What is agentic AI and how TD Bank uses it

Agentic AI represents a new generation of artificial intelligence capable of independently performing multi-step tasks with minimal human intervention. Unlike traditional chatbots that merely answer questions, agentic systems actively work — analyzing data, verifying facts, evaluating results, and preparing materials for human decision-making.

TD Bank deployed this technology in May 2026 into the process of approving mortgages and Home Equity FlexLine (the Canadian equivalent of a HELOC, a credit line secured by real estate). In practice, the agentic AI performs the following steps:

• Scanning client documents — automatically extracts data from uploaded files
• Income calculation and its validation against the bank's internal rules
• Consent checks — verifies that the client has given the necessary consents
• Income verification from various data sources
• Detecting discrepancies in the application
• Automatic summary generation (memo) for the underwriter with complete findings

The entire process, which previously took an average of 15 hours of human labor, is now handled by agentic AI in under 3 minutes. The human underwriter then receives a ready-made basis for the final decision.

Sandra Aziz and the Layer 6 team: Who is behind the development

TD Bank's first agentic AI tool was spearheaded by Sandra Aziz, Senior Machine Learning Engineer at Layer 6 — the bank's internal AI research center specializing in advanced technological solutions. Aziz joined TD in 2014 as an intern with a master's degree in statistics from the University of Montreal and has since gone through several waves of technological transformation.

"When I started, we were called the statistics team, then we became the analytics team doing predictive modeling, and then we started using AI," Aziz described. Her career at the bank mirrors the evolution of artificial intelligence itself — from basic statistical models through predictive modeling and generative AI to agentic systems.

Before the mortgage project, Aziz led the insurance fraud detection team at TD Insurance, where she deployed a predictive model capable of detecting fraud before it occurs. This collaboration later led to the development of the client-facing chatbot TDI Virtual Assistant, launched in early May 2026.

Trustworthy AI: Humans remain in the loop

TD Bank emphasizes that all AI solutions are developed within its "Trustworthy AI" approach — a governance framework that includes control mechanisms at every stage of development and deployment. A key principle is human-in-the-loop: AI prepares materials and recommendations, but the final decision always remains with a human.

"We've been able to speed things up. But at the same time, we have to make sure everything we do makes sense from the client's perspective," Aziz stated. "We want to be first to market with the fastest processes and the best technology. But we can never lose sight of our clients and what they need to make their lives easier."

Broader context: AI at TD Bank is nothing new

Agentic AI for mortgages is just the latest addition to TD Bank's growing portfolio of AI solutions. Layer 6 has previously developed several tools that now help thousands of employees and clients:

• GenAI virtual assistant in contact centers — helps operators answer client queries within seconds. Since its introduction, it has reduced wait times by 12%. It uses the RAG (retrieval-augmented generation) method, drawing answers exclusively from the bank's internal documents and providing source citations for every response.
• Branch Virtual Assistant — helps 17,000 TD branch employees respond to client queries faster.
• TD AI Prism — a predictive model that forecasts clients' future needs based on their financial behavior. It led to a more than 40% increase in marketing campaign success rates.
• TDI Virtual Assistant — the first client-facing generative AI chatbot among major Canadian banks for insurance, launched on May 1, 2026. It works without requiring login.

What this means for banking in Europe

TD Bank's example shows the direction banking is heading globally. Agentic AI in processes like mortgage approval can fundamentally change the speed and efficiency of financial institutions. In the European context, however, the situation is specific — the deployment of AI in banking is subject to the EU AI Act, which classifies creditworthiness assessment systems as high-risk.

This means European banks will have to go through a stricter approval process when introducing similar solutions and demonstrate transparency, non-discrimination, and human oversight. TD Bank, with its emphasis on human-in-the-loop and source citations, is essentially already showing the way to prepare for these requirements.

In the Czech Republic, we are already seeing the first signs — the Czech National Bank launched its own AI center in 2026, using Nvidia chips for financial market supervision. Commercial banks in Czechia are not yet deploying agentic AI in such an advanced form, but the pressure for digitalization and faster processes is growing.

TD Bank plans to expand agentic AI into further stages of the mortgage process — from document submission to loan disbursement. Similar tools will gradually appear in other areas of banking as well.