Canada faces a massive looming divide on artificial intelligence. The Canadian government has committed billions of dollars to AI. Companies are rushing to adopt the latest tools from OpenAI and Anthropic. Canada has become the second highest per-person adopter of Claude, ahead of every country in the world besides the United States. And yet Canadians remain skeptical, if not yet outright hostile, to the prospect of AI adoption, as polling results indicate. Data centre construction faces similar headwinds. And though the kinds of spectacular expressions of discontent found elsewhere (i.e., graduates protesting AI at their ceremonies, public meetings being regularly overwhelmed by protesters) have been comparatively rare in Canada, the underlying sentiment for engrained resistance is definitely present.
Why is this happening? How can reactions be so volatile? Canada is near the bottom of a 47-country list for AI trust and literacy, according to a study by KPMG. At the same time, KPMG says 70 percent of Canadians believe AI will lead to positive outcomes, and 60 percent are already seeing results in their day to day. If the benefits of the technology were communicated effectively, AI could be more widely supported in Canada; while the majority of Canadians are aware of the good that AI can do, they are also scared of the uncertainty about the future. Quite frankly, I don’t blame them.
I. Winning the Argument
CEOs of the leading AI labs have spent the last few years pushing narratives of immediate job loss and human displacement as a direct result of their products. At one conference, Sam Altman said that he “sees intelligence as a utility like electricity or water and people buy it from [OpenAI] on a metre and use it for whatever they want to use it for.” Naturally, this is not an appealing vision for the future.
All comments like these have done is sow division, both in online discourse and in local approvals for tech-backed infrastructure. The result is an incredibly polarized sentiment on AI all around the world, primarily in the United States, where one poll shows how unfavourable each AI leader actually is and another illustrates how sentiment has shifted among Americans under 30. In reality, AI adoption has not yet driven the intensity of change to the global economy promised by Altman et al., although the damage to the technology’s reputation has been done. On a recent podcast, Altman reneged on his prior stated timelines, arguing that the rollout of AI and its impact on society will take much longer than he thought, and conceded that the industry has done a poor job of messaging their technology. Unfortunately, it may be too late to save their reputation.
Canada has an opportunity to learn from the communication failures of American AI companies. So far, champions of AI adoption have not taken the time to tell Canadians why AI is important, never mind why sovereign AI matters. It’s hard to move quickly and cover all the bases when the pace at which AI is changing the fabric of our society is unlike anything we’ve seen in decades.
As a young entrepreneur in the AI field, I am comfortable with the concept of citizens having control over such powerful technology, and it’s safe to say that Canada needs a reasonable measure of control and oversight over the AI it relies on, from the data centres that generate compute to the various interfaces through which we interact with AI on a daily basis. Meeting that goal would have to start with good messaging. In a democratic society such as ours, we need buy-in from the majority of Canadians to move at the pace required to build critical infrastructure here in Canada. Quite simply, if Canadians can’t see or understand the benefit to themselves, they’ll stand up and push back.
The argument has yet to be made to the Canadian public in ways that are convincing and resonant: that rapid AI infrastructure development is how we keep Canadian businesses competitive on the global stage and protect Canada’s national security from adversaries with access to frontier models. Resistance to achieving this feels like NIMBY-ism in a new frontier. Few people want AI infrastructure built literally in their backyard.
People outside of tech keep asking me how AI benefits them. To some, it feels like entry-level jobs are being automated away. CEOs from companies cutting their headcount make it sound like AI is to blame. Neither could be further from the truth. Companies overhired during COVID and are now rightsizing. But that’s not the narrative their boards want to hear, so the stories aren’t told truthfully. This kind of convenient messaging might help the company’s leadership today, but it hurts society in the long run. It should be reported better by all involved.
Every single company that we at North Group are helping implement AI is using the technology as a reason to become more ambitious, not to replace their team en masse.
AI allows employees to automate the monotonous work that no one likes. This frees them up to spend more time on high-value activities that they enjoy and AI is not good at. Not many people chose their role because they loved repetitive pencil pushing. It’s time we use AI to allow people to spend as much of the workday as possible on their craft—the meaningful strategic work that matters to both them and their firm. With that being said, some jobs will undoubtedly be impacted at the macro level as what work looks like changes. The first layer of AI implementation leans heavily on augmentation, leaving humans firmly in control. As technology continues to improve, the potential opportunity shifts towards automation, where AI can handle complete workflows, at first merely deterministic ones, then eventually those that require judgement. It is important to note that this is not a rapid transition. Building enough trust for organizations to hand off complete workflows to technology takes time, as it has in past technological revolutions. As this shift happens, society must take care of the individuals whose jobs are impacted, both with retraining and with the gains from the prosperity that AI is driving. The best way to avoid being negatively impacted by AI is to learn how to use it and to educate yourself on what the AI era could look like.
