What Canada needs to get right in its AI strategy: DeepDive

DeepDives is a bi-weekly essay series exploring key issues related to the economy. The goal of the series is to provide Hub readers with original analysis of the economic trends and ideas that are shaping this high-stakes moment for Canadian productivity, prosperity, and economic well-being. The series features the writing of leading academics, area experts, and policy practitioners. Canada is at a strategic inflection point in its approach to artificial intelligence. The federal government’s consultation on defining the next chapter of Canada’s AI leadership—part of a broader process launched in fall 2025 to update the national AI strategy—is inviting input from Canadians on everything from research and commercialization to infrastructure, workforce skills, governance, and public trust. This conversation could not be more timely. Artificial intelligence is already reshaping productivity, labour markets, business models, and national competitiveness. Countries that successfully integrate AI into their economies are likely to enjoy faster growth, stronger firms, and higher living standards. Those that do not risk falling behind. Canada’s challenge Canada enters this moment with significant strengths but also persistent vulnerabilities. The country has long been recognized as a pioneer in AI research and early public policy. Yet it continues to struggle to translate that early leadership into widespread commercial adoption, scalable innovation, and broad-based economic impact. That challenge is not only about capital and commercialization, but how AI is used and harnessed by Canadians themselves in the workplace. In practice, early evidence from Canadian industry points in a positive direction; that AI adoption is already augmenting rather than replacing jobs. Most Canadian firms using AI report no net reduction in headcount, while many report productivity gains and new roles emerging. One of the most important and often underappreciated levers for closing this gap is the breadth of AI development and deployment models available to Canadian firms and institutions. It would be a dereliction of opportunity to tie Canada’s economic future to any single AI model or licensing approach; proprietary systems, open models, and hybrid approaches are all driving real-world impact today. The real strategic challenge, therefore, is how the government can leverage—and sequence—different approaches within a diverse AI ecosystem to ensure that AI adoption is widespread, competitive, and resilient. A new report published by the Linux Foundation argues that open source AI matters in the Canadian context because it shapes who can participate and how fast AI capabilities diffuse. According to the Foundation, more than a single model choice, “openness” can function as a guiding philosophy that prioritizes experimentation, skills development, and organizational control at the foundation of AI adoption, which only begins with open source models. Within this framework, managed and commercial AI systems can then build on those foundations, providing the reliability, performance, and infrastructure needed to scale successful use cases across sectors and markets. Canada’s AI legacy  Canada’s position in the global AI ecosystem did not emerge by accident. The country was an early mover in recognizing the importance of artificial intelligence as a general-purpose technology. It launched the world’s first national AI strategy in 2017, supported by substantial public investment and institutional coordination. That strategy helped anchor a network of globally respected research institutions, including the Quebec Artificial Intelligence Institute (Mila) in Montreal, the Vector Institute in Toronto, and the Alberta Machine Intelligence Institute (Amii) in Edmonton. Together with the Canadian Institute for Advanced Research (CIFAR), these institutions have helped make Canada home to roughly 10 percent of the world’s leading AI researchers and placed the country among the top five globally for highly cited AI research. Canada has also performed well on several indicators of AI investment. It consistently ranks near the top of global comparisons for venture capital investment in AI startups and for patent filings related to AI technologies. Recent federal budgets have reinforced this trajectory, committing billions of dollars to AI infrastructure, compute capacity, and adoption programs. And yet, despite these strengths, Canada has struggled with a familiar problem: converting research excellence into broad economic returns. Across multiple surveys and international benchmarks, Canada lags peers in AI commercialization and enterprise adoption. Only about one-quarter of Canadian firms report having fully implemented AI solutions, compared to roughly one-third globally. Many Canadian companies remain stuck in pilot projects or early experimentation, unable to scale AI across core business functions. This “research-to-market” gap has real consequences. Without widespread adoption, AI’s potential productivity gains remain concentrated in a narrow slice of the economy. Without scalable commercialization pathways, Canadian startups struggle to grow into global champions. And without sufficient infrastructure and skills diffusion, public investments in research risk generating spillovers elsewhere rather than at home. The challenge facing Canada, then, is not whether it has AI capability—it clearly does—but whether it can deploy that capability widely, responsibly, and at scale. Canada’s choices—and why they matter Canada’s AI ecosystem is shaped by a range of development and deployment models. These include open models that allow inspection and modification, proprietary systems offered through commercial platforms and APIs, and hybrid approaches that combine open components with managed services, licensing, or sector-specific customization. The strategic choice is not between “open” or “closed,” but whether Canada can build an ecosystem that captures the benefits of each to meet the needs of its unique strategic and economic profile. Proprietary systems have a place and a purpose; however, open approaches support experimentation, customization, and learning. Hybrid models increasingly allow organizations to balance control with performance and compliance. The choice is not binary. When combined, these approaches can have complementary economic, technical, and institutional implications that shape who can adopt AI, how quickly they can innovate, and how much control they retain over the technology. At the root of that growth trajectory is the principle of openness. Lowering barriers to adoption and experimentation One of the most consistent barriers to AI adoption—especially among small and medium-sized enterprises—is cost. Licensing fees, usage-based pricing, infrastructure requirements, and integration challenges can all slow uptake, regardless of the underlying model. Open source AI plays a decisive role in lowering these barriers as a crucial first step. By reducing upfront costs and allowing customization, it enables more firms to experiment, learn, and integrate AI into existing systems early on in their commercialization journey. What follows is that proprietary and managed AI services help organizations move quickly from experimentation to production. The result is not a trade-off, but a pipeline: open approaches expand participation at the front end, while commercial platforms help successful