its own limitations.[4]
- Barrier figures are percentages of all businesses, not of AI users, so they understate how often adopters hit a given obstacle.[1]
- One quarter, one country. The survey ran from 1 April to 6 May 2026, with 9,251 responses from a sample of 21,105.[1]
Sources
Every figure above comes from one of these. All were retrieved and read in September 2026, and each link goes to the primary publisher rather than to a summary.
- 1Statistics Canada. Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026. Catalogue 11-621-M, released 11 June 2026. Canadian Survey on Business Conditions, 9,251 responses. ↩
- 2Statistics Canada. Analysis on artificial intelligence use by businesses in Canada, second quarter of 2024. Catalogue 11-621-M. The first iteration of these questions. ↩
- 3Statistics Canada. Analysis on expected use of artificial intelligence by businesses in Canada, third quarter of 2025. Catalogue 11-621-M, released 11 September 2025. ↩
- 4Statistics Canada. Artificial intelligence adoption and productivity in Canadian firms. Economic and Social Reports, summarising Li and Liu (2026), Canadian Public Policy. ↩