I envision a future economy with a culture of entrepreneurship that empowers anyone to build their own company. A world where knowledge is widely available and services are more accessible. On the way there, industries may shift from a few massive companies to ones with many more lean, mid-sized businesses. Incumbents that seek to merely maintain their position and avoid AI adoption will be disrupted by smaller firms that use AI to give their employees superpowers.
Many companies will become AI-native and take advantage of the opportunity for potential disruption and innovation. AI-native companies can achieve more with less employees. As they take market share from incumbents, they won’t need to hire as quickly. Not every large company will become smaller in this process, but many will downsize as a result of this disruption. The scale of new firms and entrepreneurs created as a result of AI should balance out the loss of jobs from large incumbents. Humanity is resilient and employment has always weathered changes in technology. There are extremely valid concerns about the way that AI is being deployed. Environmental impact, grid reliability, and cognitive atrophy are all important considerations.
The solution to these issues is not to stick our heads in the sand and avoid AI entirely. This would not be productive or realistic. It goes against the best interests of national security and Canadian competitiveness in global markets. Instead, we should focus on delivering the best versions of AI infrastructure we can and prioritizing AI use cases that will tangibly improve the lives of Canadians like improving healthcare, speeding up government services, and automating the boring parts of an employee’s day to drive GDP growth. Early adoption of AI has shown that people are most in favour of the technology when it is deployed into less human reliant roles and the parts of their job they don’t like, as opposed to automating leisure.
To get there, we must build better data centres. South of the border, backlash is driven by environmental concerns, energy constraints, and worries about the impact on quality of life, according to a recent Fox poll of data centre opponents. To negate this, we must develop energy resilience and expand our grid capacity quickly. Successful data centre proposals have prioritized closed loop cooling systems and standalone energy sources to remove any reliance and strain on existing infrastructure. If data centres are to be created at the scale required for Canada to maintain control over the AI it uses, proponents need to consider the direct benefit that the infrastructure will provide to the community. Meta’s new data centre in Alberta targeted decreased electricity costs and an expanded tax base. As the demand for AI ramps up the demand for energy and compute, it is critical that Canadian infrastructure keeps up. Otherwise, we will be increasingly reliant on American data centres to service our AI demand. Given the geopolitical turbulence and Canada’s desire for sovereign AI, building the right kind of data centres is more important than ever.
As we do this, we must continue to research how increased AI usage impacts our brains and our labour market, while being honest and transparent about the implications of the technology. We’re already doing these things, but the majority of Canadians have not been won over yet. In other words, most people don’t understand that the action we take now will define the role of our country’s economy in the next few decades. Instead of being solely pro-AI or solely anti-AI, we can focus on charting a productive path through the middle by making difficult choices in how and when and what kinds of AI we build. But build we must. This part of Canada’s AI strategy has been communicated well. Ultimately, it should not solely be the job of CEOs or governments to explain why AI matters. Everyone has a part to play.
II. Toward an AI Consensus?
We must lead with concrete examples of the wins from AI adoption, the expected victories and visions for the future if we get it right, shouting them at least as loudly as the concerns. In Canada’s recent AI for All strategy, these positives could have been clearer. Simply put, AI matters to all Canadians, whether you’re a nurse in Victoria, a public servant in Ottawa, an engineer in Montreal, or a fisher in the Maritimes. As I see it, the outline of a future AI consensus in Canada should revolve around the following points:
● Canada needs AI infrastructure to retain control over AI access, remain competitive in global markets, and stay safe from bad actors with access to good AI models.
● AI will continue to create better public services with more equitable and affordable access, like reducing hold times for the CRA and wait times for healthcare.
● AI will enable greater competition and the disruption of legacy incumbents that don’t deliver quality outputs.
● AI will drive a culture of entrepreneurship and return Canada to a country of builders.
● AI will allow employees to focus on their highest value, strategic work and automate the monotony that eats up their thinking time.
● AI is the key to improving Canada’s productivity and making our country one of the most prosperous in the world.
● AI will soon be as critical as electricity and the internet. It can help us build the future we want, but only if AI development happens in the right way.
We can start by thinking through these arguments and counterarguments now, not as afterthoughts to be settled once the technology has gone out the door but as essential elements of a unified strategy for ensuring the potential of AI is realized in a manner consistent with the common good of all.
There is also the question of what role governments can play, particularly with respect to the more sensitive public safety aspects of the AI frontier. With its intervention in the release of Anthropic’s Fable and OpenAI’s Sol, the United States has set a precedent for governments to review new frontier models before the public can broadly access them. Each new frontier model will likely face multiple weeks of intense scrutiny before it becomes widely available. We may reach a point where model providers are under pressure from the government to keep the latest models only available for intelligence or cybersecurity use cases, or to a small, select number of organizations in the “commanding heights,” such as those in Anthropic’s Project Glasswing.
