A Business Commentary by: Daryl Ching
There is a specific fear moving through Canadian small business right now, and I have watched it work on a room full of owners. It is not fear of spending too much. It is the fear that a competitor is buying artificial intelligence, pulling ahead, and that by the time the gap is visible it will be too late to close.
The room was full of trades businesses. HVAC, plumbing, electrical. People who have bought equipment for decades without needing anyone’s help. The question of whether the next dollar should go into another service van or into technology had them stuck, and the fear was doing the sticking.
It shows up in the national numbers as two opposite mistakes.
Some owners freeze. The Canadian Federation of Independent Business estimates private investment contracted 6.3 per cent last quarter and will shrink a further 4.7 per cent this quarter, even as its own Business Barometer put small business confidence at a five-month high in July. Owners feel better and are spending less. Waiting feels safe when you cannot tell which purchase is right.
Others leap. A BDO report found that nearly half of Canadian firms experimenting with AI report no meaningful return. RSM found Canadian firms trailing U.S. peers on payback. That is what buying something you cannot evaluate looks like at national scale.
Two errors, one cause. Owners cannot underwrite the decision. A van has a known payback, a resale market, a lender who will finance it, a replacement cycle you can predict. An AI platform has none of that, and it carries two questions nobody can answer for you. Will it do what it promises? And how soon will you be buying it again?
So the fear fills the gap where analysis should be. And here is what makes this worth writing about: Statistics Canada has now tested what the fear assumes, and the assumption fails.
The fear assumes that the competitor who adopts AI pulls ahead because of the adoption. In April, Statistics Canada researchers published a study linking federal technology surveys to business microdata to measure exactly that. At first glance, the fear looks justified. AI adopters showed 16.8 per cent higher labour productivity than non-adopters.
Then the researchers controlled for how productive those firms were before they adopted, and the premium fell to 10.2 per cent. Stronger firms adopt first. Then they controlled for what else those firms had built, including data analytics, cloud computing, trained staff, and research capability. The premium fell to 5.1 per cent and stopped being statistically significant. Read that carefully, because it is the answer to the fear. The adopters were not winning because they adopted.
They were already winning, and adoption is what winning looks like from the outside.
The advantage lives in the boring capability underneath. Firms that know their own numbers, keep clean data, and train their people. The purchase is the visible part of an invisible advantage, and buying the visible part does not confer it.
The study comes with honest limits. It draws on survey data from 2019 and 2021, before generative AI reached most businesses, and its authors caution that gains take time and depend on organizational change. This is not evidence that AI does not matter. It is evidence that adoption alone does not.
That distinction matters for small firms more than anyone. Amazon spent more than US$100 billion on technology and infrastructure last year, better than a tenth of its revenue. At that scale you place 20 bets and let 15 fail. A plumbing contractor makes one bet, funded from the same pool as payroll, the van, and the leasehold improvements. Small firms cannot experiment their way into the underlying capability. Which is exactly why they should not be racing to buy the badge of it.
The competitive question an owner should ask is not “What did my competitor buy?” It is “How much of what I already own sits idle?” That means utilization, in the plainest sense. What share of technicians’ hours are billable, and how many hours a week the equipment actually works. Some of the owners in that room could not tell me, and it is the single figure that determines whether they need more capacity or better use of the capacity they have. An owner who knows that number is better prepared for AI than an owner with a subscription and no idea.
The race is real. It is just not a purchasing race.
There is a policy implication sitting in plain sight. Ottawa’s new AI strategy commits billions toward raising business adoption of AI from 12 per cent today to 60 per cent by 2034. That target counts purchases. Statistics Canada’s own researchers have shown that purchases, taken alone, do not move productivity. A strategy that hits 60 per cent adoption without building the capability underneath will have produced the BDO statistic at national scale, and called it success.
I run an accounting firm, and bookkeeping, one part of what we do, has been commoditized for as long as I have worked in it. Won on price, with little room to innovate. Even that business now has to look hard at AI or risk obsolescence. So I am inside this fear, not above it. What the evidence has convinced me of is that the owners who come through it will not be the ones who bought fastest. They will be the ones who could measure what they bought.
*Daryl Ching, CFA, is the founder of Vistance Accounting and speaks regularly to CEO peer groups across Canada.*
