Toronto-Dominion Bank TD-T is rolling out its first real-world generative artificial intelligence platforms as global lenders race to test the technology and its impact on the financial sector.
Canada’s second-largest lender launched two pilot projects Wednesday to implement generative AI platforms for its contact centres and engineering teams. Banks and the industry’s regulators have been working to understand the technology’s application in the sector, and how to safeguard lenders from potential risks.
AI has already been adopted widely across the industry as lenders have invested in automating functions such as risk management, trading and investment research. With the rise of generative AI, banks have been exploring how to implement the enhanced form of the technology, which learns from data-based patterns and trends to help it generate content, make decisions and solve problems.
“A lot of businesses and financial services are stepping into this area with the appropriate level of caution to understand how they can create good outcomes whilst also knowing how this thing works,” TD head of AI and analytics Luke Gee said in an interview. “We have moved forward with the appropriate pace and caution, and we’re also doing everything we can to bring our people with us so they get to enjoy this.”
Major U.S. banks have already started dabbling with generative AI programs. Last year, JPMorgan Chase & Co. said it is working with regulators to explain the bank’s generative AI pilot projects and ensure that it has implemented the proper controls. In early May, the banking giant unveiled IndexGPT, a program similar to ChatGPT that builds thematic indexes and identifies investment opportunities based on trends.
Canada’s banking regulator has also been keeping a watchful eye on the spread of AI technology. Last year, the Office of the Superintendent of Financial Institutions pointed to the threat of digital innovation, including AI, on a list of the top risks facing the stability of the financial sector.
Many of Canada’s largest banks are exploring and testing generative AI. For now, lenders are largely using the technology to boost employee productivity.
In the early stages, banks are largely testing generative AI in ways that could reduce costs. The technology could add up to US$340-billion in value across the sector, according to a 2023 report by McKinsey & Co. Inc.
In one of TD’s pilots, the bank is using the technology to streamline customer interactions with the bank’s contact centres. The generative AI virtual assistant aims to reduce wait times by helping customer service representatives find answers to day-to-day banking inquiries faster.
When representatives are unable to answer a customer question or problem, they turn to the guidance of higher-level support teams. The generative AI virtual assistant is designed to step in to create written responses for bank employees, along with links to the TD policies and procedures used to inform its answers.
TD’s research and development centre, Layer 6, developed the generative AI virtual assistant and trained the platform with the help of employees who understand the bank’s practices.
The technology is meant to help employees do their jobs more efficiently and effectively, rather than replace roles with generative AI, according to TD’s executive vice-president of North American customer operations Dawn Cooper.
“People will never be replaced,” Ms. Cooper said. “There’s always going to be a need to have conversations in the financial industry when there’s a big moment happening. When there’s a complex issue, that’s when you really want somebody to be with you, human to human, to have those conversations. Generative AI is really the complement to the people.”
The bank is also testing GitHub Copilot, a generative AI programming assistant developed by Microsoft Corp MSFT-Q. The platform helps engineers analyze code, offer recommendations on how to write it and detect bugs.
After TD did an initial and smaller scale test with GitHub Copilot late last year, employees said in a survey that the platform could save them up to 20 hours of work over two weeks.
TD’s chief engineer and architect Licenia Rojas said that purpose of the platform is to “accelerate cold development and create a robust experience for our developers and engineers that we could leverage more of their time towards creativity, features and capabilities that we would want to enable for our customers.”
TD has been actively developing technology platforms and programs. In the past three years, the bank has filed 2,500 patents, of which one in five are related to AI.
“We’re the number one patent filer among financial institutions in Canada and the top 20 in the U.S.,” said Imran Khan, head of TD Invent, the bank’s technology innovation unit. “We have an exhaustive patent portfolio and AI is a big part of that.”