Such an arrangement will raise new questions that touch on AI's relationship to democracy and popular legitimacy. If this new form of intelligence is meant to be accessible, how do we reckon with a world in which usage is controlled by a very small number of people? Does everyone actually need access to the best models? If the most advanced forms of AI are to be restricted, at least initially, what precisely are the responsibilities and prerogatives of those entrusted with holding that technology? While we don’t yet know the answers, we should keep in mind that where control over AI matters for a country’s security, it also matters for a company’s competitiveness. We need to build a future that balances prudence and progress, where the success of the economy is not tied to a flick of a switch that turns models on and off.
Another question is how we can keep our made-in-Canada breakthroughs and innovations rooted here, in order to avoid the brain-drain and tech-drain that has so often sapped Canada’s technological prowess in the past. Canadians and graduates of Canadian universities have built the foundation for AI, making outsized contributions with the creations of the Vector Institute and MILA, and by co-founding OpenAI, Anthropic, XAI, and countless other frontier AI companies. But other countries have done a better job of capitalizing on it, not just in luring talent but in creating the ecosystems and feedback loops that allow a particular application to scale into a successful competitor. We need to ask how Canada can continue to lead in AI, and harness the research it creates, especially for open-source and open-weight AI. It scarcely needs saying, but when you own your own intelligence, you have more control over your destiny.
To deal with the (understandable) degree of uncertainty and fear that many Canadians are experiencing, we need to communicate with clarity and consistency, address concerns head on, and remind ourselves that the future is yet to be determined. That open-endedness and sense of agency, as opposed to the dictates of a crude techno-determinism, can be the most effective and empowering message to impart. Industries and communities across Canada have choices in how that future will look. But we must make those choices together, and AI development must continue under a shared picture of the future.
We have to reassure skeptics and those most vulnerable to technological shocks that the transition can and will be navigated in the best way possible, one that minimizes negative impact to society and puts humans first. That begins with understanding how the technology can shape industries and public services for the better. These are, of course, general principles and their application will be subject to disagreements and compromises but alignment on the premises is a prerequisite to the kind of inclusive and harmonious AI roll-out Canadians overwhelmingly desire. And we can begin to broaden that conversation now.
One thing we are missing, though, is a platform for leaders to communicate the ways that AI will tangibly improve life for Canadians. This is one reason why we created the podcast Building The North. After recording six episodes with CEOs, politicians, and founders, we’ve already heard exciting visions for how AI could improve government, banking, healthcare, housing, and more. We will likely need more such platforms in order to develop a healthy, variegated, and constructive AI discourse in Canada.
It may seem too ambitious to even just propose such a thing as an eventual AI consensus, which would encompass labour markets, bureaucracy, national security, strategic interests, and homegrown innovation, and more. But it is a worthy objective to aspire to, one that would mirror the process by which past technologies moved from mistrusted agents of disruption and displacement to accepted facts of everyday life. In other words, it falls on the builders and advocates of AI to earn legitimacy rather than expect it to be handed to them on a silver platter. One thing we can be sure of is that without better messaging, this “social licence” will become ever more elusive, and the AI divide will continue to grow.
III. A Fork in the Path
Messaging is the next frontier that threatens AI advancement. On this topic, Canada has two options. It can follow in the footsteps of American tech CEOs and continue to market the long-term unknowns of AI, or it can learn from their mistakes and offer a more measured, honest outlook on the pace of AI adoption. This choice will determine much of the success of Canada’s AI rollout.
If the mistakes made by leaders south of the border are repeated, Canada will be unable to build the AI infrastructure needed to secure Canadian control over the AI platforms used by businesses across the country. This means Canada will need to rely on American infrastructure to meet the AI demand of Canadian businesses, something that will grow rapidly as they attempt to stay globally competitive over the coming decade. In a world where historically strong geopolitical relationships have become increasingly uncertain, the impact that overreliance on American infrastructure could have would be devastating.
If Canada learns from the communication errors of American tech CEOs, messaging would be grounded in the tangible benefits to the end user, directly targeted at skeptics to understand their perspective, and focused on the role that Canada could play in this space. The Canadian AI industry may never have a state-of-the-art frontier AI model, but it likely does not need one to thrive. As compute and energy infrastructure is built, Canadian companies can temporarily leverage frontier AI models to learn the value AI will create and train staff on how to use it. Once infrastructure has been scaled to meet demand, organizations can switch to cheaper, sovereign models to run their automated workflows. Most AI use cases don’t require the best model that exists on the market.
Canada has a moment of great opportunity. The Federal Government’s AI for All strategy correctly attempts to thread the needle between the AI strategies of the United States and China, providing a third option grounded in responsible AI. In the long run, every country, and perhaps every company, will want their own open-source or open-weight model that provides fast, cheap intelligence that no one can turn off. This is how Canada can lead in the AI era. To get there, we first need effective messaging.
Julian Wells is the co-founder of North Group, a firm that builds AI for business. He also founded Studenthaus, a consumer intelligence platform for student housing, and was the youngest person selected for the 2024 class of BCBusiness 30 Under 30.