As a child, Jesse Cresswell thought a lot about the question “what if?”
It was a core question posed by the books he devoured, from the science fiction by Isaac Asimov that posed ethical dilemmas related to technology, to the popular science books that described major breakthroughs in physics, like The Elegant Universe by Brian Greene.
Two decades later, Cresswell remains excited by imagining what the future can look like. But now he does so by leading the Trustworthy AI team within Layer 6 at TD.
Cresswell is a Staff Machine Learning Scientist at Layer 6, the cutting-edge artificial intelligence (AI) team that has been part of TD since its acquisition in 2018. Layer 6 has grown from roughly 20 colleagues to more than 200 today.
The Layer 6 team is responsible for developing AI tech solutions for the Bank that support colleagues and customers. Researchers use existing information — like TD customer service policies and procedures — and apply AI technology to them, letting the technology analyze patterns and put together information quickly. The goal is to create new solutions that help TD colleagues better serve customers.
Part of the team’s work is to develop computer models — a simplified representation of a real-world system or process using data, math, and instructions — where AI can be used to help make the lives of colleagues or customers easier.
For instance, Layer 6 recently developed a chatbot that is trained to “understand” the contents of the Bank’s customer service policies. So, when a customer calls the Bank looking for help on a particular issue for their bank account, this AI technology can help a customer service agent find the right policy information more quickly, to help them provide speedier customer service.
Building this type of technology ethically is a top priority, and that’s where Cresswell comes in.
Leading a team of eight people, Cresswell requires that as TD develops its AI technology to serve customers and colleagues, it’s built with a governance framework that includes checks and balances every step of the way. That TD approach, and the team that leads it, is called Trustworthy AI.
“We align our AI models to human needs and values,” Cresswell said of his team’s work within Layer 6.
Building human values into AI
As AI becomes more and more important to the way large organizations like the Bank operate, building the technology on a foundation of human values is a big job.
After all, Cresswell and Layer 6 are working to develop cutting-edge technology to help shape the future of banking at TD. As the technology evolves, it is critical to TD that the recommendations made by AI technology enables the Bank to stay true to its values.
Those values include privacy, security, fairness, accountability, and explainability.
“It’s incredibly important that, as a society, we understand the technology and build guardrails for it.”
Cresswell explains this is done best when considering values together. For instance, improving the security of an AI model might also make it less fair, as it could introduce barriers for some people. One particular demographic of customers might find those security features frustrating to their ability to access the technology. A robust understanding of the ways these decisions can affect customers means Cresswell and the team can evaluate tradeoffs to best serve customers.
“Everyone who comes to our bank — no matter their gender or age, for instance — is an individual with different needs, at a different stage of their life,” he said. “We need to build technology that does not disadvantage any given group, and that we do so while maintaining the privacy and confidentiality of personal information.”
Growing trust
Cresswell sees the nascent and growing AI field as an opportunity for TD to listen to public needs and take leadership on what the future of the broader responsible AI field looks like.
“As a bank, we constantly think about how we can improve the customer experience and protect customers from day-to-day risks,” he said.
Cresswell’s team does so with a human-in-the-loop approach, to involve Bank colleagues in making critical AI decisions. AI may make a recommendation, but that’s not enough. TD colleagues review recommendations to validate whether it would be right for customers . That means the experience and judgment of TD colleagues are looking out for customers, while AI complements the experience.
The Trustworthy AI team is already being recognized by the industry for its work. The approach, and team, was recently named by DataIQ as the Best Responsible AI Program or Initiative, of any company, in North America.
Cresswell feels the Bank’s Trustworthy AI approach also help spur innovation.
“Constraints breed creativity,” he said. “Requirements to make our AI models more trustworthy focus our attention on the most important aspects of how we support customers.”
His output is living proof of this philosophy. He was named TD Inventor of the Year for 2024 and has more than 100 patent filings to his name.
That work is part of TD Invent, the strategic umbrella effort to power innovation to help make banking easier for customers and colleagues.
For someone who extensively researches and advocates for what it means to build responsible AI, Cresswell remains attuned to evaluating both risks and the opportunities to better serve customers.
“Great researchers like the ones on the Trustworthy AI team are motivated by serving the public good,” Cresswell said.
“Everybody on the team recognizes the power of AI and the potential that it brings for change, but also the necessity of making sure that we do it right.”
It’s a mindset that comes from a lifetime of thinking about the question “what if?”
