Building trust in AI: Q&A with Jim Banting

With Canada’s National Artificial Intelligence Strategy: AI for All, the federal government hopes to position AI as an integral driver of innovation in Canada. Teresa Reguly, co-leader of Torys’ Intellectual Property and Food and Drug Regulatory practices, and Jim Banting, Assistant Vice-President of Innovations, Partnerships, and Entrepreneurship at the University of Toronto, discuss the advancement of AI innovation and adoption in Canada.

Teresa Reguly: Canada's AI for All strategy hopes to build public trust in AI, accelerate widespread adoption, and invest in Canadian innovation, talent, and infrastructure. In your view, what will be a key factor in moving from plan to execution?

Jim Banting: Every major technology shift has produced this pattern: the calculator, the personal computer, and the internet all faced resistance before becoming indispensable. AI is following the same arc.

The factor that moves strategy into execution is building a shared understanding that AI is a leverage tool, not a replacement for expertise. Domain knowledge remains essential. AI amplifies what experts can do; it doesn't substitute for them. Trust grows from that understanding, and adoption grows from trust.

Looking ahead, AI adoption may invert the SaaS adoption trajectory. SaaS moved from large enterprises down to SMEs. AI could move in the opposite direction, or at least follow a different path, because enterprise adoption is being slowed by legitimate concerns over confidential and proprietary data. The winners in the execution phase will be organizations that solve for safe, governed ways for AI to interact with proprietary data, alongside those building the strongest models.

Organizations that provide support for adoption, whether through secure sandboxes or training on data privacy and security, will be increasingly important. For example, U of T President Melanie Woodin recently announced that the university will use its three campuses as a living lab for responsible AI adoption.

Teresa Reguly: You have been heavily involved in advancing support for research translation and commercialization. What are your thoughts on how the Canadian government supports AI innovation? Where could they improve?

Jim Banting: There are four areas of focus that I believe would help further support AI innovation:

• Training. Build the pipeline of talent equipped to deploy AI in applied settings, including forward-deployed engineers who bridge model capability and real-world implementation.
• Government adoption. Government use of AI, done well, is a credible demonstration to Canadians of where AI creates real value.
• Bridge funding. Targeted capital to close the gap between promising research and commercial viability at the stage often called the Valley of Death, when good technology stalls for lack of patient capital.
• Compute talent retention. Train high-performance computing specialists and keep them in Canada, so the country has the people needed to build and operate world-class supercomputing infrastructure, including the creation of testbeds to incubate Canadian next-generation hardware. GPU demand will exceed supply for some time, but Canadian researchers should start to have a line of sight to compute jobs that need 10,000+ GPUs. 

On the startup side, the discipline worth building is designing AI native from day one: build in optionality across model providers and approaches rather than architecting around a single vendor. Founders should avoid locking core product logic into deterministic processes. The value of an AI-native architecture is continued enhancement as the underlying models improve.

Sustainable growth is not possible without continued investment in basic research. New materials, therapeutics, and technologies destined for commercialization originate upstream, in fundamental research with no guaranteed near-term application. Bridge funding and adoption incentives address the middle and late stages of the pipeline, but they depend on a well of foundational discovery to draw from. Underfund basic research today, and the commercialization pipeline runs dry in a decade.

AI as we know it today would not be here without made-in-Canada research in neural networks and machine learning from Dr. Geoffrey Hinton, which is work that began decades ago.

Teresa Reguly: Are there any playbooks from other jurisdictions that you think Canada could look to in terms of building AI sovereignty?

Jim Banting: The UK and Sweden offer instructive models. Both show how sovereign compute can support AI and digital transformation while maintaining control over sensitive data, regulatory compliance, and national governance requirements. Their approaches combine hyperscale cloud infrastructure with data residency, trusted operational controls, security safeguards, policy enforcement, and transparency measures, enabling government and regulated industries to run critical workloads within a framework aligned to national law and public trust. Global-scale infrastructure paired with national governance controls is a credible template for Canada.

Teresa Reguly: There has been a significant acceleration of AI adoption and investment. Where do you see the greatest opportunities and challenges within this innovative pocket?

Jim Banting: The challenge is cutting through the “noise”. A large share of investment is going into the picks and shovels (infrastructure and tooling), rather than differentiated applications.

The opportunity is using AI to compress innovation cycles. The cost of building software has never been lower, so products and services can keep improving when built with an AI-native default.

The bigger opportunity is getting that approach to permeate deep tech: ventures addressing large, unmet needs, protectable with patents, capable of building a durable competitive moat. AI-native speed paired with defensible deep-tech IP is where the most sustainable value will be created.

Teresa Reguly: What unique opportunities does the Canadian market present?

Jim Banting: Ontario is one of the most interesting healthcare markets in the world: a large, diverse population with a similar level of access to care. That combination is rare among health systems and creates real potential for insight into care delivery and treatment outcomes. Unlocking that dataset for researchers, under strong privacy safeguards and clear consent frameworks, could generate discoveries in care delivery and new treatments that would be difficult to replicate in more fragmented or access-unequal health systems.

The federal government’s $100-million investment in VITAL, a new multi-provincial platform, is an example of the opportunity here: larger and faster clinical trials that bring new therapies to patients, more efficient health care systems, and continuous learning among clinicians and researchers.

Teresa Reguly: The University of Toronto has been an integral driver of AI research and advancement for decades. What are key insights that founders building AI ventures should know?

Jim Banting: There's no substitute for building a business that is a “painkiller” rather than a “vitamin”. Solving problems users are willing to pay for remains the north star for any venture, AI included, and founders should prioritize revenue from the outset.

AI venture-building has taken on a unicorn-or-bust mentality, but the heartbeat of the Canadian economy is SMEs. Profitable businesses that employ 30 to 200 people are a valuable, durable outcome, not a consolation prize relative to a unicorn outcome. AI is also a tool for building more of these businesses, faster. A commercialization ecosystem built solely around unicorns misses where most durable economic value in Canada gets created, and universities have a role to play in supporting founders toward either outcome.

With AI as a cross-cutting theme, U of T is positioned to support a distinct set of emerging sectors: self-driving labs (with AI in the loop), advanced materials, critical minerals, quantum, AI, health, and Arctic research. Each map to existing institutional strengths, and strong established and emerging industries in Canada and Ontario, that are looking to integrate our research to grow faster and to open new markets. University researchers are working with industry partners to make current federal or provincial priorities that could lead to sustained societal and economic impact.

Teresa Reguly: What is your 20-year forecast for the industry?

Jim Banting: Software AI will migrate into physical AI, embedded everywhere. LLMs will become a commodity layer, with differentiated value sitting around them, utilizing proprietary data sets to create products and services with a competitive moat.

Jim Banting leads the University of Toronto's Innovation, Partnerships & Entrepreneurship Office, connecting researchers, industry partners, investors, and entrepreneurs across the University's innovation ecosystem. His mandate includes strategic partnerships, research commercialization, and entrepreneurship development. He brings leadership experience in academic innovation, research partnerships, and  technology transfer, as well as industry experience in the Canadian and U.S. biotechnology sectors spanning partnerships, licensing, and mergers and acquisitions.

Prior to joining U of T, he led partnerships, technology transfer, and high-performance computing at a Canadian university, and co-founded a life sciences spin-off that was later acquired by a U.S. specialty pharmaceutical company. Jim holds a PhD in Pharmacology from Queen's University.