TD Bank Group has crossed a threshold that most financial institutions are still approaching from a distance. Canada’s second-largest bank has deployed a live agentic AI model inside its mortgage and home equity lending operations, compressing a 15-hour manual underwriting task down to under three minutes.
The system, built by TD’s internal AI lab Layer 6, handles the pre-adjudication phase of mortgage and home equity line of credit (HELOC) applications autonomously. It classifies documents, extracts borrower data, calculates income, validates figures against policy requirements, performs consent checks, flags discrepancies, and delivers a structured summary memo to a human underwriter. No shortcuts, no narrow task automation. This is end-to-end workflow execution with a human receiving the output rather than driving it.
The same workflow logic TD is applying to mortgage origination maps almost directly onto processes that are central to this industry: account onboarding, income verification, document classification, creditworthiness assessment, and dispute resolution. If agentic AI can collapse 15 hours of underwriting prep into three minutes with improved accuracy, the same architecture can plausibly do the same for collections intake, account review queues, litigation support documentation, and compliance audits.
TD is not treating this as a pilot. The bank says it has mapped every stage of the real estate lending journey and intends to introduce agentic AI at each step. It is also actively exploring deployment across other business lines. Chief Analytics and AI Officer Luke Gee framed the vision plainly: a hybrid model where human colleagues and AI work together to get clients to a faster decision.
What separates TD’s announcement from the usual AI press release is the governance structure behind it. The bank’s Trustworthy AI team evaluates every model on privacy, security, fairness, accountability, and explainability before deployment and continues monitoring after. That framework earned recognition as the Best Responsible AI Program in North America by DataIQ in 2025. In a regulatory environment where consumer financial protection scrutiny is intensifying, that kind of structured oversight is not just good ethics. It is a competitive differentiator and a potential model for how institutions in this industry approach AI deployment.
TD has set a target of $1 billion in annual value from AI. The mortgage automation launch is the opening move. Organizations across lending, servicing, and collections should be studying it closely.
