Tony Frost and Christian Dippel are associate professors of business economics and public policy at Western University’s Ivey School of Business.
Could artificial intelligence be the silver bullet for Canada’s chronic productivity woes? Techno-optimists think so. Policymakers certainly hope so. But the answer does not depend on research labs, frontier models, or a few large corporations that control many sectors of Canada’s economy. It will determine whether small businesses can actually adopt these tools and use them to modernize their operations, enter new markets, and innovate.
Small and medium-sized enterprises (SMEs) employ approximately two-thirds of Canada’s private sector workforce and generate nearly half of private sector GDP. Unless productivity in this sector increases, national productivity will not increase either.
Canada’s current AI debate feels out of place. The productivity gains that make a real difference will come from the ordinary businesses that make up the bulk of the economy: manufacturers, distributors, construction companies, professional service providers, and retailers. Many of these companies have never hired data scientists or built AI capabilities in-house.
Companies with 10 to 300 employees may be the most important targets for AI transformation. Many are sophisticated in their field, integrated into North American supply chains, meet rigorous quality standards, and operate with real discipline. However, data is rich but systems are often poor. They store years of sales and production data, inventory records, purchase history, compliance logs, and customer information, much of it on older platforms and used primarily for record-keeping and transaction processing. AI holds the promise of turning that stuck data into something far more valuable.
As is the case for countless struggling small businesses, business knowledge is locked away in emails, spreadsheets, and the heads of a few veteran employees. Estimating, sourcing, and scheduling often rely on manual workarounds, making it difficult to scale and sell beyond your core customer base or geographic region. These companies struggle to build tailored solutions for their customers because they don’t have a way to systematically leverage historical data.
For these companies, AI promises more than just efficiency. It’s the ability to operate like a much larger company without building a massive back office. Instead of relying on a small number of experts to price complex custom orders, often at a snail’s pace, AI-enabled quoting tools can instantly tap into past contracts, supplier input, and production constraints, allowing companies to respond faster, win more business, and expand into new products and geographic markets with confidence. When applied effectively, AI can reduce fixed costs associated with growth.
However, the companies that would benefit most from AI are often the ones most hesitant to adopt it. That hesitation is often mistaken for conservatism or lack of ambition. In our experience, this is usually a reasonable response to uncertainty and downside risk. Unlike large companies, most small businesses don’t have in-house teams to experiment, absorb failures, and distinguish between real technological advances and impressive vendor claims. For them, implementing AI doesn’t seem like a series of small bets. It feels like a leap into the unknown.
The problem involved is similar to the classic “lemon problem” in economics. Companies that hire their 10th software engineer already know what good looks like. However, this is not the case for the first employer. You may not know how to screen candidates, define suitable roles, or even evaluate whether the output is truly useful. One bad hire can derail the entire effort.
And risk is not the only issue. Many small and medium-sized businesses still don’t have a clear understanding of what AI can actually do for their business, and AI is framed too narrowly as a back-office efficiency tool rather than a way to redesign how companies quote, schedule, sell, interact with customers, develop products, and expand into new markets.
The predictable result is a delay. Many small businesses say they are waiting for AI tools to become cheaper, simpler, or more “proven.” Some companies adopt simple off-the-shelf solutions and conclude that they have “made AI happen.”
Small and medium-sized businesses don’t need a national AI strategy; they need practical initiatives to reduce the risk of getting started with AI. In most cases, the first step is not an advanced application. They get the fundamentals right: understanding what data a company already has, where it resides, and which operational decisions can realistically be improved. This kind of fundamentals are not sexy, but they are what transform AI from a buzzword into a practical feature.
A further challenge is that many small and medium-sized businesses cannot avoid the risks of this transition on their own. They lack the expertise to scope pilots, evaluate vendors, hire technical talent, and translate AI tools into a coherent operational or growth strategy. For them, AI adoption does not seem like a series of manageable experiments. It feels like a high-stakes gamble.
Paradoxically, at the very moment that small and medium-sized businesses are struggling to meaningfully implement AI, the market for entry-level junior software and AI talent is softening as large employers and tech giants ramp up hiring and AI begins to replace some traditional junior jobs. Young engineers want real problems and responsibilities, and small businesses are full of them. But the bridge between them is missing.
Ultimately, Canada’s AI future will be won on the manufacturing floors, warehouses, and back-offices of the hundreds of thousands of small and medium-sized businesses that make up the backbone of Canada’s economy.
This column is part of Globe Careers’ Leadership Lab series, where executives and experts share their insights and advice about the world of work. To find all Leadership Lab stories, please visit: tgam.ca/Leadership Lab and guidelines on how to submit to columns. here.