use cases scale across the economy. Transparency, trust, and digital sovereignty Trust, accountability, and sovereignty are central to sustained AI adoption. As AI systems are deployed in health care, finance, and public administration, governments and institutions need confidence in how these systems operate and evolve. Open source AI offers distinct advantages in this regard. Auditable models, transparent documentation, and the ability to adapt systems to Canadian legal, democratic, and cultural contexts improve oversight and public confidence. These benefits complement, rather than replace, the guarantees offered by commercial providers. In this sense, openness supports digital sovereignty because it increases resilience and imbues control from the outset. Rather than relying exclusively on proprietary platforms developed and controlled abroad, Canada can build AI systems that reflect its own linguistic diversity, cultural contexts, and regulatory frameworks. This includes fine-tuning models for Indigenous languages, regional dialects, and sector-specific needs when ready-made solutions may not be available. Accelerating innovation and closing the commercialization gap Open models’ value in commercialization is often misunderstood in debates about artificial intelligence. They are sometimes framed as an alternative to “commercial” AI platforms, or as an ideological position about how AI should be built. In practice, their value is more pragmatic. That is precisely because they function as a shared innovation layer. By making foundational tools widely available, they allow companies, researchers, and public institutions to experiment rapidly, test use cases, and build internal expertise without committing prematurely to a single vendor or deployment path. This collaborative dynamic has long been a feature of open source software. In AI, it is proving equally powerful. Open models enable faster experimentation, cross-sector learning, and the rapid diffusion of best practices. They also make it easier for Canadian firms to adapt global advances to local problems—whether optimizing energy grids, improving agricultural yields, or enhancing financial risk management. There is also growing evidence that open source aligns well with investor preferences. Studies show a positive relationship between commercial open source startups and early-stage valuations, with investors often viewing open architectures as signals of technical credibility and long-term flexibility. For Canada, where publicly funded research plays a large role in AI innovation, open source offers a pathway to ensure that knowledge generated in labs can be translated into market-ready applications. It reduces friction between research and commercialization and helps retain economic value domestically. Workforce development and skills diffusion Canada’s AI ambitions will rise or fall on people, not just technology. Employers consistently cite talent shortages as a key constraint on adoption. At the same time, surveys show that Canada lags peers in AI literacy and training penetration. Open source AI helps address this challenge by lowering the cost of learning and experimentation. Students, workers, and developers can access real models, real code, and real tools—not just abstract concepts—and build hands-on experience that translates directly into workplace capability. This matters for reskilling as well as education. As AI reshapes job roles, workers across sectors will need opportunities to acquire new skills. Open source environments make it easier for training institutions, employers, and individuals to design practical learning pathways without prohibitive licensing costs. In this sense, workforce development is not a defensive exercise but a growth strategy. Current evidence indicates that firms are using AI to enhance output and redefine roles rather than reduce overall employment. Open source AI can accelerate this transition by making skill development more accessible and inclusive Sectoral applications across the economy AI adoption across a wide range of Canadian sectors is already contributing to progress in certain areas and holding out unrealized potential in others. Agriculture uses AI for precision farming, yield optimization, and climate adaptation, but adoption rates remain low, suggesting significant room for growth. Energy is deploying AI to manage grids, optimize consumption, and support the rapid expansion of data centre infrastructure—a growing priority under Canada’s Sovereign AI Compute Strategy. Financial services lead in adoption, using AI for fraud detection, personalization, and risk management, with several Canadian banks ranking among global leaders. Health care and government services are experimenting with AI to reduce administrative burden and improve service delivery, though trust and governance remain central concerns. ICT and manufacturing are increasingly integrating AI into customer experiences and production processes, but manufacturing adoption remains uneven. Economic impact: From potential to performance The macroeconomic stakes of case studies like these are substantial. Independent forecasts suggest that AI could add up to 9 percent to Canada’s GDP by 2035 and more than $180 billion annually by 2030. Generative AI alone could significantly boost productivity and support the creation of tens of thousands of innovation-driven jobs. But these outcomes are not automatic. They depend on the pace and breadth of adoption across Canada’s firms, regions, and sectors, the availability of infrastructure and skills, and the policy environment shaping investment decisions. Implications for Canada’s next AI strategy The federal government’s current consultation process asks the right questions: how to strengthen research excellence, accelerate adoption, scale Canadian firms, build infrastructure, and ensure trustworthy AI. The evidence suggests that the next national AI strategy should treat open source not as a niche concern but as a strategic foundation. This implies several policy directions: Recognizing open source AI as a core enabler of adoption and commercialization. Investing in open infrastructure, including shared compute resources and open models. Strengthening pathways from publicly funded research to market-ready applications. Supporting workforce training and reskilling using open tools and environments. Embedding transparency and accountability into AI governance frameworks. These steps align closely with the themes emerging from the consultation and with international best practices. Conclusion: Turning leadership into lasting advantage Canada’s early leadership in AI research and policy is a genuine achievement. But leadership in today’s AI landscape is measured not by publications or pilot projects, but by adoption, impact, and inclusion. Open source AI offers Canada a pragmatic path to convert promise into performance. It lowers barriers to experimentation and participation, builds trust, and offers innovators choice on diverse AI models that can scale their business. As the government defines the next chapter of its AI strategy, embedding openness into policy, investment, and practice could help ensure that AI strengthens Canada’s economy—and that its benefits are widely shared. The opportunity is real. The choices made now will determine whether Canada remains an AI pioneer in name only or becomes a leader in practice, which would grow the Canadian economy and advance scientific innovation in the process.