How the promise of AI is taking hold at Canada’s biggest banks

The market is still underestimating how AI monetization in banks compounds through distribution, not just cost takeout. The first-order story is productivity, but the second-order effect is higher conversion on existing balance sheets: better prediction improves cross-sell, deposit gathering, and risk-adjusted lending, which can lift revenue without proportionate balance-sheet growth. That matters because the most valuable AI applications in banking are the ones that sit inside workflows already producing fees and NII, so the earnings leverage can arrive faster than headline AI budgets suggest.

TD looks like the clearest relative winner on execution momentum, but the bigger signal is competitive re-rating across the Canadian bank complex. RBC’s credit-underwriting use case is arguably the more economically powerful model because it can expand originations while staying inside guardrails, which is a cleaner path to ROE than pure employee productivity. CIBC’s deposit-generation evidence is important because deposits are the cheapest funding source; if AI materially improves deposit gathering, the benefit can flow through multiple years of margin resilience rather than a one-time revenue bump.

The key risk is not technical feasibility, it is governance drag and customer trust. If AI-driven recommendations create even a small uptick in mis-selling, bias, or model explainability issues, deployment will slow and the monetization timeline stretches from quarters into years. Cybersecurity is the other underappreciated tail risk: as these models become more embedded in customer-facing workflows, the attack surface grows and could force heavier compliance spend, diluting the operating leverage story.