Want unlimited access to top ideas and insights?
Maksims Volkovs, senior vice president and chief AI scientist at TD Bank, and his team take a "pragmatic buy, build and blend approach" to developing AI tools — building in-house when they see a "data advantage," and leaning on off-the-shelf technology for speed and simplicity.
In the case of TD's new AI Prism predictive foundation model, Volkovs and his more than 130-member team develop the tool entirely in-house following three years of R&D.
Volkovs, who joined TD eight years ago, recently oversaw consolidation of myriad single-purpose predictive models into AI Prism to better anticipate customers' needs.
"Operating one foundation model instead of dozens of individual predictive models allows us to more efficiently implement the standards and guardrails set by our 'Trustworthy AI' team," Volkovs said, "and creates the opportunity for that team to centralize their model-monitoring processes."
AI Prism has resulted in 20% to 30% improvement in predictive accuracy and 53% improvement in lead quality, according to Volkovs. AI Prism sharpens lead quality by targeting only those customers who'd be receptive to marketing offers, The result thus far: a more than 40% lift in response rates for marketing campaigns.
Including AI Prism, which launched last year, TD extracted about $170 million in value from the bank's AI initiatives in 2025. Volkovs has been central to delivering those gains.
In 2018, TD acquired Layer 6, a Toronto-based company co-founded by Volkovs that developed an AI prediction engine for enterprise data.
Volkovs directs R&D at Layer 6, whose Toronto and New York City offices function as TD's AI center of excellence. He holds inventor or co-inventor credit on more than 300 patents and has published more than 50 research papers.
Outside TD, Volkovs advocates for responsible use of enterprise AI through partnerships with the likes of MIT, Columbia University, the Fields Institute, the Vector Institute, Mila - Quebec Artificial Intelligence Institute and the University of Toronto, where he earned a doctoral degree in machine learning.
All of TD's digital transformation efforts, including the more than 20 launched in the past 12 months, rest on a foundation of privacy, security, reliability, accuracy and trust. For instance, AI Prism lives in TD's secure cloud, with no customer data exposed to third-party vendors. More broadly, TD's digital initiatives build in human oversight to improve accuracy, decrease risk and cut costs, according to Volkovs.
"For our virtual assistants, for instance, a colleague queries a model and must verify the result before acting on that response for a client," he said. "This human-in-the-loop approach is vital to manage the risks and ethical issues that could arise with these systems, mitigating bias and providing feedback to build trust in them."
