In October 2023, Imperial Oil reported that its Kearl oil sands mine had “completed its multiyear program to convert its 81 haul trucks to autonomous operation.” For that sentence to be true, a lot had to go right. The trucks had to run in production rather than in a demonstration, through northern Alberta winters, hauling ore that pays the bills, with the savings landing somewhere a finance department could find them.
Imperial didn’t sell it as artificial intelligence. It counted the trucks. That’s the standard the AI pitch to Alberta’s builders and growers should be held to, and so far most of that pitch hasn’t met it.
Key facts
- Statistics Canada reported on May 27, 2026, that 19.2% of Canadian businesses used artificial intelligence to produce goods or deliver services in the 12 months before the second quarter of 2026.
- A Statistics Canada analysis published June 11, 2026, found that 9.2% of construction businesses and 4.5% of agriculture, forestry, fishing and hunting businesses reported that use.
- Statistics Canada reported on May 27, 2026, that 40.0% of Canadian businesses said artificial intelligence was not relevant to their goods or services.
- Imperial Oil reported on Oct. 27, 2023, that its Kearl mine had converted 81 haul trucks to autonomous operation.
The optimistic case for AI adoption has real numbers behind it
Start with the strongest version of the other side, because it’s good. Across all industries, the share of Canadian businesses using AI to make goods or deliver services went from 6.1% in the second quarter of 2024 to 12.2% a year later and 19.2% in the second quarter of 2026, according to Statistics Canada. That’s roughly a tripling in two years. On that curve, construction at 9.2% looks late rather than unwilling, and late buyers often get a more finished product at a lower price.
Geography supports the same reading. Urban businesses reported 21.0% use and rural businesses 9.9%. Adoption tracks distance from the vendors, which makes it a sales problem more than a character flaw.
That much of the optimistic case holds. The trouble starts with what the late adopters say about why they’re late.
Construction and farm businesses are telling vendors the product doesn’t fit
Start with what those businesses say. In the same survey, 40.0% of Canadian businesses said AI wasn’t relevant to what they make or sell. A contractor who says that has usually looked at what’s on offer and can’t find the line on an estimate it would change.
Productivity research points the same way. A Statistics Canada analysis published in April found that the productivity edge of AI-using firms fell to 5.1%, and stopped being statistically significant, once the researchers accounted for the other capabilities those firms already had. The agency’s own summary is that “AI adoption in isolation is likely insufficient.”
Employment data adds the last piece. Among businesses using AI in 2025, 89.4% reported no change to employment. The tools aren’t replacing crews. They have to earn their place the harder way, by saving hours someone can count.
Put those together and the gap between 19.2% and 9.2% says more about the product than the customer. Too much of the pitch still runs on the announcement and the pilot. An estimating desk is a throughput problem, bids in and wins out, and a tool that cuts hours per bid or lifts the win rate can be checked against last year’s numbers in an afternoon. Very little of what gets demonstrated to builders is sold that way.
AI vendors should price their tools by the hours they save
Kearl worked because the measure was physical, counted in trucks and loads. The same discipline is available to anyone selling into a machine shop or a grain operation. Price the tool against a number the owner already tracks, and stand behind it.
Here is a prediction that can be checked. If vendors keep selling to Alberta contractors on demonstrations instead of hours saved, construction will still sit below half the all-industry AI rate when Statistics Canada publishes its second-quarter 2027 figures next spring.
