The government’s national AI strategy is largely framed around the notion that Canada has an AI adoption problem. At the launch last month, Prime Minister Mark Carney said that “only 12% of Canadian businesses are using AI today” and presented the strategy as a plan to address concerns that Canada lags behind other countries. AI Minister Evan Solomon echoed the same issue and put a specific number on it, targeting an increase from 12 per cent to 60 per cent. The AI adoption issue helped justify billions in spending and the suite of new legislative reforms. I’m supportive of many of the measures, but a closer look at the statistics and comparisons the government used shows that it relied on outdated data and dubious comparisons. In fact, Statistics Canada had actually released new data on business adoption of AI days before the strategy’s release that placed it at 19.2%, yet the government instead pointed to the older figure of 12 per cent, which made a stronger case that adoption was lagging and that new government support was needed.

The Statistics Canada data comes from the Canadian Survey on Business Conditions, which shows AI adoption growing from 6.1 per cent in 2024 to 12.2 per cent in 2025 to 19.2 per cent in 2026. That places Canada roughly in line with other peer countries. For example, the OECD reports that firm-level AI use across member countries with available data increased from 8.7 per cent in 2023 to 14.2 per cent in 2024 and 20.2 per cent in 2025. Eurostat places use among EU enterprises at 19.95 per cent in 2025. While those figures are not directly comparable with the Canadian survey given differences in the size of enterprises surveyed, it suggests Canada is not the laggard some suggest.

Relying on survey data about AI knowledge and public attitudes, the Canadian national strategy claims that low literacy and low trust are the “binding constraints” on deeper AI integration. But the latest Statistics Canada survey found that two in five businesses regarded AI as irrelevant to their operations. In an earlier survey, 78.1 per cent of businesses with no adoption plans cited lack of relevance, while privacy and security concerns only registered at 8.1 per cent. The story the data tells is that the problem for most non-adopters was not a lack of trust, but the absence of a convincing use case. The Canadian sectoral results reinforce the point. AI use reached 42.3 per cent in information and cultural industries and 40.4 per cent in finance and insurance, compared with 4.5 per cent in agriculture, 7.9 per cent in wholesale trade and 9.2 per cent in construction. In other words, adoption rates are far higher among businesses that have more obvious use cases and lower among those that don’t.

The headline numbers also mask the distinctly different outcomes depending on the AI use question. The Canadian Federation of Independent Business found that 45 per cent of businesses use generative AI at least occasionally to complete tasks, and KPMG reported that 93 per cent of surveyed business leaders said their organizations were using AI in some form, up from 61 per cent a year earlier, and that more than half of Canadian workers now use generative AI on the job. Those numbers do not contradict the 19.2 per cent so much as measure different things. Statistics Canada asked if the business was using AI to produce goods and deliver services, which may not capture occasional use, experimentation, or partial deployment.

Canada was also portrayed as an international laggard in a misleading way. As I noted above, the updated data shows Canada largely in line with other OECD countries. But beyond that data point, the comparison the strategy relies on most heavily sets 8 per cent adoption among Canadian small and medium-sized businesses against Nordic leaders at 29 to 42 per cent, Germany at 26 per cent, and France at 18 per cent. However, Canada has other data points that suggest SME adoption is higher, and the comparisons are apples to oranges, since the European numbers feature all-enterprise rates drawn from a survey that counts only firms with at least ten employees and includes large firms with higher adoption rates, while Canada’s survey includes businesses with one to four employees, who are less likely to adopt at this stage. In fact, the G7 and OECD sources for the European figures carry an explicit warning that survey methodologies vary by country and that comparability is limited and should be approached with caution.

Meanwhile, comparisons with the U.S. are affected by differences in definitions, as U.S. data illustrates how quickly a change in wording can change the apparent adoption rate. The U.S. Census Bureau broadened its survey in November 2025 to ask about AI use in any business function, and American adoption rose from single digits to between 17 and 20 per cent, a point The Hub made in comparing the two countries.

The government may have been working with the best data available while the strategy was being drafted, but by the time it was released, the yardstick had moved. That yardstick is itself being shut down as Statistics Canada has placed the Canadian Survey on Business Conditions on its list of discontinued programs, and its successor, the TechStat program, is not due to publish business data until 2027, which leaves the government’s 60 per cent target pegged to a survey that will no longer exist. The new numbers do not establish that Canada has solved every adoption or productivity challenge, but they confirm that adoption is moving far faster than the government’s narrative admits and that the real policy question is increasingly shifting from whether Canadian businesses will use AI to whether they can turn rapidly expanding use into meaningful gains. Canadian AI policy should begin with that question, not an adoption crisis that the government’s own data has already overtaken.