The Number That Matters
TD Bank dropped mortgage pre-adjudication time from 15 hours to under 3 minutes, 99.7% reduction, with an agentic AI agent built by Layer 6, their internal AI research lab. This is not a pilot or a proof-of-concept deck. It is a production deployment inside a $2.1 trillion-asset bank with 28.1 million clients, processing real loan files under a documented Trustworthy AI governance regime that DataIQ already named Best Responsible AI Program in North America for 2025. The agent classifies client documents, extracts income data, validates consent, flags policy discrepancies, and writes the summary memo a human underwriter signs off on.
If you are a CIO, CFO, or CRO at a mid-market lender and your board has not yet asked you how your governance program compares to TD's, they will by Q3. The announcement landed ten weeks before the EU AI Act's August 2, 2026 full enforcement date, which classifies credit scoring as a high-risk system requiring documented human oversight and explainability. TD's timing is not a coincidence. Their Trustworthy AI program, with its privacy, security, fairness, accountability, and explainability gates, was already running before this agent was scoped. The mortgage agent is not a bolt-on; it graduated from a governance pipeline that regulators will now use as a reference architecture.
Why Mortgage Pre-Adjudication Is the Textbook Hard Case
Mortgage origination is the worst-case scenario for agentic AI in regulated finance. High-stakes credit decisions, heavy documentation requirements, consumer protection rules, disparate-impact risk, and a regulatory perimeter that extends to every step of the workflow. If you can ship an agent here, you can ship it almost anywhere in retail banking. TD's agent does not approve the loan, it does the pre-adjudication work that historically consumed underwriter hours before a human ever made a credit decision. That bounded scope is the single most important architectural choice in the entire deployment.
The team structure signals how serious the build was. Sandra Aziz joined Layer 6 as Technical Product Owner in March 2025 and led a cross-functional group of research scientists, data scientists, and TD process experts. The launch came roughly 14 months later. That is not a quarterly hack, that is a multi-quarter productization effort using TD's own bank knowledge and Layer 6's research depth. Layer 6 itself has been operating as TD's AI center of excellence since the $100 million acquisition in January 2018, led by founders Jordan Jacobs, Tomi Poutanen, and Maks Volkovs, who also co-founded the Vector Institute. When Aziz says "We built where nothing else existed. Everything is new," she means the team wrote the workflow orchestration, the document classification logic, the income extraction rules, and the policy discrepancy detection from scratch, no off-the-shelf agent framework.
The Governance Regime That Makes It Copyable
The reason this deployment matters beyond TD is that the governance regime is documented and pre-existing. TD's Trustworthy AI program evaluates models on privacy, security, fairness, accountability, and explainability before customer contact, and continues monitoring post-deployment. The agent is not operating in a policy vacuum, it is operating inside a risk-rating taxonomy that maps directly to the kinds of questions U.S. regulators will start asking by Q4 and that EU AI Act Article 6 high-risk system requirements already mandate.
Every mid-market lender should be walking their board through this case study in the next 60 days, not as a product pitch but as a forcing function for three questions: Do we have a named Trustworthy AI program with documented evaluation gates? Do we have a risk-rating taxonomy that classifies AI systems by customer impact and regulatory exposure? Do we have a human-in-the-loop discipline that defines when an agent hands off to a person and what that person is accountable for? If the answer to any of those is no, the mortgage agent deployment is the artifact you use to get budget.
What Mid-Market Lenders Should Do This Quarter
TD's Chief Analytics and AI Officer Luke Gee stated that TD has targeted approximately $1 billion in annual AI value, with roughly $500 million in annualized revenue at stake. That is not a rounding error, that is a board-level commitment with named executive ownership. Mid-market lenders do not need to match TD's dollar figure, but they do need to match the governance posture. The playbook is now visible: bounded scope, cross-functional team with both ML and process expertise, pre-existing governance pipeline, documented risk evaluation, and human-in-the-loop handoff rules.
If you are running a B2B SaaS, FinTech, or HealthTech platform that touches lending workflows, the TD case is the reference architecture your buyers will cite when they ask how your Agentforce or Claude-based agent integrates with their compliance program. The governance regime is no longer optional, it is the product.
The takeaway: TD just published the governance playbook every regulated lender will be measured against by year-end.
What to do next
Book a Strategy Session