Adoption by midsized business will be AI’s biggest economic impact in Canada
"The most important question isn't what AI can do, but what we can do with AI."
Kathryn Hume, Vice-President, AI Engineering
Vector Institute
Canada’s economy is highly dependent on its small and medium-sized enterprises. Like all businesses, these companies are always on the lookout for ways to boost their efficiency and productivity by automating tedious tasks and adopting cutting-edge innovations. Until recently, that’s been accessible only to the largest companies with the deepest pockets. But now AI is leveling the playing field.
Kathryn Hume leads the AI engineering team at the Vector Institute, translating frontier research into practical, production-grade AI. With a career spanning RBC, Tangerine and Borealis AI, she has spent over a decade making AI work not just in demos, but inside some of the most complex regulated environments in North America. She sees AI giving smaller businesses the opportunity to transform from the inside out - democratizing access to innovation that was previously available only to organizations with deep pockets and dedicated data science teams.
"Before, you'd have to convince someone your idea was worth a try and wait for your team to have bandwidth. What's changing now is that the distance between an idea and a working proof of concept has collapsed. That changes who gets to participate in innovation – and that matters enormously for Canadian business."
Sometimes, that might not even look like a formal pilot. Hume says any employee can think of something that would make their job easier and then use AI to build it for themselves, acting as their own tester and proving the solution’s viability just by using it. While some solutions might be useful just for the person who created them, others might get rolled out to the rest of the company or even become external product offerings.
"The talent challenge isn't just about hiring data scientists. It's about building organizations where people with deep domain expertise can work productively alongside AI systems - and where your most experienced people are empowered, not sidelined, by the technology."
The technical foundations of successful AI adoption
While getting started is easy, not every AI implementation will ultimately be successful. Helping companies isolate and scale the successful AI projects is a big part of what Vector does. Through her team’s work at Vector, Hume has observed four technical foundations that can make the difference between a company that sees great results from AI adoption and one that struggles to achieve the returns they’d like:
Data strategy: From records to action
All AI models run on data, so your data foundations have to be solid to gain the benefits. Areas where you’ve already been collecting and maintaining high-quality data are often good candidates for AI solutions. But you can also look for ways to collect new data or combine different data sets to enable new insights and new tools.
Those new combinations are also underpinning a shift away from “systems of records” and toward “systems of action.” An example of a system of records is a customer relationship management system that just logs and stores data for you to use later. A system of action is layered on top of the system of records, proactively analyzing that data and feeding it into your systems and workflows so you can act on it and make data-informed decisions.
“These systems can take information from your team's Slack channels, your codebases and your documents and, based on the interactions between them, provide you with a status update,” says Hume. “You might still need to ask your team to confirm the update. But it’s much faster than asking them to go through all the information to provide you with the answer. That leaves them more time to do their regular work.”
A strong data foundation is what enables this shift. You need to understand the data you have, know where your gaps are, and have a plan to fill those gaps and keep your databases clean and current.
An experimental sandbox
If you want people to experiment and try new things, they need a safe place to do it. In other words, something that is segregated from your actual production/operational systems, but behaves just like those real environments. That gives employees a way to test their AI-powered creations without risking your mission-critical data or processes.
A sandbox environment also doubles as a training ground. The skills employees build during their experiments are what they’ll need when it comes time to implement their solutions in your production environment.
The right mix of talent
“There’s an interesting decoupling of skills and experience happening right now,” says Hume. “People with decades of experience in their fields have highly valuable expertise, but they don’t always have the specialized AI skills that new graduates bring to the table.”
That doesn’t mean discounting experience, but it can mean looking externally to bring in specialized AI experts who can help take your rough pilots and scale them into revenue drivers. And yet, competing for that talent with major players like Google and Meta means you likely can’t rely entirely on new hires even if you wanted to. Upskilling the people already in your organization is also highly effective. It essentially turns them into data scientists who also have deep expertise in your field. Combining these sources will produce complementary skillsets that will drive more innovation and learning in a self-reinforcing cycle.
Technical leadership
Business and technology are becoming more entwined than ever, with nearly every company becoming a tech company to some degree. To make the best decisions, senior and executive leadership teams need to include technical expertise like engineers who can help direct the company’s technology-related innovations and development.
“Taking the risk away gives people the confidence to try bigger and bolder things. That ends up enhancing the culture of experimentation across your whole organization. That’s how we start to see real wins.”
Fostering Canadian innovation
With the right technical foundations and an innovative mindset, Canada’s small and medium-sized enterprises can build AI solutions that may help Canada extend its AI leadership from research to commercial markets.
Through its work with Bell and other partners, Vector offers a variety of support to help businesses do just that. One example is their Machine Learning Associates program, which pairs recent graduates with critical AI skills and small businesses that need those skills for a specific project. Hume says the program has been highly successful so far, with most participants successfully commercializing their products and many hiring the graduate on full-time after the project is over.
"Canada's persistent AI adoption gap is well documented; recent data from Statistics Canada confirms that only 12 per cent of Canadian businesses have integrated AI into their operations. While usage doubled from 2024 to 2025, it remains significantly below G7 and OECD averages despite enormous interest,” says Hume.
“Closing that gap is one of the most important things we can do for Canadian competitiveness right now and what we think the AI for all strategy is about. That's what drives this work."